You are on page 1of 23

This is an open access article published under an ACS AuthorChoice License, which permits

copying and redistribution of the article or any adaptations for non-commercial purposes.

Review
www.acsnano.org

Pathways and Challenges for Biomimetic


Desalination Membranes with Sub-Nanometer
Channels
Cassandra J. Porter, Jay R. Werber,* Mingjiang Zhong, Corey J. Wilson, and Menachem Elimelech*
Cite This: ACS Nano 2020, 14, 10894−10916 Read Online
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

ACCESS Metrics & More Article Recommendations *


sı Supporting Information

ABSTRACT: Transmembrane protein channels, including ion


Downloaded via 65.35.165.247 on November 11, 2020 at 18:30:25 (UTC).

channels and aquaporins that are responsible for fast and


selective transport of water, have inspired membrane scientists
to exploit and mimic their performance in membrane
technologies. These biomimetic membranes comprise discrete
nanochannels aligned within amphiphilic matrices on a robust
support. While biological components have been used directly,
extensive work has also been conducted to produce stable
synthetic mimics of protein channels and lipid bilayers.
However, the experimental performance of biomimetic
membranes remains far below that of biological membranes.
In this review, we critically assess the status and potential of
biomimetic desalination membranes. We first review channel
chemistries and their transport behavior, identifying key characteristics to optimize water permeability and salt rejection. We
compare various channel types within an industrial context, considering transport performance, processability, and stability.
Through a re-examination of previous vesicular stopped-flow studies, we demonstrate that incorrect permeability equations
result in an overestimation of the water permeability of nanochannels. We find in particular that the most optimized
aquaporin-bearing bilayer had a pure water permeability of 2.1 L m−2 h−1 bar−1, which is comparable to that of current state-
of-the-art polymeric desalination membranes. Through a quantitative assessment of biomimetic membrane formats, we
analytically show that formats incorporating intact vesicles offer minimal benefit, whereas planar biomimetic selective layers
could allow for dramatically improved salt rejections. We then show that the persistence of nanoscale defects explains
observed subpar performance. We conclude with a discussion on optimal strategies for minimizing these defects, which could
enable breakthrough performance.
KEYWORDS: biomimetic membrane, molecular sieving, reverse osmosis, desalination, aquaporin, carbon nanotube, water purification,
nanochannel, vesicle, stopped-flow

A s stressors like population growth, industrialization, and


climate change threaten to deplete and contaminate our
freshwater resources, larger bodies of saline water could
provide a vast supply of water for drinking, agricultural, and
industrial use.1 However, desalination of these waters requires
energy requirements for the RO stage have drastically reduced
from ∼15 kWh m−3 using the original cellulose acetate
membranes of the 1970s down to only ∼2 kWh m−3, only
∼25% above the practical minimum energy.2
Despite the substantial reduction in energy consumption and
more energy and financial resources than traditional freshwater
overall cost, seawater RO still has room for improvement. While
purification methods.2 Currently, the state-of-the-art technology
for desalination is reverse osmosis (RO) using thin-film- current water permeabilities enable near-optimal performance,
composite (TFC) polyamide membranes.3,4 Fully aromatic
TFC-RO membranes are readily produced at industrial scale Received: July 10, 2020
through interfacial polymerization, whereby a rapid reaction Accepted: September 4, 2020
occurs at the interface of immiscible organic and aqueous phases Published: September 4, 2020
to form a highly cross-linked polyamide selective layer on a
porous support5 (Figure 1). With the advent of the TFC-RO
membrane and energy recovery devices, seawater desalination

© 2020 American Chemical Society https://dx.doi.org/10.1021/acsnano.0c05753


10894 ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

Figure 1. Transition in desalination research from focusing on dense polymers that reject salt by a solution-diffusion mechanism to considering
sub-nanometer channels capable of molecularly sieving ions. In the solution-diffusion panel (left), common reactants for TFC-RO membranes
are represented, which rapidly react at an organic−aqueous interface during interfacial polymerization to form a cross-linked, fully aromatic
polyamide selective layer with characteristic ridge-valley morphology. Salt rejection determined by a solution-diffusion mechanism results from
the higher partitioning and/or diffusion rates of water over ions. In the ion sieving panel (right), common molecular sieves that have been
considered for desalination are shown, with ideal water pathways illustrated. In pores similar in size to water, single-file water transport is
induced. Nanotubes and nanochannels can be synthetic (e.g., carbon nanotubes) or biological (e.g., aquaporins). To produce nanoporous
sheets, sub-nanometer pores where only a few atoms are vacant have been etched in single-layer graphene using chemical oxidation, electron
beam irradiation, doping, and ion bombardment.12 For 2D laminates, the water pathway is through interlayer spaces between sheets. Studies so
far have primarily considered graphene oxide nanosheets for 2D laminates.13,14 The molecular sieving mechanism for ion rejection is by size
exclusion, where highly uniform pores exclude larger solutes and ideally transport only molecules similar in size to water.

increased water-solute selectivity would allow for reduced far exceed salt permeability during solution diffusion, as it does
operational costs, improved reliability and efficiency, and for polyamide, historical data suggest that it will be difficult to
enhanced product water quality.4 For example, TFC-RO significantly advance performance with polymeric systems.
membranes inadequately retain chloride and some small neutral Commercial desalination and water purification membranes
solutes, such as boron in seawater desalination and trace organic typically exhibit a permeability−selectivity trade-off, similar to
contaminants in wastewater reuse, necessitating extra purifica- the Robeson plot for polymeric gas separations.8−11 Further-
tion steps which increase the cost of desalination.4 more, despite many decades of extensive research, no polymeric
Transport through the polyamide layer is well described by material has yet surpassed the desalination performance (i.e.,
the solution-diffusion model, in which permeants (i.e., water and water permeability, water-salt selectivity, and cost-effectiveness)
solutes) partition into the dense polyamide layer and diffuse of fully aromatic polyamide.
through it (Figure 1).6 The resultant permselectivity of the To overcome the limitations of the solution-diffusion-based
membrane is attributed to differences in abilities and rates of polyamide membranes, research focus has shifted toward the
species to dissolve into and diffuse through the polyamide development of desalination membranes that remove solutes via
membrane material.7 Although intrinsic water permeability can molecular sieving. In this mechanism of ion rejection, highly
10895 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

uniform, rigid pores that are smaller than the diameter of istics of the aquaporin that explain its ultraselectivity and fast
hydrated salt ions transport water and nearly completely reject water transport, comparing this biological channel to several
ions by size exclusion (Figure 1). Recent formats of molecular synthetic channels and placing each in industrial context. Using
sieves considered for desalination include nanotubes and corrected analysis of reported permeability measurements, we
nanochannels, two-dimensional (2D) laminates, and nano- then show that the water permeabilities of many channels have
porous sheets (Figure 1).14 However, these top-down efforts been overestimated. Subsequently, we predict best-case-
have failed so far to achieve adequate salt rejection due to the scenario outcomes for common biomimetic formats, including
persistence of defects coupled with the daunting challenges of membranes with intact vesicles and membranes with planar
tuning interlayer spacing or pore size.12,13,15 Biomimetic biomimetic layers. For the more promising planar format, the
membranes, or composites comprising an amphiphilic matrix biggest challenge is the presence of nanoscale defects. Through
with discrete, aligned nanochannels on a robust support, may mathematical models, we estimate the defect density for several
provide a platform for industrial-scale molecular sieves that reported biomimetic membranes. We then discuss synthesis
overcome the limitations of solution-diffusion-based polyamide pathways that could limit both the presence of defects and the
membranes. effect of defects on transport performance. We conclude with a
After over 3.5 billion years of evolution,16,17 the cell discussion on the practicality of biomimetic desalination
membranes of modern organisms can perform an array of membranes and how to best exploit the strengths of discrete
highly complicated functions, which rely on a system of complex nanochannels as molecular sieves in other applications beyond
transmembrane proteins aligned within the amphiphilic lipid desalination.
bilayer. In this system, water and only select ions pass through
channel pores and pumps, depending on the energy and nutrient TRANSPORT BEHAVIOR OF BIOMIMETIC
needs of the cell.18,19 In pioneering work, Preston et al. DESALINATION MEMBRANES
determined that an integral membrane protein formed a
Transport through the Amphiphilic Matrix. The key
biological channel that selectively transports water in and out
components for transport in biological membranes are the
of many types of cells. This protein was the CHannel-forming
selective transmembrane proteins and the amphiphilic matrix
Integral Protein of 28 kDa (CHIP28),20,21 later called the
(i.e., the lipid bilayer) in which these proteins assemble. For
aquaporin. For these discoveries, Peter Agre and Roderick
industrial biomimetic desalination membranes, these compo-
MacKinnon shared the 2003 Nobel Prize in Chemistry.
nents need to cooperate to reject salt and transport only water.
Through additional biophysical studies, ion channels also
Molecular transport should predominantly occur through the
showed impressive selectivity, inspiring the design of synthetic
incorporated channels that reject solutes by molecular sieving.
ion channels.22−25 Eventually, researchers realized the potential
Meanwhile, salt flux should be minimized through the
implications of these channels for industrial-scale water
surrounding amphiphilic matrix. The most promising matrices
purification, especially aquaporin in the use of desalination,
in biomimetic membranes comprise amphiphilic block copoly-
and attempted to produce biomimetic membranes, or materials
mers rather than lipids, with the resulting matrix enabling the
that mimic the structure and performance of biological
alignment of the nanochannels while providing mechanical and
membranes.26−35 While much of the work has focused on chemical robustness.
water-solute separations, the biomimetic membrane format also Molecular transport through both amphiphilic matrices and
presents opportunities to develop membranes with tunable dense polymer membranes (e.g., polyamide selective layers) is
selectivity based on a chosen channel type. described by the solution-diffusion mechanism.43 In this
However, translating biological mechanisms into industrial- mechanism, species dissolve or partition into the membrane
scale technology necessitates scale-up by orders of magnitude matrix, diffuse in the direction of decreasing chemical potential,
from the microscopic size of a cell membrane to tens of square and desorb into the permeate stream. Water flux, Jw, for dense
meters.36 For industrial relevance, the synthesis of a biomimetic polymeric membranes is described by44
membrane would need to be cost-effective and simple.
Simultaneously, such a membrane would need to be Jw = A(ΔP − Δπm) (1)
mechanically stable under RO pressures exceeding 70 bar and
chemically stable during repeated membrane cleaning and where A is the water permeability coefficient, ΔP is the applied
usage.37 Notably, even at the lab scale, sufficiently high-salt hydraulic pressure, and Δπm is the osmotic pressure difference
rejection has not yet been achieved for biomimetic desalination across the membrane.
membranes after over a decade of research.37,38 Therefore, Rates of both dissolution and diffusion affect overall water and
channels, selective layer formats, synthesis strategies, and salt permeabilities, Pw and Ps. For materials with low-water
support layer types must be carefully considered to attain the content, permeability P (for both water and solute) can be
capabilities of this technology. While certain aspects of approximated as7,43,45
biomimetic desalination membranes have been reviewed
recently,37−42 a critical analysis of their performance and their K pD
potential application in water-treatment processes remains P=
d (2)
necessary.
In this critical review, we examine efforts toward biomimetic where Kp is the partition coefficient, D is the diffusivity, and d is
desalination membranes for water purification in order to the membrane thickness. The thickness-dependent permeability
identify the best strategies to realize their full potential for both as defined here and in the biophysical literature43 is often called
desalination and solute−solute selectivity. We first examine permeance in industrial membrane literature.7,45 Polyamide and
molecular transport, contrasting solution-diffusion with molec- other glassy polymers predominantly have diffusion selectivity
ular sieving and assessing transport through the mixed matrix of for aqueous separations, whereby differences in D result in
biomimetic selective layers. We next identify the key character- selectivity (e.g., between water and ions). Bilayers predominately
10896 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

have solubility selectivity, whereby differences in Kp result in where vw is the volume of one water molecule and z is the
selectivity.43 distance between each water molecule. Internal and entrance
The water permeability coefficient A is related to the water resistances are incorporated into the effective diffusivity, as
permeability Pw by45 resistances from pore-lining chemical functionalities and
constriction zones cannot be decoupled for a water column.
PwVw Additionally, these resistances are currently unpredictable a
A=
̂
RT (3) priori, requiring molecular dynamics (MD) simulations to
resolve the overall Dw. The single-channel water permeability
where Vw is the molar volume of water, R̂ is the ideal gas can be converted to the channel water permeability coefficient,
constant, and T is the absolute temperature. Solute flux Js is Ac, by dividing by the overall channel projected area (i.e., the area
related to diffusive salt permeability by the Fickian diffusion on the basis of the outer and not inner pore diameter) and
equation:45
subsequently using eq 3.
Js = BΔCm (4)
Overall Transport Through the Biomimetic Mem-
brane. For a high-rejection membrane with relatively large
where B is the salt permeability coefficient (equivalent to Ps) and water flux, the concentration and osmotic pressure changes
ΔCm is the solute concentration difference across the across the membrane are not accurately described by the
membrane. Together, water and solute fluxes under pertinent difference in concentration between permeate and bulk feed due
operating conditions determine the observed solute concen- to the development of an unstirred, more-concentrated layer on
tration in the permeate, Cp, which in turn is related to the the feed side near the membrane surface. This concentration
observed solute rejection, Ro, by polarization effectively increases the osmotic pressure and
depends on membrane permeability as well as operational
Js /Jw Cp factors like applied driving force (typically hydraulic pressure),
Ro = 1 − =1−
Cb Cb (5) crossflow rate, and feed concentration. Film theory is commonly
used to account for concentration polarization for isotropic
where Cb is the bulk feedwater concentration of solute. membranes when using eqs 1 and 4.44 In the case of discrete,
Transport through Channels. A nanochannel at approx- highly permeable, and selective nanochannels that are spaced
imately the size of a water molecule could completely reject relatively far apart within an amphiphilic matrix, an unstirred
essentially all ions by molecular sieving. If these selective pores hemisphere surrounding the entrance of each nanochannel on
also have sufficient water permeability, typically involving single- the feed side forms.52 However, this unstirred hemisphere is
file water transport, they could drastically increase the overall Jw predicted to negligibly exacerbate concentration polarization
of a biomimetic layer without affecting Js, thereby reaching salt due to a sufficiently fast rate of water diffusion from surrounding
rejections above those attainable by current desalination regions of the bulk feed, resulting in similar concentration
membranes (see eq 5). To these aims, it is imperative that polarization to conventional polymeric membranes. For densely
desalinating nanochannels: (i) highly reject ions, (ii) permeate packed nanochannels, these unstirred hemispheres could
water at a sufficiently fast rate, and (iii) load within the converge into a much thicker unstirred layer and require special
amphiphilic matrix at acceptable densities. consideration.52 However, high-density packing is currently
For convective flow through relatively large channels (i.e., >1 atypical for biomimetic layers. In other words, concentration
nm in diameter), the macroscopic rules of hydrodynamics apply, polarization for biomimetic membranes is expected to be similar
whereby the key determinants to flow are channel length and to that of conventional polyamide membranes.
diameter, described by Hagen−Poisueille’s law.46,47 For For both isotropic membranes and molecular sieves, the
narrower nanochannels with single-file water transport, bulk energy needed for transport across the membrane, or activation
water properties and continuum hydrodynamics cannot apply. energy, can be determined using the Arrhenius equation by
Many models envision nanotubes and nanopores as highly varying the temperature during transport. While the activation
slippery cylinders with no internal resistances. In these energy is necessary to estimate permeability at varied temper-
stochastic models, single-file water transport is described by atures, it is also an important metric for understanding the
either the collective diffusion of a water column based on mechanism of flow. The activation energy for water flow through
random walks or by the movement of a vacancy in the opposite nanochannels (∼4 kcal/mol) is typically much less than for
direction of water flow.46,48,49 Both these treatments ignore the water flow through lipid bilayers (∼15 kcal/mol), as most
entrance effects as well as energy barriers often present in nanochannels are relatively static, whereas thermally driven lipid
biological pores, like hydrogen-bonding sites and increased motion is required for molecular diffusion.53,54
resistance through constriction zones within noncylindrical (e.g., To calculate the overall water permeability coefficient of the
hourglass) pores. biomimetic layer, Abm, the channels and amphiphilic matrix are
To overcome these limitations, a collective diffusion model considered resistors in parallel through the summation of their
was proposed in order to find water permeability through each respective water permeability coefficients (Aam and Ac):
channel, Pw,c.50 The most recent of these collective diffusion
models was derived on the basis of the Finkelstein model that Abm = ϕcAc + (1 − ϕc)A am (7)
predicts the work required to move a column of water molecules
(represented as hard spheres) through a pipe of a given length by where ϕc is the channel areal coverage. To achieve an overall
osmotic force.51 In this newest collective diffusion derivation, acceptable water permeability of a biomimetic layer for a given
the water diffusivity, Dw, is related to the water permeability:46 application, the target single-channel permeability and pore-
loading density are contingent on one another, as evident from
Dw vw
Pw,c = eq 7. In other words, a pore with lower water permeability may
z2 (6) be sufficient if channel density is relatively high, and vice versa.
10897 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

Figure 2. Channel types for biomimetic membranes, including biological as well as bioinspired and bioderived. (a) Single channel water
permeability versus pore interior diameter. The pore interior diameter here is defined as the inner diameter of the most constricted region. For
PAH[4], the pore diameter shown refers to the average width of dynamic voids that formed in channel clusters. Channel permeabilities from
stopped-flow data and simulations were adjusted to 25 °C and corrected for any previous errors in permeability calculation (see SI, Section S1
for details). (b) (left) Overhead view of AqpZ tetramer and (right) side view of single AqpZ channel with characteristic hourglass shape. (c)
GramA dimer as it exists in biological and vesicular environments. In organic solvents, the monomers can intertwine to form a parallel or
antiparallel helix. AqpZ and GramA diagrams were drawn using PyMOL69 with protein sequences from the Research Collaboratory for
Structural Bioinformatics Protein Data Bank (RCSB-PDB, https://www.rcsb.org/), PDB-ID codes 1RC270 and 1NRM.71 (d) Dependence of

10898 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

Figure 2. continued

water diffusivity through pores on the number of interior hydrogen-bond points. Data extracted from ref 56. (e) Cyclic peptide nanotubes. (left)
Hydrogen-bonding pattern of pore-forming, stacked cyclic peptides. A polyglycine structure is shown for simplicity; side residues would
typically be present. (right) Modified cyclic peptide with interior peptide-mimicking functional groups. The analogous unmodified cyclic
peptide is radially symmetrical with a fourth primary amine side chain (cyclo[(D-Ala-Lys)4]).72 (f) Single-walled carbon nanotube porins (wide
and narrow) with armchair pattern. Number of carbons approximate those of wCNTP and nCNTP. (g) (left) PAP[5] and (right) PAH[4]
nanochannels with peptide appendages that form interarm hydrogen bonds. (h) (left) Aquafoldamer subunits with “sticky” ends. Differences in
end groups that comprise Aqf1 and Aqf2 subunits are illustrated. (right) Weak hydrogen-bonding pattern of pore-forming, stacked
aquafoldamers. Six subunits are needed to cross a DOPC membrane. (i) Pure water permeability versus total channel areal coverage in a DOPC
bilayer. Single channel permeabilities from (a) were divided by channel cross-sectional area and converted into channel water permeability (A)
coefficients using eq 3. Overall biomimetic layer A coefficients were calculated for various densities of channels within a DOPC bilayer using eq
7. DOPC bilayer hydraulic permeability was taken as 0.15 L m−2 h−1 bar−1.73 The shaded region indicates the water permeability of current
commercial TFC-RO membranes, an adequate range for desalination performance. Permeabilities are listed in Table S1.

KEY MATERIAL COMPONENTS OF BIOMIMETIC inhibit anion transport.63 The aquaporin completely rejects
DESALINATION MEMBRANES protons as well, either by a strong electrostatic repulsion
provided by the NPA motif and the arginine in the constriction
Biological Channels. Biological membranes contain an
array of passive transport channels as well as active zone or by dipole inversion (i.e., the rotation of water molecules)
cotransporters and pumps with varying functions.42,55 The that breaks the water wire;60−63 the exact proton-exclusion
broad range of biological activity occurring across the same mechanisms are currently under debate. Although exclusion of
amphiphilic matrix suggests that a successful biomimetic protons is critical for biological function, it is likely less
membrane could be tuned for specific applications, from important for industrial separations.
containing self-aligned channels that reject nearly all solutes Simulations and mathematical analyses have further eluci-
for desalination to more niche applications requiring solute- dated some special features of the aquaporin, especially
selective transport. The proper channel type must be selected for regarding the influence of channel shape and hydrogen bonding
stability, processability, and performance. during water permeation. A continuum hydrodynamics analysis
For biomimetic desalination membranes, aquaporins such as using finite calculations and subsequent MD simulations
human Aqp1 and bacterial AqpZ that transport only water have demonstrated that hourglass or cone-shaped channels with
received particular interest. Single-channel water permeabilities short constriction regions mitigate the large energy barrier
of 2.7 × 10−13 cm3 s−1 for AqpZ and 4.9 × 10−13 cm3 s−1 for Aqp1 present at the entrance of a water-selective cylindrical pore,
have been measured (Figure 2a).56 Aquaporins typically while small cone angles of around 10−30° reduce resistance
assemble into tetramers, with each individual protein subunit when transitioning from funnel to constriction zones.65,66,68 The
containing a water pore that is formed by several membrane- smaller the channel length, the larger the optimal funnel angle,
spanning alpha helices (Figure 2b).42 Aquaporins have been with the greatest observed angle in nature being ∼27° for
extensively explored for direct use in biomimetic membranes Aqp5.66 For a cylindrical pore that induces single-file water
and have been functionally reconstituted in lipid and block transport, the large entrance effects hinder rapid transport since
copolymer systems. Single channel permeability has been shown each water molecule must simultaneously shed two hydrogen
to be similar in different matrices, although reconstitution bonds.66,68 There are a multitude of water-selective protein
efficiency can vary substantially.57 In addition to serving as a channels with varying numbers of hydrogen-bonding sites at
potential membrane selective agent, aquaporins provide a model their pore mouths and interiors, including aquaporins that
for synthetic channels in terms of both the attainable transport convey only water (e.g., Aqp1 and AqpZ) and aquaporins for
performance and structure−function relationships. Accordingly, water and glycerol (GlpF) as well as peptide-based gramicidin
we will briefly review some of the main distinguishing features of dimers for transporting water and small monovalent cations
aquaporins. (Figure 2c). Notably, the diffusivity of water through each
Aquaporin’s water-solute selectivity with efficient water biological channel is negatively correlated with the number of
permeability is the result of four main attributes: (i) a hydrogen-bonding sites within the channel itself (Figure 2d).56
predominantly hydrophobic interior, (ii) an hourglass shape In terms of pore diameter of aquaporins, the constriction size
tapering to only ∼0.28 nm in diameter at the constriction zone, correlates with greater permeability, with GlpF having markedly
(iii) a series of amino acid residues, asparagine-proline-alanine higher water permeability. However, when comparing channels
(NPA), located before the constriction zone,58−62 and (iv) an of varying chemistries, there is no correlation between
abundance of inward-facing carbonyl groups.63 A channel with constriction size and permeability, as other factors predominate
extreme internal hydrophobicity induces a slip boundary (Figure 2a).
condition where only viscosity is the key limiting factor in Synthetic Nanochannels. Although protein nanochannels
transport rate.64 Moreover, for aquaporin, water is funneled into such as aquaporins can be synthesized using well-established
a single file, gradually reducing the hydrogen-bond number from bioprocessing techniques, these transmembrane proteins can
four to two and dissipating bulk-water viscosity.65,66 At a size easily denature and lose functionality, making production much
around that of water, the constriction zone rejects ions via size costlier than for water-soluble proteins. Additional stability
exclusion.58,67 Meanwhile, several features hinder ion transport challenges will arise during membrane synthesis and use.
via Donnan and dielectric exclusion. Positive poles at the Furthermore, biological channels are not necessarily the most
channel center induced by the two NPA-bearing helices and a optimal in terms of pore-loading efficiency and functionality for
positively charged arginine within the constriction zone prevent industrial processes. Synthetic, bioinspired channels could
cation transport, while several pore-accessible carbonyl oxygens overcome the instability and poor processability of their
10899 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

biological exemplars and provide a greater opportunity to tune linking a defined number of benzenediol units with methylene
performance for target applications. bridges. In the first example of pillararene water channels, central
The first generation of self-aligning, synthetic bioinspired rings with five hydroquinone units were used with hydrazide or
channels were cyclic peptide nanotubes (CPNs), or rings of peptide-appended arms, which were stabilized by interarm
amino acids designed in flat, stackable units that can crystallize hydrogen bonds to form a tubular structure.80,81 The peptide-
into nanotubes hundreds of nanometers long with an internal appended pillar[5]arene (PAP[5]; Figure 2g) exhibited a sharp
pore diameter of ∼1 nm (Figure 2e).72,74 CPN units can in situ molecular weight cutoff (MWCO) of ∼500 Da and a relatively
self-assemble into a channel within an amphiphilic matrix.75 low single-channel water permeability of 4.6 × 10−15 cm3 s−1
Bottom-up, in situ assembly helps avoid issues with physical (Figure 2a). PAP[5]s assembled into dense clusters of parallel
mismatch that occurs when lengths of the overall channel and channels, increasing the channel pore packing density. However,
hydrophobic/hydrophilic regions do not align with the ions readily permeated PAP[5] owing to its inner pore diameter
corresponding surrounding matrix lengths.41 While typical of 0.5 nm.
CPNs transport ions, the pore size can be reduced using a In the aim of achieving salt rejection, peptide-appended
non-natural aromatic amino acid, thereby inhibiting ion passage. hybrid[4]arenes (PAH[4]s) were developed to yield a smaller
Simulations have shown that manipulation of the interior-facing central constriction (Figure 2g). The stackable subunit
functional group can tune the single-channel water permeability, comprises an arene of alternating recorcinol and catechol
reaching a maximum of 1.3 × 10−13 cm3 s−1 (Figure 2a,e).72,76 subunits (2 each, 4 subunits total), linked alternately at the meta
Recently, it was demonstrated that carbon nanotubes can be and ortho positions, with 8 peptide appendages.82 Similar to
cleaved by ultrasonication to lengths of ∼10 nm, after which PAP[5]s, PAH[4]s self-assemble to form dense clusters within
these “carbon nanotube porins” (CNTP) self-align within amphiphilic matrices. The central constriction of PAH[4] is
amphiphilic matrices due to a predominantly hydrophobic relatively impermeable; however, once self-assembled into
exterior with hydrophilic oxygen functionalities lining the ends. clusters, lateral windows provide dynamic voids of 0.2−0.4 nm
Narrow carbon nanotube porins (nCNTP) (Figure 2f) of ∼0.7 width (the 0.5−0.7 nm widths reported in the study referred to
nm internal diameter exhibit a water permeability of 3.7 × 10−13 distances between atomic centers), allowing for interconnected
cm3 s−1, while wider carbon nanotubes (wCNTP) of a similar water-wire pathways to hop between channels. The resulting
length but with a larger pore diameter of ∼1 nm had lower equivalent “per channel” permeability of PAH[4] is 2.7 × 10−14
single-channel water permeabilities of 1.9 × 10−14 cm3 s−1 cm3 s−1.82 Sodium permeation through the channels was
(Figure 2a). These permeability values and several others in undetectable, while chloride permeability was measurable but
this section were adjusted to account for a fitting error low. Based on reported chloride and water permeabilities,
commonly described in biomimetic membrane literature, pristine biomimetic membranes incorporating PAH[4] could
which will be discussed later in detail. The significant increase theoretically achieve water-salt selectivity of at least ∼108, far
in permeability for a narrower nanotube porin indicates the exceeding the current selectivity of desalination membranes of
reduction of water−water hydrogen bonding even within pores 104−105.83
well above the diameter of water. On the basis of measured water Recently developed aquafoldamer-based synthetic water
and sodium-chloride permeabilities,54,77 nCNTPs are predicted channels also have promising performance (Figure 2h).84 In
to have a water-salt selectivity (Pw/PNaCl) on the order of 105, these aquapores, the helical, crescent-shaped oligomer subunits
resulting in salt rejections >99%. Although CNTPs are more have chain-end functionalities which intermolecularly hydrogen
stable and processable than biological pores, it remains a major bond, assembling into stacked, one-dimensional nanotubes.
challenge to produce large quantities of narrow carbon Two versions of these aquafoldamer-based water channels were
nanotubes with uniform pore diameter. created (Aqf1 and Aqf2), both comprising pyridine-based
Other artificial channels of recent interest are imidazole- aromatic polyamide chains beginning with an ester and
derived quartet channels (Imdzl) that form planar supra- terminating with a benzene ring.81 This pyridine chain is similar
molecular matrices with regularly spaced pores.78,79 These in composition to fully aromatic polyamide, but with the
channels are highly selective for water, forming connected replacement of one methine group in each aromatic ring with a
single-file water wires. Within polymerization solutions and nitrogen. The nitrogen atom forms hydrogen bonds with
casting dopants, they may highly cluster to form a planar matrix surrounding amide hydrogens to promote a helical curl (Figure
with pores without needing the assistance of surrounding 2h). Aqf1 differs from Aqf2 in that it has an additional oxygen
amphiphiles for alignment. However, with a reported pure water and methylene unit before its terminus, giving Aqf1 a rounded
single-channel permeability of at best only around 1.8 × 10−16 pore with a diameter of 0.28 nm. Meanwhile, Aqf2 is more
cm3 s−1 for Imdzl with left-hand chirality ((S)-Imdzl) cuboidal at 0.21 × 0.27 nm internally. With these small pore
(Figure2a),78 a selective layer solely comprising these channels constrictions, Aqf1 and Aqf2 both transport water in single file
would exhibit a permeability of only ∼0.16 L m−2 h−1 bar−1. This and completely reject ions. Under osmotic gradient, Aqf1
transport rate is only slightly higher than that observed for water exhibited a higher single-channel permeability of 3.0 × 10−14
passing through a 1,2-dioleoyl-sn-glycero-3-phosphatidylcholine cm3 s−1 compared to 2.2 × 10−15 cm3 s−1 for Aqf2 (Figure 2a).
(DOPC) lipid bilayer (0.15 L m−2 h−1 bar−1) and roughly an The more hydrophobic, slippery interior of Aqf1 likely explains
order of magnitude below necessary to approach minimal this phenomenon, where the hydrogen-bonding energy of water
desalination energy.4,73 Highly polar imidazole subunits with with the pore walls was found to be 15.1 kcal mol−1 for Aqf1 and
secondary amines provide a large number of hydrogen-bonding 16.1 kcal mol−1 for Aqf2. These channels have only been
sites, which again is inversely correlated with water permeability incorporated in lipid bilayers; hence, self-assembly and stability
(Figure 2d). in block copolymer matrices are to be determined.
In contrast, pillararene-based channels provide a hydrophobic When maximizing pore loading, the permeability of channels
interior to allow for relatively fast water transport. The central normalized by their cross-sectional areas should be considered.
constriction in these channels is the pillararene ring, formed by That is, a channel with a relatively high flow rate through its pore
10900 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

Figure 3. Design considerations for common biomimetic membrane formation strategies. (a) Biomimetic compositions for the amphiphilic
channel insertion matrix. (b) Common nanochannels for desalination, from biological aquaporin and gramicidin to bioinspired channels. All
channels self-align within the channel insertion matrix due to matching amphiphilic pattern. (c) Vesicle-encapsulated mixed-matrix formats,
whereby permeate passes through two layers of a channel-laden matrix in intact vesicles. Lipid or polymer vesicles are either imprinted or fully
encased by nonporous polymer, typically a polyamide formed through interfacial polymerization. (d) Pore-spanning planar layer format,
whereby the solution passes through one layer of channel-laden lipid or polymer matrix. Planar layers are classically formed by rupturing
channel-laden vesicles. Note that vesicles are precursors for all biomimetic membrane formats in (c,d).

may have such a large outer diameter that obtaining a reasonable methods and materials used and the overall format of the
overall permeability would require a channel areal coverage that biomimetic membrane.
would destabilize the amphiphilic matrix. For instance, through In general, there are two types of molecules that self-assemble
simulations, minigramicidin A (MiniGram), a gramicidin to form channel-incorporating matrices: amphiphilic lipids and
channel with the same diameter as biological gramicidin A block copolymers (Figure 3a). Typical lipids used include:
(GramA) but with 8 fewer amino acids in length, has a natural lipids like Escherichia coli phospholipid,53,88,89 chicken
permeability slightly less than that of AqpZ (Figure 2a).85 egg phosphatidylcholine,57,78,79 or porcine brain phosphatidyl-
However, MiniGram has an outer diameter of only 0.96 nm per serine57,79 as well as synthetically produced lipids like
channel, whereas each AqpZ subunit is approximately a 3 nm × 3 DOPC,28,29,32,35,90 1,2-dioleoyl-3-trimethylammonium-pro-
nm square (∼9 nm2 per channel).86,87 As a result, to reach the pane (DOTAP),30,32,35 or 1,2-dimyristoyl-sn-glycero-3-phos-
typical TFC water permeability of 2−3 L m−2 h−1 bar−1, levels phocholine (DMPC).91 Polymer matrices may be formed from
that allow for near-optimal operation of desalination processes, diblock or triblock copolymers, most frequently consisting of
MiniGram would need a channel areal coverage of <0.25%, poly(2-methyl-2-oxazoline) (PMOXA) as the hydrophilic block
whereas AqpZ would need a coverage of 2−4% (Figure 2i). As and poly(dimethylsiloxane) (PDMS) as the hydrophobic
another example, PAH[4] has a relatively low single channel block26,27,33,34,92,93 but sometimes comprising poly(1,2-buta-
permeability (Figure 2a), but only occupies ∼2 nm2 per channel. diene) and poly(ethylene oxide) (PB-b-PEO).57,73 These
Accordingly, just 5−8% areal coverage with PAH[4] would be particular polymer systems are not necessarily unique but
needed to reach the target water permeability levels (Figure2i). nonetheless have been extensively explored, although the higher
The propensity of PAH[4] to form clusters may facilitate fluidity of PDMS over other hydrophobic polymers allows for
achieving this relatively high pore density without destabilizing better channel insertion efficiency without denaturing chan-
the biomimetic matrix. nels.94 Amphiphilic polymer matrices are generally more stable
Common Biomimetic Selective Layer Formats. Pro- than lipid bilayers; lipid bilayers can be disrupted by
gress has been made in the development of selective water perturbations as mild as air bubbles. As such, while lipids are
channels, particularly synthetic channels that may offer useful in model systems for preliminary explorations, they have
substantial advantages in terms of processability and cost. no apparent use in industrial applications. For the best channel
Currently, biological and synthetic channels are available that insertion efficiency, the matrix must reasonably match the
could allow for the development of ultraselective biomimetic physical dimensions of the channel, although bilayer thickness
membranes, in principle. However, the performance of a has been shown to conform to channel heights for some
biomimetic selective layer is dependent not only on the chosen systems.95 Additionally, the matrix must chemically match the
water channel but also to a large extent on the matrix in which it degree of hydrophobicity/hydrophilicity of the channel
is incorporated. Therefore, it is imperative to consider the exterior.57
10901 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

With a channel type and matrix selected, the synthesis strategy the areal coverage of the support, namely the minimization of
and overall format will critically impact the performance of the defects in the planar selective layer. The addition of attractive
biomimetic membrane. Channel properties are typically functional groups on the support and/or vesicle outer
assessed within model bilayers, usually using lipids. These surface31−35 and the determination of an optimal rupturing
bilayers are produced as spherical, nanoscale liposomes through pressure31 have improved planar coverage. However, sodium
the rehydration and extrusion of lipid layers or as planar formats chloride rejection achieved in this format for AqpZ-incorporated
spanning across small apertures, formed by painting lipid/ membranes is still only in the range of 20−75%, suggesting full
solvent solutions.96−100 For industrial-scale applications, a coverage has still not yet been achieved. The highest reported
biomimetic active layer would need to be produced on a more salt rejection for a planar biomimetic layer was 97% for a
stable support. The most common formats of biomimetic gramicidin-laden selective layer.31 In this case, a small
membranes consist of two general strategies: (i) intact vesicles proportion of cationic lipid 1,2-dimyristoyl-3-trimethylammo-
encapsulated in a polymeric matrix (Figure 3c)26−30 and (ii) nium propane was incorporated into the liposome to aid in
pore-spanning planar biomimetic layers (Figure 3d).31−35 In rupturing on a sulfonated poly(ether sulfone) NF membrane,
both methods, channel-laden vesicles are precursors, as this and an optimized liposome rupturing pressure was determined.
nanostructure is relatively straightforward to synthesize, self- The nonporous nature and relatively high salt rejection of the
assembling in water based on amphiphile−solvent interaction. NF support, at 60%, likely also helped minimize the impact of
In aqueous solutions, the amphiphilic biomimetic matrix the defects. In subsequent sections, we will further discuss the
thermodynamically favors a sphere to minimize interactions of typical selective layer defect coverage and its impacts.
unlike phases. In the development of these formats, aquaporins Other Biomimetic Formats. Some alternative biomimetic
have most frequently been used because they show relatively membrane formats have been explored, aiming to overcome
high stability in both polymer and lipid matrices, coupled with issues with channel-insertion efficiency and incomplete selective
promising transport properties for desalination.26−35 layer coverage (Figure 4). In general, many of these ideas are
For mixed-matrices incorporating intact vesicles, aquaporin- based on lamination of multiple layers with the notion that if one
laden liposomes or polymersomes are first deposited on a porous layer does not fully cover the support, then multiple should.
support. Frequently, cross-linkable polymers are incorporated to At a particular channel-to-polymer ratio, block copolymers
stabilize these intact vesicles, especially for pressure-assisted and channels can self-assemble into crystalline planar sheets
deposition.27,30 End functionalities on the vesicles can often aid during slow dilution of detergent or solvent, after which the
in bonding the vesicle with the support membrane.26,27,30 The sheets can be stacked to form 2D laminates. Carboxylic-
highest deposition of vesicles included strategies for both cross- functionalized PAP[5] channels were incorporated into flat
linking the vesicular matrix and bonding the vesicle to the crystalline sheets of PB-b-PEO diblock copolymers as
support, resulting in ∼35% surface coverage by vesicles after aggregated microphases (Figure 3a). Because the PAP[5]s
incubation.30 Because vesicle coverages are not typically included anionic carboxy groups, the micrometer-sized sheets
reported, we have estimated the coverage based on published could be laminated onto a poly(ether sulfone) microfiltration
microscopy images (Figure S1). support by a layer-by-layer (LbL) technique involving the
After immobilizing vesicles, they are surrounded with cationic polyelectrolyte polyethylenimine. In general, LbL
(imprinted)26,27 or fully covered in (encased)28−30 a nonporous consists of depositing layers of alternately charged polyelec-
polymer. In some cases, the polymeric matrix surrounding the trolytes that adhere via electrostatic attraction. Only four
vesicles is a polyamide formed via interfacial polymerization,28,29 lamination cycles were needed for full coverage, with a sharp
but other polymers have been used as final coating layers. For MWCO of ∼450 Da and water permeability of around 65 L m−2
example, alternately deposited layers of dopamine and histidine h−1 bar−1, which is 3-fold greater than industry-leading
were used to imprint vesicles.26 In another study, vesicle- composite NF membranes.
immobilized membranes were dip-coated in initiators and then The production of these laminated crystalline sheets can take
monomers of methyl methacrylate and ethylene glycol up to 6 days by way of a slow dilution of detergents. In a
dimethacrylate to promote redox free-radical polymerization.27 subsequent study, a solvent method reduced the synthesis time
Additional methods have involved the immobilization of DOPC of nanochannel-packed sheets to only 2 h. In this study, other β-
vesicles with PEO conjugates and cross-linking the vesicles with barrel membrane protein nanochannels were incorporated, each
glutaraldehyde to each other as well as a dopamine-coated showing sharp and channel-dependent MWCOs with signifi-
support.30 In areas with vesicles, water would theoretically pass cantly higher water permeability than some commercial NF
through two biomimetic selective layers (i.e., into and out of the membranes.101 While these results are exciting, processability
vesicle) before reaching the support (Figure 3c). with this approach may be an issue. Additionally, applicability of
The degree to which the addition of vesicles improves the this biomimetic format to RO-type separations is unclear, as the
performance of nonporous polymers, especially when consider- channels used thus far do not inherently reject salt. As with
ing the highest-performing polyamide material, hinges on the graphene oxide and MoS2 2D laminates, we can expect that the
areal coverage of deposited vesicles. As we will later show, rejection performance of any laminated membraneeven when
maximized vesicle coverage is requisite for any improvement in incorporating salt-excluding channelswill be limited by
salt rejection over pure polyamide. However, literature suggests interconnected framework defects.13,102 Through these defects,
achieving this level of vesicle coverage is not trivial. We will later water and solutes can bypass the biomimetic layer, thereby
discuss in-depth the limitations of this approach. hindering overall water-salt selectivity. Such defects would likely
For planar formats where the biomimetic active layer spans be particularly problematic for small solutes such as sodium and
across a porous support, channel-laden vesicles are typically chloride ions that readily permeate through interlayer materials
ruptured on the support. Transport would then occur through (e.g., swollen cationic polyelectrolytes).
just one biomimetic selective layer (Figure 3d).31−35 The Another method to promote multilayering for full coverage of
performance of a membrane in this configuration depends on the active layer relied on a revision of the matrix composition. A
10902 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

with GramA and PAP[5] increased the water permeability.


Membranes with PAP[5] again exhibited a MWCO of ∼450 Da.
However, for membranes with GramA, the permselectivity of
sodium over chloride was only 0.34, whereas in nature, GramA is
highly selective for monovalent cations over anions,104
suggesting the presence of nanoscale defects. Multilamellar
morphologies such as those used in these studies could
stochastically minimize the overall defect content, but at the
cost of decreased permeability due to the increased selective-
layer thickness.
Biological and synthetic water channels have also been used in
mixed-matrix membranes, whereby channels were directly
incorporated in polyamide films (Figure 4c).105−107 Although
this format lacks the biomimetic channel-aligning matrix, it
resembles biomimetic membranes in that discrete biomimetic
channels are incorporated within a matrix of dissimilar chemical
composition. AqpZ has been protected by complexation with a
β-barrel-forming amphiphilic peptide before incorporation into
polyamide, enhancing permeability.106 The incorporation of
zwitterion-functionalized carbon nanotubes with an average
diameter of 1.5 nm somewhat enhanced salt rejection and
greatly increased water permeability of a TFC-RO mem-
brane.105 While improved performance over controls has been
reported in experimental studies,105−107 no such mixed-matrix
membrane has achieved better performance than the best-
performing commercial membranes. Additionally, this approach
predominantly relies on the intrinsic transport performance of
the polyamide matrix and is generally limited by the defects that
form at the interface between the channel and the glassy
polyamide material.

PERMEABILITY MEASUREMENTS OF CHANNELS AND


BILAYERS
Biophysical Methods to Measure Membrane Perme-
ability. In classic biophysical studies, methods of quantifying
transport through reconstituted membranes were developed to
understand transport through bilayers and protein channels. For
industrial biomimetic membranes, these techniques are essential
Figure 4. Alternative biomimetic membrane formats. (a) Laminated for optimizing membrane design. By studying the bilayer alone,
crystalline sheets, where sheets of up to ∼1 μm in side-length are channel loading efficiency can be optimized, where degree of
layered for full coverage. (b) Co-assembly of multilayer lamellar hydrophobicity and length of each type of block should be
block copolymer. Biological membranes follow an ABA (hydro- matched between channel exteriors and amphiphilic channel-
philic−hydrophobic−hydrophilic) pattern, whereas, here, a BAB insertion matrix.57 For testing the performance of the channel-
(hydrophobic−hydrophilic−hydrophobic) block copolymer is used laden matrix, two general methods have been utilized: (i)
to promote multilayered self-assembly. (c) Blended mixed matrix, aperture-spanning patch clamp and (ii) vesicle stopped-flow
where channels are blended into a polymer solution before casting studies.
or polymerization of a nonporous polymer, typically polyamide The state-of-the-art system for measuring ion conductance
formed through interfacial polymerization. Random alignment of
nanochannels occurs.
across a membrane is the patch clamp. To produce a patch
clamp, a pipet tip is touched to the surface of a vesicle or cell
membrane, and a light suction is applied to form a gigaohm-
cross-linkable triblock copolymer of polyisoprene-b-poly- strength seal. Beyond this point, by adjusting the strength of
(ethylene oxide)-b-polyisoprene was spin-coated onto a silicon pulses of suction or voltage and contact with air, the system can
wafer for self-assembly into well-aligned, multilayered lamellae be setup to attach a full cell/vesicle or a small patch. Extensive
laying parallel to the surface, confirmed through grazing- relationships between current and voltage changes and single
incidence small-angle X-ray scattering.103 These lamellae channel transport rates have been developed for apertures
could then be transferred to a porous support, in this case containing one channel108−114 and those containing multiple
anodized aluminum oxide (Figure 4b). In this system, the BAB channels.115,116 However, in the case of developing a
pattern (i.e., hydrophobic−hydrophilic−hydrophobic) was desalination membrane, typical channels should reject all
utilized to promote and stabilize multilayer formation by solutes, prohibiting the measurement of ion conduction across
fostering cross-linking (polymer bridging) between adjacent their orifices. Furthermore, during transport through aquapor-
hydrophobic domains. Lithium bis(trifluoromethanesulfonyl)- ins, water molecules are flipped so that electron conductivity
imide content in the casting solution was optimized to eliminate through a fully intact water wire cannot be used as a
detectable defects by reducing PEO crystallinity. Co-assemblies confirmation of pore loading. In these cases, methods to
10903 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

Figure 5. Concepts for stopped-flow measurements to predict biomimetic membrane permeability and determine efficiency of channel
insertion into amphiphilic matrix. (a) Measuring signal in a stopped-flow apparatus to determine permeability of biomimetic membrane.
Vesicles are loaded with a buffer consisting of a solute that is impermeable through the vesicle selective layer at small time scales. Once exposed
to hyperosmotic solution, vesicles rapidly shrink via water exiting through nanochannels to equilibrate interior and exterior osmolar
concentrations. Either light side-scattering or self-quenching fluorescent dyes are utilized to determine real-time vesicle volume. (b)
Comparison of water permeability results for matrices with AqpZ using the directly fitted flux eq 8, the commonly misused exponential eq 14,
and the good approximation from the exponential eq 12. Permeabilities in stopped-flow were typically taken at lower temperatures and were
temperature-corrected to 25 °C using an activation energy of 3.3 kcal mol−1, an average value from several studies.53,88,89,92 Permeability data
from references a89, b,28 c,35 d,120 e,90 f,123 g,122 and h.57 (c) Comparison of resultant pure water permeability from stopped-flow studies
conducted at various changes in osmolarity across the vesicles, for DOPC and PMOXA-PDMS-PMOXA (MDM) vesicles. Direct fitting of the
correct flux equations as well as the incorrect exponential approximation and two good exponential approximations are provided. Data taken
from literature.73 (d) An example of how AqpZ-loaded vesicles shrink over time when exposed to hyperosmotic solution. Models based on the
commonly used erroneous single exponential fit as well as the use of a differential equation are presented alongside measured data points. Data
was extracted from ref 89. (e) Example of stopped-flow curves to measure transport rates of relatively permeable solutes through MDM using
self-quenching encapsulated fluorophore. Data taken from ref 73.

quantify the transport of water through small ionic concen- each loaded with multiple channels, can be tracked over a
tration changes close to recontituted membranes under osmotic discrete window of time. In general, a stopped-flow apparatus
stress have proven advantageous. A recent example of these allows for the rapid mixing of small quantities of two different
methods uses ion-selective microelectrodes.117,118 Additionally, solutions and subsequent measurement of optical changes. By
the patch clamp can be used to study the slower diffusive mixing a solution of vesicles with either a hyperosmotic or
permeabilities of water and neutral solutes through amphiphilic hypoosmotic solution and tracking the average osmotically
pure matrices without channels.51,119 However, patch clamping driven volume change with time, the permeability can then be
is not ideal for measuring water permeability since the calculated.96
membrane area is small. After successfully characterizing the function of the aquaporin
Stopped-flow methods for vesicular solutions are commonly in biophysical work,89 stopped-flow methods with vesicles were
used to assess water and neutral-solute permeability. In these heavily used to determine the aquaporin loading efficiency and
methods, the average volume change of a population of vesicles, function in efforts toward biomimetic mem-
10904 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

branes.26−28,30,32−35,53,57,88−93,120−123 More recent efforts have where Cin,S(t) and Cout,S are the intra- and extra-vesicular
paired stopped-flow permeability measurements with techni- concentrations of the permeable solute. In eqs 8 and 9, the
ques such as fluorescence correlation spectroscopy that enable intravesicular concentrations, Cin(t) and Cin,S(t), are relative to
direct measurement of the number of channels per vesicle.56,81 the initial vesicle volume, while the resulting ratios with
Signals from either light side-scattering (scattering at 90°) or normalized vesicle volumeCin(t)/Ṽ (t) and Cin,S(t)/Ṽ (t)
fluorescence from encapsulated, concentration-sensitive fluo- give the absolute osmolarity and permeable solute concen-
rescent dyes are correlated to vesicle volume. With decreased tration, respectively.
vesicular volume, light side-scattering increases, while the Critically, eqs 8 and 9 are algebraic reformulations of eqs 1 and
fluorescence of loaded self-quenching fluorescent dyes decreases 4, respectivelythe fundamental flux equations for membrane
(Figure 5a). Fluorescence methods avoid complications with transport from the solution-diffusion modelwhich have been
light scattering, such as optical artifacts from particle motion translated for vesicles with permeability driven solely by
immediately after mixing and changes in refractive index.124 concentration gradients (i.e., no hydraulic pressure). Accord-
Vesicles can also be used to determine solute permeability, ingly, these permeabilities measured from studies on vesicles
which is especially important for biomimetic membranes provide the expected permeability for a planar, defect-free
targeting solute−solute selectivity and in processes like membrane with the same composition as the vesicular walls (i.e.,
wastewater reuse that aim to eliminate organic, neutral the same amphiphilic material and similar channel density). In
molecules. In one method used for solutes that are less other words, Pw and Ps can be used to estimate water flux and
permeable than water, the permeable solute of interest is loaded solute rejections for a defect-free membrane, using eqs 1−5.
within the vesicles, which are then mixed with an isotonic With modern computer software, it is relatively straightfor-
solution containing a relatively impermeable solute. As the ward to directly solve for Pw and Ps using eqs 8 and 9 by
permeable solute exits, vesicles shrink due to water efflux, which numerical iterative fitting. We emphasize that eqs 8 and 9 are the
can be used to determine the solute permeability.125,126 A fundamental flux equations from the solution-diffusion model,
second method for permeable solutes is to mix vesicles with a and the direct use of these equations is the most preferred and
hyperosmotic solution of the pertinent solute.73 After a relatively most accurate fitting method. However, many early studies used
single exponentials as approximations of eqs 8 and 9.88,130−135
fast exit of water, a more gradual increase in vesicle volume
The observed first-order signal rate constant, ks, has frequently
occurs as the solute and water flow back into the vesicle. In cases
been determined through stopped-flow signal data fitting
where the permeation of a basic or acidic solute is higher than
using27,30,32,34,91,93
water permeability, that is, vesicle volume changes are too slow
to reflect species permeability, a pH-sensitive dye can be loaded S(t ) = A e(−kst ) + B (10)
into vesicles before mixing with solutions of interest. This
technique has been used for permeation of protons and where S(t) is the signal intensity with time, A is the amplitude,
ammonia as well as model carboxylic acids (e.g., acetic, and B is the final signal value. Using exponential fits, an early
propionic, and butyric acids).73,127 approximate solution was136
In cases involving previously unexplored matrix materials or r0ks(1 − b/V0)
synthetic channels, initial vesicle stopped-flow studies can be Pw ≈
3VwCout (11)
used to optimize these materials and channel incorporation
methods. Once the maximum permeability is determined for where r0 is the initial average vesicular radius, b is the per vesicle
these systems, best-case-scenario desalination performance for average volume of the bilayer, and Cout is the initial extravesicular
composite, industrial-scale formats of these same materials can osmolarity. More recently, it was shown that the permeability
be predicted. However, to accurately predict performance, it is can be approximated within 5% error of the numerical solution
necessary to correctly translate the measured stopped-flow using137
signal into permeability, which deserves careful attention.
r0ks(Cin + Cout)
Common Use of Vesicle Stopped-Flow Equation Pw ≈
Overestimates Permeability. Based on stopped-flow experi- 6VwCout 2 (12)
ments with osmotic gradients, the osmotic water permeability,
where Cin is the initial intravesicular osmolarity.
Pw, is related to the rate of change in the normalized vesicle
Stopped-flow experiments are frequently conducted below
volume, Ṽ (t):128,129

A s,0 ji Cin(t ) y
ambient temperature to improve resolution. To predict

jj
j − Cout zzzz
performance under normal desalination conditions at ambient
dṼ (t )
V0 k V (t ) {
= PwVw temperature, the osmotic water permeability can be adjusted

ÅÄÅ ÑÉ
dt ̃

1 yzzÑÑÑÑ
using the Arrhenius expression:
ÅÅ −Ea ij 1
(8)

Pw(T2) = Pw(T1)expÅÅ Å j
j − zzzÑÑ
ÅÅ R̂ jj T2 T1 {ÑÑÑÖ
ÅÇ k
where Vw is the molar volume of water, As,0 is the initial average
outer surface area of the vesicles, V0 is the initial average vesicle (13)
volume, Cin(t) is the inner vesicular osmolarity with time, and
Cout is the outer osmolarity after mixing, typically taken as a where T1 is the absolute stopped-flow experimental temper-
constant due to the relatively low vesicle concentration. ature, T2 is absolute ambient temperature, and Ea is the
Similarly, the following equation can be used to determine activation energy of water transport through the channel of
solute permeability, Ps:73 interest. The water permeability coefficient of the biomimetic

A s,0 ij Cin,S(t ) yz
layer is then calculated from Pw using eq 3.
jjC zz
jj out,S − z
Ṽ (t ) z{
dCin,S(t ) Although the biophysical community predominantly directly
V0 k
= Ps fits eq 8 to stopped-flow data for determining water
dt (9) permeability, a longstanding confusion within the biomimetic
10905 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

membrane community has led to the incorrect use of an vesicles faster than the permeable solute can enter. A second,
exponential approximation. In the 1990s, a couple of biophysical more gradual curve appears as the permeable solute slowly
studies presented the osmotic permeability calculation:88,138 enters the vesicle and water flows back in to reach osmotic
equilibrium. To find solute permeability, eq 9 is fitted to the
roks second gradual curve.73 Notably, water permeability can be
Pw =
3Vw(Cout − Cin) (14) found from the same stopped-flow data by fitting eq 8 to the
initial sharp-decline curve.
This approximation is nearly identical to eq 11, except that the
volume of the bilayer is neglected andmuch more critically EVALUATING THE PERFORMANCE OF BIOMIMETIC
the permeability is related to the initial concentration difference, MEMBRANES
Cout − Cin, rather than just the external concentration, Cout. The
approximation in eq 14 has been extensively used in biomimetic Modeling Transport Properties Through Composite
membrane literature to estimate Pw.26−28,30,32−35,57,90−93,120−123 Layers. Modeling the expected performance of membranes can
Unfortunately, this fitting procedure systematically leads to provide a benchmark for membrane fabrication efforts as well as
dramatically higher permeabilities than when using the illustrate the upper performance limits of a particular material.
fundamental water flux equation, eq 8.73,137 We modeled transport behavior to predict desalination
To stress the magnitude of error stemming from the use of eq performance for biomimetic membrane configurations with
14, we have recalculated the vesicular permeabilities for stopped- both intact vesicles and pore-spanning planar layers. We focus
flow signals provided by numerous studies involving AqpZ- on biomimetic desalination membranes with perfectly water-
laden matrices (Figure 5b; Figure S2). We compared results selective AqpZ. However, similar modeling approaches could
from the correct use of iterative numerical methods in Python serve for predicting performance of other biomimetic formats.
coding software (see SI for code) to solve for best fit with eq 8 With knowledge of the transport behavior for each individual
with the incorrect use of eq 14. Use of the incorrect equation led layer in the biomimetic desalinating composite, we modeled the
to water permeabilities 2−12-fold greater than from a correct overall permeability and salt rejection using the resistance
fitting procedure. We also present the use of eq 12 as a good model,139 similar to models used elsewhere for biomimetic
approximation using single-exponential fitting of stopped-flow membrane performance.140 The general premise of the model is
signals, which was generally (but not always) close to the results to treat molecular flow through a membrane as analogous to
from directly fitting eq 8. Typically, we saw that the magnitude of electron flow through a circuit, where the goal is to project the
the error from using eq 14 was mostly related to the solute overall resistance to flow for a given species. This overall
concentrations used. This can be more directly illustrated when resistance is then simply the inverse of the overall permeability.
the same stopped-flow data are used to solve for water As with circuits, membrane components can be considered as
permeability over a varied range of Cout − Cin using directly resistors-in-series for vertically stacked components or as
fitted eq 8 as well as exponential approximations, eqs 11, 12, and resistors-in-parallel for components that are laterally arranged
14 (Figure 5c).73 Using direct fitting and the approximations in (e.g., channels in an amphiphilic matrix, in eq 7). As an example,
eqs 11 and 12 tend to result in similar, concentration- an encased vesicle format comprises three types of membrane
independent permeability, whereas permeability found through composite areas: (i) areas with two bilayers (i.e., entering and
eq 14 shows marked concentration-dependence, which is exiting a given vesicle) covered with encasing polymer over the
aphysical, and especially overestimates permeability at a smaller support, (ii) areas with only encasing polymer over the support
difference between intra- and extra-vesicular concentrations where vesicles are absent, and (iii) defective regions with just the
(Figure 5c). We stress that eq 14 with its concentration- support layer exposed.
dependent fitting not only yields an inaccurate permeability but More specifically, the permeability of each layer, Pj, in a total
also cannot be used for a valid comparison between channels of n layers for a given type of composite x is incorporated in the
from different studies due to dissimilar testing conditions. summation:
These recalculations suggest that all amphiphilic matrices n
with AqpZ have exhibited, at best, permeabilities in the range of 1 1
= ∑
1−2 L m−2 h−1 bar−1, resulting from the limited maximum Px j=1
Pj (15)
possible channel loading density. The recalculated permeabil-
ities for unoptimized protein insertion efficiency averaged to This resistance model is only valid for composites comprising
around 1 L m−2 h−1 bar−1,28,35,89,90,120,123 while studies that solution-diffusion-based membranes. For porous support layers,
optimized insertion reached bilayer permeabilities of around 2 L we assumed rejection was negligible in cases where the top layer
m−2 h−1 bar−1 at 25 °C.57,122 The highest water permeability rejects significantly more than the support. In cases where a
achieved was 2.1 L m−2 h−1 bar−1 (Figure 5b),123 equivalent to support layer is expected to provide significant resistance to
an AqpZ areal coverage of 2.5% in a DOPC bilayer (Figure 2i). solute permeation, a real rejection of the support or solute
This water permeability is comparable to that of commercial, permeability coefficients should be used.140
state-of-the-art TFC-RO desalination membranes. From here, the salt and water flux contributions for each
In contrast to water permeability, solute permeability has not composite type are determined using eqs 1−4. To determine
been as well-studied in biomimetic membrane literature. total water and salt fluxes for the entire membrane, Jt, transport
However, stopped-flow can nonetheless be used to determine through the different composite types are considered as
the transport behavior of more permeable solutes and predict resistors-in-parallel:
resultant biomimetic membrane performance. Examples of
permeable solute stopped-flow curves with self-quenching m
encapsulated fluorophore carboxyfluorescein are presented in Jt = ∑ ψxJx
Figure 5e. Rapid self-quenching occurs as water first exits the x=1 (16)

10906 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

Figure 6. Predicted performance of planar biomimetic membrane formats. (a) Water/salt (sodium chloride) selectivity (left) and water/boron
selectivity (right) with varied permeability of biomimetic layer with nanochannels. For water/salt selectivity, the trends shown are for various
densities of AqpZ, PAH[4], and nCNTP in DOPC. For boron selectivity of AqpZ-loaded biomimetic membranes, matrix types represented
include DOPC, ABA triblock copolymer MDM, and AB diblock copolymer PB-PEO, as boric acid permeates these matrices at different rates.
Amphiphilic matrix water and salt permeability coefficients were retrieved from ref 73 with sodium as the limiting ion. For performance
comparison, data from polyamide RO membranes is also presented, calculated from manufacturer standards.142 Data for optimized channel
insertion areal coverage of 2.5% are pinpointed for AqpZ, Aqf1, PAH[4], and nCNTP as well as the most permeable biological water- and
glycerol-transporting channel GlpF. Single-channel AqpZ and GlpF water permeabilities were taken from stopped-flow studies.56 Water and
salt permeability coefficients of PAH[4] were taken from ref 82 with chloride as the limiting ion. (b) Predicted rejection of sodium chloride
(left) and pure water permeability (right) versus defect coverage for biomimetic membranes in seawater RO. The model includes a biomimetic
active layer on a support. The active layer was modeled with the best-achieved real AqpZ density, resulting in a biomimetic layer permeability of
2.1 L m−2 h−1 bar−1. The range of supports include ultrafiltration and nanofiltration membranes with pores 20 and 1 nm in diameter,
respectively, as well as looser, high-permeability TFC-RO polyamide and tighter, high-rejecting TFC-RO. Reference lines/regions for best-
observed TFC-RO salt rejection (99.85%) as well as the range of water permeability for current commercial TFC-RO polyamide membranes are
included for comparison.140 Models were conducted at standard seawater desalination test conditions (55 bar, 0.55 M NaCl). (c) Permeability
(P in m/s) trends through amphiphilic matrix materials (i.e., no channels) for neutral solutes of varied octanol/water partition coefficients Kow.
These solutes more readily permeate through the channel-incorporating matrix than ions. Data taken from ref 73.

where ψx is the fractional coverage of each composite type x, fluxes, we can calculate the overall observed rejection using eq 4
with m total types of composites. With total water and solute (see SI for more details).
10907 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

We used our permeability and rejection models to predict slightly higher energy usage during seawater desalination.4
best-case-scenario performance of defect-free desalination While either polyamide support layer could enable promising
biomimetic membranes. We also used the models to illustrate biomimetic membranes, there remains a minor permeability/
the detrimental effect of defects and the importance of support selectivity trade-off.
layer choice in alleviating defects.140 Adequate neutral solute rejection is critically important in
The Potential of Biomimetic Membranes. A perfectly most RO applications. Although discrete channels with water-
formed biomimetic membrane with aquaporins could promise sized pores can molecularly sieve out essentially all solutes,
unparalleled salt rejection. Based on matrix and AqpZ water and lipophilic solutes can permeate through the amphiphilic matrix
salt permeabilities from vesicular stopped-flow studies, sodium by solution diffusion. The affinity for a neutral solute to partition
chloride rejection is predicted to be >99.9999999% for typical into and diffuse through the amphiphilic matrix is well-
AqpZ loading quantity (Figure 6a, left).73 Meanwhile, boron correlated with the solute’s octanol/water partition coefficient
rejection can be as high as >99.9% when the matrix comprises an Kow or the ratio of the chemical’s equilibrium concentration in
AB diblock copolymer of PB46-b-PEO20, (Figure 6a, right). the octanol phase versus the water phase (Figure 6c).
Considering this potential solute rejection, it is clear that no Hydrophilic solutes with log Kow less than −0.2 are predicted
studies have produced defect-free biomimetic layers that fully to be well rejected (>90%) by a bilayer with 2.5% AqpZ, but for
cover the support. Thus far, salt rejection for AqpZ-loaded pore- solutes with Kow higher than this value, rejection would
spanning planar layers has been in the range of only 20− significantly suffer. Typical neutral organic contaminants
75%.32−35 Although optimal vesicle rupturing pressures have found in wastewater are hydrophobic and have log Kow much
improved rejection for gramicidin-containing membranes,31 the greater than −0.2. These molecules would therefore be poorly
reported rejection of 97% sodium chloride (or 3% salt passage) rejected by a biomimetic active layer.73 Likewise, although
in this single study still cannot compete with current TFC-RO boron rejection in a PB−PEO/AqpZ system is predicted to be
membranes that exhibit as low as ∼0.15% salt passage. >99.9%, it is predicted to be 95−98% when DOPC lipid or an
The impact of defects on performance is markedly influenced ABA triblock copolymer of PMOXA-PDMS-PMOXA (MDM)
by the transport properties of the underlying support layer(s).140 are used (Figure 6a). These differences in solute transport or
In particular, a relatively low-permeability, salt-rejecting support rejection behavior demonstrate the importance of matrix
layer mitigates the negative impact of defects on salt rejection. material in certain applications.140
To illustrate this point, we predicted the performance of planar- A salt-selective support layer can minimize the impact of
layer biomimetic membranes with increasing defect area, formed defects on salt rejection while maintaining sufficient perme-
on top of support layers varying from relatively selective RO- ability for seawater RO.73,140 The combination of a biomimetic
type membranes to highly permeable porous membranes, layer with a high-rejection RO membrane can marry the benefits
modifying the results of our previous study by considering the of both layers while avoiding their downsides. In seawater RO
true best-case AqpZ-matrix permeability of 2.1 L m−2 h−1 bar−1 and wastewater reuse, a high-rejection polyamide RO
(Figure 6b).140 As discussed earlier, increased selectivity and salt membrane has a relatively high rejection of salt and superb
rejection is a critical research goal. Therefore, we consider the rejection of larger neutral pollutants due to size-based removal
defect areas that would allow for increased salt rejection above mechanism (i.e., slower diffusion of larger solutes). The
that of currently employed RO membranes. For all support molecular weights of most neutral solutes present in wastewater
types, except RO-type layers, just 1% defect area would are larger than 80 Da, exceeding typical MWCO for polyamide
undermine the performance of the biomimetic membrane. As RO membranes.73 However, TFC-RO membranes are still not
a result, RO-type membranes (i.e., polyamide thin-film completely selective for salt, ammonia, boron, or low molecular
composite) may be ideal supports for biomimetic membranes weight wastewater organic contaminants, thereby requiring
to minimize the impact of defects. Additionally, the synthesis postprocessing to meet regulatory restrictions for irrigation or
methods of these polymeric membranes are often modified to wastewater reuse as well as ultrapure levels for the semi-
adjust their transport properties, with some having relatively conductor and power industries. The combination of both a
high water permeability and low salt rejection and others having high-rejection polyamide RO and biomimetic active layer in a
relatively low water permeability and high salt rejection. In our defect-free membrane is projected to reject >90% of the most
model, we consider a high permeability membrane with a water common neutral organic pollutants and >99.9% for the majority.
permeability of 6.1 L m−2 h−1 bar−1 and sodium chloride Even with a 1% defect area, ∼99.7% of boron would still be
rejection of ∼99% as well as a high rejection membrane with a rejected.73
permeability of 2.3 L m−2 h−1 bar−1 and a salt rejection of Defects Explain Subpar Performance. So far, biomimetic
∼99.8%, which are characteristic of the commercial brackish membranes have not approached the theoretical desalination
water and seawater RO membranes XLE and SW30-XLE performance (Figure 6a, b), suggesting the presence of defects
(DuPont Water), respectively.141 based on observed transport behavior (Figure 6b). Previously,
When using a support layer composed of either polyamide defects were not quantifiedeven when vesicle and planar
material, highly selective performance may be achievable with membrane permeabilities were explicitly measured within the
meaningful defect levels. For example, for a defect density of 1%, same studybecause stopped-flow measurements were (i)
a biomimetic membrane with a support of high-permeability decoupled from planar measurements, meaning the osmotic
polyamide would reject ∼99.98% of sodium chloride, while water permeability was not related to the final biomimetic
using a support of high-rejection polyamide would result in membrane permeability, and (ii) the fitting equations used to
99.996% rejection, corresponding to roughly two orders-of- predict permeability were wrong, systematically overestimating
magnitude lower salt passage than for current commercial possible permeability for a defect-free membrane. We have
desalination membranes. Meanwhile, the former would have a estimated the defect coverage to better understand observed
permeability of 1.6 L m−2 h−1 bar−1, while the latter would performance. For formats with intact vesicles, the influence of
permeate at only 1.1 L m−2 h−1 bar−1, which would necessitate defects was modeled as areas lacking polymer matrix material,
10908 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

Figure 7. Defects for typical biomimetic membrane formats. (a) Defects defined for vesicle-encapsulated mixed-matrix formats, which occur
within the nonporous polymer. (b) Defects defined for pore-spanning planar layer format, which occur within the amphiphilic channel insertion
matrix. Note that in all cases, a defect is essentially an area where the porous support layer directly contacts bulk feed solution. (c) Defect-free
theoretical effective water permeability compared to observed performance. (d) Defect-free theoretical sodium chloride rejection compared to
observed performance. Selection of pure water permeability versus water permeability in the presence of saltwater feed depended on which was
reported in literature. Lower observed rejection and higher water permeability compared to theoretical values suggest presence of defects. (e)
Model-projected coverage of defects based on rejection and permeability models. Defect coverage was deduced by simultaneously minimizing
the sum of square errors between model-predicted water permeability and salt rejection values compared to observed performance reported in
the literature. *Defects projected were 0% for ref 28. **Li et al.35 results did not match model; simple defects could not explain both observed
lower rejection and permeability than the starting NF substrate. For (c−e), blue italic letters refer to the following references: a,26 b,27 c,28 d,29
e,30 f31 g,32 h,33 i,34 and j.35 Note that this study f31 incorporated GramA instead of AqpZ.

while for pore-spanning planar layers, defects were defined as the former and between 1 and 10% defect coverage for the latter.
areas lacking any channel-laden matrix (i.e., exposed support However, as we will further show, formats with intact vesicles
layer in both cases) (Figure 7a,b). The coverage of vesicles and have little chance of surpassing current RO salt rejection unless
whether or not they were imprinted or encased was accounted vesicle packing density is increased to unrealistic values.
for in the models. We first predicted performance of defect-free Vesicle-Encapsulated Mixed-Matrices Negligibly Im-
membranes for five previous studies incorporating intact vesicles prove Membrane Performance. Just as the performance of
and five with pore-spanning planar layers, showing that in nearly biomimetic membranes hinges on the negative impact of
every case, defect-free membranes are predicted to have lower defects, the enhancement of salt rejection by adding vesicles
permeability and higher rejection than achieved (Figure 7c,d). within a polymer layer hinges on the areal coverage of such
To estimate the defect density, we used the previously defined vesicles. The question then arises whether biomimetic
transport equations to conduct a least-squares analysis on the membranes with intact vesicles have the potential to surpass
reported permeate flux during salt rejection studies or pure water the performance of current state-of-the-art desalination
permeability (depending on what was reported) and salt membranes. Thus, we modeled the influence of vesicle coverage
rejection (Figure S3). The predicted defect density of across a range of 0% to 90%, with 90% corresponding to the
biomimetic membranes reported in several studies is presented theoretical maximum packing density of monodisperse circles
in Figure 7e. In general, we see that lower defect densities are on a 2D plane, which is currently unrealistic to achieve. Typical
estimated for formats incorporating intact vesicles than pore- coverages are in the range of only 7−20%,26−28,30,92 although
spanning planar layers, typically with 0−1% defect coverage for one study achieved a vesicular coverage of ∼35% by using
10909 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

Figure 8. Best-case-scenario performance of biomimetic membranes with vesicle-encapsulated mixed matrices. (a) Pure water permeability
(top) and maximum possible salt rejection with no defects (bottom) for varied 2D packing densities of aquaporin-laden vesicles in vesicle-
encapsulated mixed matrices. Two different common commercial TFC-RO membranes, XLE and SW30-XLE (DuPont Water) or high-
permeability and high-rejection RO, were used as model nonporous polymers surrounding the vesicles. Vesicles both encased and imprinted in
nonporous polymer are represented. The biomimetic membranes were modeled on top of an ultrafiltration membrane with pores of 3 nm in
diameter (UF 3). The blue region represents the typical vesicle surface coverage found in literature,26−28,30,92 while the yellow region is the
permeability of current commercial TFC-RO. Note that by using cross-linked vesicles of polyethylene-glycol-conjugated DOPC on a dopamine-
coated support, one study achieved a vesicular coverage of ∼35%.30 The theoretical maximum possible surface coverage of vesicles is assumed
to be the maximum packing density of spheres on a plane, or 0.9. The permeability of the bilayer was assumed to be 2.1 L m−2 h−1 bar−1. (b) A
typical surface coverage of 15% projected area, with calculated flux contributions for water passing through vesicles versus just polyamide over
support layer. (c) Theoretical maximum surface coverage of vesicles at 90% projected area, with calculated flux contributions for each type of
area.

polyethylene-glycol-conjugated DOPC to self-cross-link the vesicle coverage would have to dramatically increase. Consid-
vesicles for stability and later cross-link them to a dopamine- ering that vesicle surface coverage is typically 5−20%, achieving
coated support.30 We used high-rejecting polyamide and high- nearly 90% ordered packing will certainly be nontrivial (Figure
permeability polyamide as the model polymers that either S1).26−28,30,92
imprint (i.e., surround) or fully encase the vesicles, assuming no We emphasize that the model we used for vesicle-laden
defects are present. membranes assumes unidirectional transport and treats vesicles
For both imprinted and encased vesicles, which show with channels as an isotropic material. Realistically, multidirec-
insignificant differences in performance, pure water permeability tional diffusion may cause species to avoid the well spaced-out
linearly decreases with increasing vesicle coverage, while salt channels in the vesicle bilayers and frequently bypass the vesicles
rejection improves most rapidly within higher ranges of vesicular completely. The only way to avoid this possibility would be with
coverage (Figure 8a). We see minimal benefits of added vesicles close to 100% vesicular surface coverage, which is not possible
for higher permeability polymer matrices. With maximum for spheres on a 2D plane. Finally, this model assumes that
imprinted vesicular coverage, high-permeability polyamide defects are avoided, which may be unattainable considering the
exhibits up to a ∼0.5% increase in rejection, not quite reaching large interfacial areas between vesicle walls and the surrounding
the initial rejection of high-rejection polyamide, even with the polymer at high vesicle areal coverage. Collectively, our
unrealistic 90% vesicular coverage. At this vesicular coverage, the modeling results and discussion indicate that vesicle-encapsu-
permeability of the membrane drops to below that of the high- lated mixed matrices negligibly improve membrane performance
rejection polyamide system, rendering the technology overly and, compounded with the involved complex synthesis, would
complicated with no added benefit in desalination. likely not be worth pursuing further.
For a high-rejection, tight polyamide matrix, maximum vesicle Optimal Biomimetic Desalination Membrane Design.
coverage (i.e., 90%) is projected to yield a 0.12% increase in salt Although methodological changes can decrease defect density,
rejection to ∼99.97%. Although pure water permeability would to date, even the most well-formed planar bilayers have suffered
drop to 1.2 L m−2 h−1 bar−2 or 0.86 L m−2 h−1 bar−2 for formats from a sufficient defect density to completely undermine the
with imprinted and encased vesicles, respectively, 99.97% salt capabilities of the aquaporin channel. Currently, an absolutely
rejection is an unparalleled level. However, to achieve this target, defect-free bilayer would be difficult to achieve, especially at an
10910 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

Figure 9. Concepts for high-performance biomimetic membranes. (a) Synthesis of an ion-rejecting or ion-selective biomimetic membrane. The
presence of functional groups assists in rupturing nanochannel-laden vesicles. Nanoscale defects are filled with a hydrogel reaction solution,
which is subsequently polymerized. The choice of channel would determine whether ions are fully rejected or selectively rejected. (b) Synthesis
of an ultrahigh rejection membrane. Synthesis begins with the formation of a smooth polyamide layer with a dense layer of functional groups.
Functional groups assist in the bottom-up polymerization of the channel insertion matrix, with subsequent insertion of nanochannels, or
rupturing of nanochannel-laden vesicles.

industrial scale. Thus, we propose two approaches that may could be particularly beneficial in applications such as ultrapure
allow for defects while still achieving ultrahigh salt rejection water production, seawater desalination, and wastewater reuse.
(Figure 9). In the first case, a biomimetic selective layer is The proposed designs build on previous analyses on ideal
formed on a surface-functionalized porous membrane by permeability and salt-water selectivity to increase desalination
traditional vesicle rupture or by growing a polymer brush efficiency4 and the influence of support layer on biomimetic
layer,143 followed by channel insertion within the solid- membrane performance.140 Although the synthesis details of
supported brush layer.144 These processes must be optimized how to effectively produce these envisioned membranes would
to minimize defects. A hydrogel coating layer would then fill need to be experimentally resolved and may prove challenging,
defects after biomimetic active layer formation,145 analogous to the suggested schemes piece together methodologies that have
silicone defect-sealing layers used in gas separation mem- proven effective in other applications. The chemical stability in
branes.139 This hydrogel would decrease the water permeability various environments as well as mechanical strength to
of defects, driving more flow through the selective channels. withstand operating pressures and shear stresses would need
Depending on the channel type chosen, the membrane would to be investigated and optimized, as is the case for all biomimetic
either highly or selectively reject ions. However, lipophilic membranes. The polymeric matrix material will need to be
solutes would easily pass through the amphiphilic matrix in this carefully chosen to enable both functional channel reconstitu-
design (Figure 9a). In the second vision, a smooth polyamide tion and scalable selective layer synthesis. Furthermore, with the
layer is functionalized with grafting sites to either improve the addition of a polyamide layer, the biomimetic selective layer
top-down deposition and subsequent rupturing of vesicles or to would ideally have increased permeability to overcome the
grow the channel-incorporating matrix from a bottom-up additional resistance. Synthetic channels should be developed
approach. We envision that such a membrane would highly with low outer-to-inner pore diameter ratios so more pores per
reject ions with the biomimetic active layer and also neutral channel areal density are present. Perhaps strategically designed
solutes with the polyamide layer (Figure 9b). This latter design external channel features or matrix compositions could also
10911 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

increase channel incorporation. The biomimetic matrix and More niche applications that could benefit from molecular
channel pairing should be optimized first by use of stopped-flow sieving include recovery of valuable chemicals from waste
experiments with correctly applied permeability equations. streams, analytical separations of chemically analogous and size-
Before attempting new biomimetic formats, best-case-scenario similar species, and high-precision industrial separations. Even
performance as well as outcomes with typical defect density can in these situations, strategies to ensure that all transport takes
be predicted using similar transport modeling methods as we place through the molecular sieves and not through the
demonstrated here. surrounding matrix would be necessary. For instance, as we
have shown, fluid-like biomimetic matrices can readily transport
CONCLUSION AND OUTLOOK lipophilic, neutral solutes. For a more impermeable matrix,
Breakthrough desalination membranes will likely rely on perhaps synthetic channels could be incorporated and self-
molecular-sieve-based transport, as is achieved in biological assemble in polymers above the glass transition temperature and
and biomimetic membranes. In this review, we have shown that become immobilized within a glassy polymer upon cooling. In
an optimized biomimetic desalination membrane could achieve cases of more random alignment of nanochannels, the addition
salt rejections above those attainable by current RO membranes, of channel exterior functionalities could promote stronger
while still maintaining sufficient water permeability. However, interactions between channel and surrounding matrix, decreas-
the major challenge for common biomimetic membrane formats ing the presence of interfacial defects.
resides in fully covering the substrate with a defect-free, planar The promise of biomimetic membranes lies in the perform-
channel-laden bilayer. The performance of a biomimetic ance of a well-engineered, highly stable discrete nanochannel,
desalination membrane essentially hinges on maximizing while its biggest threat resides in the surrounding matrix
transport through highly selective nanochannels while minimiz- material. In general, defects persist in all current efforts toward
ing defects. molecular sieves that incorporate next-generation nanomaterials
A significant portion of biomimetic membrane research is during top-down synthesis. For 2D laminates, interconnected
devoted to the development of synthetic channels for higher framework defects override the intended molecular sieving of
stability and smaller per-channel physical footprints than interlayer spaces or nanosheet pores, while mixed matrices
transmembrane integral proteins. Although the development incorporating nanotubes or biomimetic nanochannels suffer
of nanochannels is important, we argue that equal attention from defects like those described in this review. There has been
should be paid to the general biomimetic platform, including over a decade of research toward biomimetic membranes using
polymeric matrices that enable functional channel reconstitu- the top-down approaches discussed in this review. Based on
tion and robust membrane performance. Such work must start vesicular formats assessed via stopped-flow, several nano-
with the optimization of channel reconstitution in vesicles using channels theoretically promise water-salt selectivity dramatically
stopped-flow studies with correct permeability equations and exceeding that of current desalination membranes, yet planar
then transition to lab-scale planar coupons that include lab-scale biomimetic membranes have not achieved similar
strategies for overcoming defects. Once successful, only then performance. This struggle to transition from a vesicular format
can scientists turn to industrial-scale biomimetic formats. Unless to a small-area planar membrane coupon suggests that scaling up
synthesis strategies are optimized, the persistence of defects in to the industrial level will be a daunting challenge. However,
the matrix can completely undermine the capabilities of the most through strategic reformatting, biomimetic membranes could
extraordinary molecular sieves. enable a breakthrough performance in RO and find applications
Most biomimetic membrane studies have focused on beyond desalination.
desalination rather than targeting other applications, likely
because the transport mechanisms of the inspiring aquaporin ASSOCIATED CONTENT
initially suggested desalination. However, it is important to *
sı Supporting Information
assess if the possibility of biomimetic desalination is worth the The Supporting Information is available free of charge at
probable convoluted synthesis schemes. Fully aromatic https://pubs.acs.org/doi/10.1021/acsnano.0c05753.
polyamide TFC-RO membranes already allow for highly energy Descriptions of data extraction for channel permeabilities
efficient seawater desalination and reject most solutes and stopped-flow recalculations, Python codes for direct
sufficiently to reach less stringent standards for potable use, fitting of the fundamental water flux equation for vesicular
including sodium chloride rejection up to 99.85%. Furthermore, stopped-flow studies (eq 8) and erroneous exponential
polyamide membranes are relatively easy and economical to approximation (eq 14), further details on transport
produce at a large scale and are familiar to the desalination modeling; tables with modeling parameters, brief
industry. The most significant room for improvement in descriptions of studies that were considered for
desalination membranes lies with increasing rejection of boron calculation of defect density, and single-channel perme-
in seawater desalination, increasing salt rejection in the ability values; figures with example vesicle coverage
production of ultrapure water, withstanding higher pressures estimation using ImageJ, stopped-flow graphs with
in high-pressure RO toward zero liquid discharge, and overlaid models for eqs 8 and 14, and model sodium
developing highly chlorine-resistant and antifouling materials, chloride rejection and water permeability (based on
especially for wastewater treatment and reuse. If biomimetic predicted defect density) compared to reported perform-
membranes can achieve one or more of these goals, the relatively ance (PDF)
complex synthesis strategies may be worthwhile. Scientists
developing synthesis strategies must consider these particular
goals as well as technoeconomic analyses in comparison with AUTHOR INFORMATION
current commercial TFC-RO standards. Corresponding Authors
New and emerging applications away from desalination may Jay R. Werber − Department of Chemical and Environmental
best exploit the strengths biomimetic membranes have to offer. Engineering, Yale University, New Haven, Connecticut 06520,
10912 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

United States; Department of Chemistry, University of Minnesota, REFERENCES


Minneapolis, Minnesota 55455, United States; Email: jwerber@ (1) Elimelech, M. The Global Challenge for Adequate and Safe Water.
umn.edu Aqua 2006, 55, 3−10.
Menachem Elimelech − Department of Chemical and (2) Elimelech, M.; Phillip, W. A. The Future of Seawater Desalination:
Environmental Engineering, Yale University, New Haven, Energy, Technology, and the Environment. Science 2011, 333, 712−
Connecticut 06520, United States; orcid.org/0000-0003- 717.
4186-1563; Email: menachem.elimelech@yale.edu (3) Petersen, R. J. Composite Reverse Osmosis and Nanofiltration
Membranes. J. Membr. Sci. 1993, 83, 81−150.
Authors (4) Werber, J. R.; Deshmukh, A.; Elimelech, M. The Critical Need for
Cassandra J. Porter − Department of Chemical and Increased Selectivity, Not Increased Water Permeability, for Desalina-
Environmental Engineering, Yale University, New Haven, tion Membranes. Environ. Sci. Technol. Lett. 2016, 3, 112−120.
(5) Rozelle, L. T.; Cadotte, J. E.; Cobian, K. E.; Kopp, C. V., Jr.,
Connecticut 06520, United States Nonpolysaccharide Membranes for Reverse Osmosis: NS-100
Mingjiang Zhong − Department of Chemical and Environmental Membranes. In Reverse Osmosis and Synthetic Membranes. National
Engineering, Yale University, New Haven, Connecticut 06520, Research Council of Canada Ottawa: Ottawa, Ontario, 1977; pp 249−
United States; orcid.org/0000-0001-7533-4708 261.
Corey J. Wilson − School of Chemical and Biomolecular (6) Lonsdale, H. K.; Merten, U.; Riley, R. L. Transport Properties of
Engineering, Georgia Institute of Technology, Atlanta, Georgia Cellulose Acetate Osmotic Membranes. J. Appl. Polym. Sci. 1965, 9,
30332, United States; orcid.org/0000-0003-3040-1676 1341−1362.
(7) Wijmans, J. G.; Baker, R. W. The Solution-Diffusion Model: A
Complete contact information is available at: Review. J. Membr. Sci. 1995, 107, 1−21.
https://pubs.acs.org/10.1021/acsnano.0c05753 (8) Mehta, A.; Zydney, A. L. Permeability and Selectivity Analysis for
Ultrafiltration Membranes. J. Membr. Sci. 2005, 249, 245−249.
Author Contributions (9) Geise, G. M.; Park, H. B.; Sagle, A. C.; Freeman, B. D.; McGrath, J.
C.J.P. primarily wrote the manuscript and conducted transport E. Water Permeability and Water/Salt Selectivity Tradeoff in Polymers
for Desalination. J. Membr. Sci. 2011, 369, 130−138.
modeling. J.R.W. and M.E. gave significant input for overall (10) Robeson, L. M. The Upper Bound Revisited. J. Membr. Sci. 2008,
article content and organization as primary advisors for this 320, 390−400.
work. M.Z. and C.J.W. provided technical feedback and helped (11) Werber, J. R.; Osuji, C. O.; Elimelech, M. Materials for Next-
revise the manuscript. Generation Desalination and Water Purification Membranes. Nat. Rev.
Mater. 2016, 1, 16018.
Notes (12) Dervin, S.; Dionysiou, D. D.; Pillai, S. C. 2D Nanostructures for
The authors declare no competing financial interest. Water Purification: Graphene and Beyond. Nanoscale 2016, 8, 15115−
15131.
(13) Ritt, C. L.; Werber, J. R.; Deshmukh, A.; Elimelech, M. Monte
ACKNOWLEDGMENTS Carlo Simulations of Framework Defects in Layered Two-Dimensional
The authors acknowledge the support received from the Nanomaterial Desalination Membranes: Implications for Permeability
and Selectivity. Environ. Sci. Technol. 2019, 53, 6214−6224.
National Science Foundation through the Engineering Research
(14) Patel, S. K.; Ritt, C. L.; Deshmukh, A.; Wang, Z.; Qin, M.;
Center for Nanotechnology-Enabled Water Treatment Epsztein, R.; Elimelech, M. The Relative Insignificance of Advanced
(EEC1449500) and via Grant CBET 1437630. The authors Materials in Enhancing the Energy Efficiency of Desalination
also acknowledge funding from the National Science Founda- Technologies. Energy Environ. Sci. 2020, 13, 1694−1710.
(15) Zheng, S.; Tu, Q.; Urban, J. J.; Li, S.; Mi, B. Swelling of Graphene
tion Graduate Research Fellowship Program under grant no.
Oxide Membranes in Aqueous Solution: Characterization of Interlayer
DGE-1752134, awarded to C.J.P. Any opinions, findings, and Spacing and Insight into Water Transport Mechanisms. ACS Nano
conclusions or recommendations expressed in this material are 2017, 11, 6440−6450.
those of the authors and do not necessarily reflect the views of (16) Gould, S. J. The Evolution of Life on the Earth. Sci. Am. 1994,
the National Science Foundation. 271, 84−91.
(17) Schopf, J. W.; Packer, B. M. Early Archean (3.3-Billion to 3.5-
Billion-Year-Old) Microfossils from Warrawoona Group, Australia.
VOCABULARY Science 1987, 237, 70−73.
(18) Tanford, C. The Hydrophobic Effect: Formation of Micelles and
Biomimetic membrane, a composite membrane composed of a Biological Membranes, 2nd ed.; Wiley-Interscience: New York, 1980.
biomimetic active layer on a robust support; the active layer (19) Eisenberg, B. Ionic Channels in Biological Membranes: Natural
mimics biological membranes by comprising an amphiphilic Nanotubes. Acc. Chem. Res. 1998, 31, 117−123.
matrix that incorporates and aligns biological or bioinspired (20) Denker, B. M.; Smith, B. L.; Kuhajda, F. P.; Agre, P.
nanochannels; aquaporin, the class of biological transmem- Identification, Purification, and Partial Characterization of a Novel
brane proteins responsible for relatively fast and selective Mr 28,000 Integral Membrane Protein from Erythrocytes and Renal
transport of water and, in some variants, glycerol; reverse Tubules. J. Biol. Chem. 1988, 263, 15634−15642.
osmosis, process by which hydraulic pressure is applied to drive (21) Preston, G. M.; Carroll, T. P.; Guggino, W. B.; Agre, P.
a solution through a semipermeable barrier against the osmotic Appearance of Water Channels in Xenopus Oocytes Expressing Red
pressure gradient, which is used in desalination; solution- Cell CHIP28 Protein. Science 1992, 256, 385−387.
(22) Fyles, T. M. Synthetic Ion Channels in Bilayer Membranes.
diffusion model, fundamental model for molecular transport
Chem. Soc. Rev. 2007, 36, 335−347.
through isotropic membranes, whereby permeation of a (23) Sisson, A. L.; Shah, M. R.; Bhosale, S.; Matile, S. Synthetic Ion
particular species is dictated by rates of partitioning into the Channels and Pores (2004−2005). Chem. Soc. Rev. 2006, 35, 1269−
membrane material and diffusing across it; molecular sieving, 1286.
size-based molecular exclusion and transport by a semi- (24) Sakai, N.; Matile, S. Synthetic Ion Channels. Langmuir 2013, 29,
permeable barrier with rigid, uniform pores 9031−9040.

10913 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

(25) Gokel, G. W.; Negin, S. Synthetic Ion Channels: From Pores to (47) Bellissent-Funel, M.-C.; Hassanali, A.; Havenith, M.; Henchman,
Biological Applications. Acc. Chem. Res. 2013, 46, 2824−2833. R.; Pohl, P.; Sterpone, F.; van der Spoel, D.; Xu, Y.; Garcia, A. E. Water
(26) Wang, H. L.; Chung, T.-S.; Tong, Y. W.; Jeyaseelan, K.; Determines the Structure and Dynamics of Proteins. Chem. Rev. 2016,
Armugam, A.; Duong, H. H. P.; Fu, F.; Seah, H.; Yang, J.; Hong, M. 116, 7673−7697.
Mechanically Robust and Highly Permeable AquaporinZ Biomimetic (48) Berezhkovskii, A.; Hummer, G. Single-File Transport of Water
Membranes. J. Membr. Sci. 2013, 434, 130−136. Molecules through a Carbon Nanotube. Phys. Rev. Lett. 2002, 89,
(27) Xie, W.; He, F.; Wang, B.; Chung, T.-S.; Jeyaseelan, K.; 064503.
Armugam, A.; Tong, Y. W. An Aquaporin-Based Vesicle-Embedded (49) Chou, T. How Fast Do Fluids Squeeze through Microscopic
Polymeric Membrane for Low Energy Water Filtration. J. Mater. Chem. Single-File Pores? Phys. Rev. Lett. 1998, 80, 85−88.
A 2013, 1, 7592−7600. (50) Zhu, F.; Tajkhorshid, E.; Schulten, K. Collective Diffusion Model
(28) Zhao, Y.; Qiu, C.; Li, X.; Vararattanavech, A.; Shen, W.; Torres, for Water Permeation through Microscopic Channels. Phys. Rev. Lett.
J.; Hélix-Nielsen, C.; Wang, R.; Hu, X.; Fane, A. G.; Tang, C. Y. 2004, 93, 224501.
Synthesis of Robust and High-Performance Aquaporin-Based Bio- (51) Finkelstein, A. Water Movement through Lipid Bilayers, Pores, and
mimetic Membranes by Interfacial Polymerization-Membrane Prep- Plasma Membranes: Theory and Reality; Wiley: New York, 1987.
aration and RO Performance Characterization. J. Membr. Sci. 2012, (52) Freger, V. Selectivity and Polarization in Water Channel
423−424, 422−428. Membranes: Lessons Learned from Polymeric Membranes and
(29) Zhao, Y. Aquaporin Based Biomimetic Membrane for Water
CNTs. Faraday Discuss. 2018, 209, 371−388.
Reuse and Desalination. IDA Journal of Desalination and Water Reuse
(53) Borgnia, M. J.; Agre, P. Reconstitution and Functional
2014, 6, 159−159.
Comparison of Purified GLPF and AQPZ, the Glycerol and Water
(30) Sun, G.; Chung, T.-S.; Jeyaseelan, K.; Armugam, A. Stabilization
Channels from Escherichia Coli. Proc. Natl. Acad. Sci. U. S. A. 2001, 98,
and Immobilization of Aquaporin Reconstituted Lipid Vesicles for
Water Purification. Colloids Surf., B 2013, 102, 466−471. 2888−2893.
(31) Saeki, D.; Yamashita, T.; Fujii, A.; Matsuyama, H. Reverse (54) Tunuguntla, R. H.; Henley, R. Y.; Yao, Y.-C.; Pham, T. A.;
Osmosis Membranes Based on a Supported Lipid Bilayer with Wanunu, M.; Noy, A. Enhanced Water Permeability and Tunable Ion
Gramicidin a Water Channels. Desalination 2015, 375, 48−53. Selectivity in Subnanometer Carbon Nanotube Porins. Science 2017,
(32) Wang, M.; Wang, Z.; Wang, X.; Wang, S.; Ding, W.; Gao, C. 357, 792−796.
Layer-By-Layer Assembly of Aquaporin Z-Incorporated Biomimetic (55) Sperelakis, N. Cell Physiology Sourcebook: A Molecular Approach,
Membranes for Water Purification. Environ. Sci. Technol. 2015, 49, 3rd ed.; Elsevier: Netherlands, 2001.
3761−3768. (56) Horner, A.; Zocher, F.; Preiner, J.; Ollinger, N.; Siligan, C.;
(33) Duong, P. H. H.; Chung, T.-S.; Jeyaseelan, K.; Armugam, A.; Akimov, S. A.; Pohl, P. The Mobility of Single-File Water Molecules Is
Chen, Z.; Yang, J.; Hong, M. Planar Biomimetic Aquaporin- Governed by the Number of H-Bonds They May Form with Channel-
Incorporated Triblock Copolymer Membranes on Porous Alumina Lining Residues. Sci. Adv. 2015, 1, No. e1400083.
Supports for Nanofiltration. J. Membr. Sci. 2012, 409−410, 34−43. (57) Ren, T.; Erbakan, M.; Shen, Y.-x.; Barbieri, E.; Saboe, P.; Feroz,
(34) Zhong, P. S.; Chung, T.-S.; Jeyaseelan, K.; Armugam, A. H.; Yan, H.; McCuskey, S.; Hall, J. F.; Schantz, A. B.; Bazan, G. C.;
Aquaporin-Embedded Biomimetic Membranes for Nanofiltration. J. Butler, P. J.; Grzelakowski, M.; Kumar, M. Membrane Protein Insertion
Membr. Sci. 2012, 407−408, 27−33. into and Compatibility with Biomimetic Membranes. Adv. Biosyst.
(35) Li, X.; Wang, R.; Tang, C.; Vararattanavech, A.; Zhao, Y.; Torres, 2017, 1, 1700053.
J.; Fane, T. Preparation of Supported Lipid Membranes for Aquaporin (58) Amiry-Moghaddam, M.; Ottersen, O. P. The Molecular Basis of
Z Incorporation. Colloids Surf., B 2012, 94, 333−340. Water Transport in the Brain. Nat. Rev. Neurosci. 2003, 4, 991−1001.
(36) Campbell, N. A.; Reece, J. B.; Mitchell, L. G.; Taylor, M. R. (59) Kozono, D.; Yasui, M.; King, L. S.; Agre, P. Aquaporin Water
Biology: Concepts & Connections, 4th ed.; Benjamin Cummings: San Channels: Atomic Structure Molecular Dynamics Meet Clinical
Francisco, CA, 2003. Medicine. J. Clin. Invest. 2002, 109, 1395−1399.
(37) Tang, C.; Wang, Z.; Petrinić, I.; Fane, A. G.; Hélix-Nielsen, C. (60) de Groot, B. L.; Frigato, T.; Helms, V.; Grubmüller, H. The
Biomimetic Aquaporin Membranes Coming of Age. Desalination 2015, Mechanism of Proton Exclusion in the Aquaporin-1 Water Channel. J.
368, 89−105. Mol. Biol. 2003, 333, 279−293.
(38) Tang, C. Y.; Zhao, Y.; Wang, R.; Hélix-Nielsen, C.; Fane, A. G. (61) Tajkhorshid, E.; Nollert, P.; Jensen, M.Ø.; Miercke, L. J. W.;
Desalination by Biomimetic Aquaporin Membranes: Review of Status O’Connell, J.; Stroud, R. M.; Schulten, K. Control of the Selectivity of
and Prospects. Desalination 2013, 308, 34−40. the Aquaporin Water Channel Family by Global Orientational Tuning.
(39) Shen, Y.-x.; Saboe, P. O.; Sines, I. T.; Erbakan, M.; Kumar, M. Science 2002, 296, 525−530.
Biomimetic Membranes: A Review. J. Membr. Sci. 2014, 454, 359−381. (62) Ilan, B.; Tajkhorshid, E.; Schulten, K.; Voth, G. A. The
(40) Yang, Z.; Ma, X.-H.; Tang, C. Y. Recent Development of Novel
Mechanism of Proton Exclusion in Aquaporin Channels. Proteins:
Membranes for Desalination. Desalination 2018, 434, 37−59.
Struct., Funct., Genet. 2004, 55, 223−228.
(41) Song, W.; Lang, C.; Shen, Y.-x.; Kumar, M. Design
(63) Sui, H.; Han, B.-G.; Lee, J. K.; Walian, P.; Jap, B. K. Structural
Considerations for Artificial Water Channel−Based Membranes.
Basis of Water-Specific Transport through the Aqp1 Water Channel.
Annu. Rev. Mater. Res. 2018, 48, 57−82.
(42) Wagh, P.; Escobar, I. C. Biomimetic and Bioinspired Membranes Nature 2001, 414, 872−878.
for Water Purification: A Critical Review and Future Directions. (64) Sahraoui, M.; Kaviany, M. Slip and No-Slip Velocity Boundary
Environ. Prog. Sustainable Energy 2019, 38, No. e13215. Conditions at Interface of Porous, Plain Media. Int. J. Heat Mass
(43) Hannesschlaeger, C.; Horner, A.; Pohl, P. Intrinsic Membrane Transfer 1992, 35, 927−943.
Permeability to Small Molecules. Chem. Rev. 2019, 119, 5922−5953. (65) Gravelle, S.; Joly, L.; Ybert, C.; Bocquet, L. Large Permeabilities
(44) Hoek, E. M. V.; Kim, A. S.; Elimelech, M. Influence of Crossflow of Hourglass Nanopores: From Hydrodynamics to Single File
Membrane Filter Geometry and Shear Rate on Colloidal Fouling in Transport. J. Chem. Phys. 2014, 141, 18C526.
Reverse Osmosis and Nanofiltration Separations. Environ. Eng. Sci. (66) Gravelle, S.; Joly, L.; Detcheverry, F.; Ybert, C.; Cottin-Bizonne,
2002, 19, 357−372. C.; Bocquet, L. Optimizing Water Permeability through the Hourglass
(45) Geise, G. M.; Paul, D. R.; Freeman, B. D. Fundamental Water and Shape of Aquaporins. Proc. Natl. Acad. Sci. U. S. A. 2013, 110, 16367−
Salt Transport Properties of Polymeric Materials. Prog. Polym. Sci. 16372.
2014, 39, 1−42. (67) Kocsis, I.; Sun, Z.; Legrand, Y. M.; Barboiu, M. Artificial Water
(46) Horner, A.; Pohl, P. Single-File Transport of Water through ChannelsDeconvolution of Natural Aquaporins through Synthetic
Membrane Channels. Faraday Discuss. 2018, 209, 9−33. Design. npj Clean Water 2018, 1, 13.

10914 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

(68) Razmkhah, M.; Ahmadpour, A.; Mosavian, M. T. H.; Moosavi, F. Mediated Interactions in Lipid Bilayers Containing Gramicidin.
What Is the Effect of Carbon Nanotube Shape on Desalination Process? Biophys. J. 1999, 76, 937−945.
A Simulation Approach. Desalination 2017, 407, 103−115. (87) Verkman, A. S.; Mitra, A. K. Structure and Function of Aquaporin
(69) The Pymol Molecular Graphics System; Schrödinger, Inc.: New Water Channels. Am. J. Physiol.: Renal Physiol. 2000, 278, F13−F28.
York, 2019. (88) Borgnia, M. J.; Kozono, D.; Calamita, G.; Maloney, P. C.; Agre, P.
(70) Savage, D. F; Egea, P. F; Robles-Colmenares, Y.; O'Connell, J. D., Functional Reconstitution and Characterization of Aqpz, the E. Coli
III; Stroud, R. M Architecture and Selectivity in Aquaporins: 2.5 Å X- Water Channel Protein. J. Mol. Biol. 1999, 291, 1169−1179.
Ray Structure of Aquaporin Z. PLoS Biol. 2003, 1, e72. (89) Zeidel, M. L.; Ambudkar, S. V.; Smith, B. L.; Agre, P.
(71) Townsley, L. E. The Three-Dimensional Structure of Gramicidin Reconstitution of Functional Water Channels in Liposomes Containing
Analogs in Micellar Environments Determined Using Two-Dimen- Purified Red Cell Chip28 Protein. Biochemistry 1992, 31, 7436−7440.
sional Nuclear Magnetic Resonance Spectroscopic Techniques. Ph.D. (90) Li, X.; Wang, R.; Wicaksana, F.; Tang, C.; Torres, J.; Fane, A. G.
Defense, University of Arkansas, 2001. Preparation of High Performance Nanofiltration (NF) Membranes
(72) Ruiz, L.; Wu, Y.; Keten, S. Tailoring the Water Structure and Incorporated with Aquaporin Z. J. Membr. Sci. 2014, 450, 181−188.
Transport in Nanotubes with Tunable Interiors. Nanoscale 2015, 7, (91) Wang, H.; Chung, T.-S.; Tong, Y. W.; Meier, W.; Chen, Z.; Hong,
121−132. M.; Jeyaseelan, K.; Armugam, A. Preparation and Characterization of
(73) Werber, J. R.; Elimelech, M. Permselectivity Limits of Pore-Suspending Biomimetic Membranes Embedded with Aquaporin
Biomimetic Desalination Membranes. Sci. Adv. 2018, 4, eaar8266. Z on Carboxylated Polyethylene Glycol Polymer Cushion. Soft Matter
(74) Ghadiri, M. R.; Granja, J. R.; Milligan, R. A.; McRee, D. E.; 2011, 7, 7274−7280.
Khazanovich, N. Self-Assembling Organic Nanotubes Based on a Cyclic (92) Kumar, M.; Grzelakowski, M.; Zilles, J.; Clark, M.; Meier, W.
Peptide Architecture. Nature 1993, 366, 324−327. Highly Permeable Polymeric Membranes Based on the Incorporation
(75) Clark, T. D.; Buehler, L. K.; Ghadiri, M. R. Self-Assembling of the Functional Water Channel Protein Aquaporin Z. Proc. Natl. Acad.
Cyclic Β3-Peptide Nanotubes as Artificial Transmembrane Ion Sci. U. S. A. 2007, 104, 20719−20724.
Channels. J. Am. Chem. Soc. 1998, 120, 651−656. (93) Wang, H.; Chung, T.-S.; Tong, Y. W.; Jeyaseelan, K.; Armugam,
(76) Hourani, R.; Zhang, C.; van der Weegen, R.; Ruiz, L.; Li, C.; A.; Chen, Z.; Hong, M.; Meier, W. Highly Permeable and Selective
Keten, S.; Helms, B. A.; Xu, T. Processable Cyclic Peptide Nanotubes Pore-Spanning Biomimetic Membrane Embedded with Aquaporin Z.
with Tunable Interiors. J. Am. Chem. Soc. 2011, 133, 15296−15299. Small 2012, 8, 1185−1190.
(77) Li, Z.; Li, Y.; Yao, Y.-C.; Aydin, F.; Zhan, C.; Chen, Y.; Elimelech, (94) Garni, M.; Thamboo, S.; Schoenenberger, C.-A.; Palivan, C. G.
M.; Pham, T. A.; Noy, A. Strong Differential Monovalent Anion Biopores/Membrane Proteins in Synthetic Polymer Membranes.
Selectivity in Narrow Diameter Carbon Nanotube Porins. ACS Nano Biochim. Biophys. Acta, Biomembr. 2017, 1859, 619−638.
2020, 14, 6269−6275. (95) Mecke, A.; Dittrich, C.; Meier, W. Biomimetic Membranes
(78) Le Duc, Y.; Michau, M.; Gilles, A.; Gence, V.; Legrand, Y.-M.; Designed from Amphiphilic Block Copolymers. Soft Matter 2006, 2,
van der Lee, A.; Tingry, S.; Barboiu, M. Imidazole-Quartet Water and 751−759.
Proton Dipolar Channels. Angew. Chem., Int. Ed. 2011, 50, 11366− (96) Bangham, A. D.; De Gier, J.; Greville, G. D. Osmotic Properties
11372. and Water Permeability of Phospholipid Liquid Crystals. Chem. Phys.
(79) Licsandru, E.; Kocsis, I.; Shen, Y.-x.; Murail, S.; Legrand, Y.-M.; Lipids 1967, 1, 225−246.
van der Lee, A.; Tsai, D.; Baaden, M.; Kumar, M.; Barboiu, M. Salt- (97) Rigaud, J.-L.; Lévy, D. Reconstitution of Membrane Proteins into
Excluding Artificial Water Channels Exhibiting Enhanced Dipolar Liposomes. Methods Enzymol. 2003, 372, 65−86.
Water and Proton Translocation. J. Am. Chem. Soc. 2016, 138, 5403− (98) Cass, A.; Finkelstein, A. Water Permeability of Thin Lipid
5409. Membranes. J. Gen. Physiol. 1967, 50, 1765−1784.
(80) Shen, Y.-x.; Song, W.; Barden, D. R.; Ren, T.; Lang, C.; Feroz, H.; (99) Holz, R.; Finkelstein, A. The Water and Nonelectrolyte
Henderson, C. B.; Saboe, P. O.; Tsai, D.; Yan, H.; Butler, P. J.; Bazan, G. Permeability Induced in Thin Lipid Membranes by the Polyene
C.; Phillip, W. A.; Hickey, R. J.; Cremer, P. S.; Vashisth, H.; Kumar, M. Antibiotics Nystatin and Amphotericin B. J. Gen. Physiol. 1970, 56,
Achieving High Permeability and Enhanced Selectivity for Angstrom- 125−145.
Scale Separations Using Artificial Water Channel Membranes. Nat. (100) Mueller, P.; Rudin, D. O.; Tien, H. T.; Wescott, W. C. Methods
Commun. 2018, 9, 2294. for the Formation of Single Bimolecular Lipid Membranes in Aqueous
(81) Shen, Y.-x.; Si, W.; Erbakan, M.; Decker, K.; De Zorzi, R.; Saboe, Solution. J. Phys. Chem. 1963, 67, 534−535.
P. O.; Kang, Y. J.; Majd, S.; Butler, P. J.; Walz, T.; Aksimentiev, A.; Hou, (101) Tu, Y.-M.; Song, W.; Ren, T.; Shen, Y.-x.; Chowdhury, R.;
J.-L.; Kumar, M. Highly Permeable Artificial Water Channels That Can Rajapaksha, P.; Culp, T. E.; Samineni, L.; Lang, C.; Thokkadam, A.;
Self-Assemble into Two-Dimensional Arrays. Proc. Natl. Acad. Sci. U. S. et al. Rapid Fabrication of Precise High-Throughput Filters from
A. 2015, 112, 9810−9815. Membrane Protein Nanosheets. Nat. Mater. 2020, 19, 347−354.
(82) Song, W.; Joshi, H.; Chowdhury, R.; Najem, J. S.; Shen, Y.-x.; (102) Lu, X.; Gabinet, U. R.; Ritt, C.; Feng, X.; Deshmukh, A.;
Lang, C.; Henderson, C. B.; Tu, Y.-M.; Farell, M.; Pitz, M. E.; Maranas, Kawabata, K.; Kaneda, M.; Hashmi, S. M.; Osuji, C. O.; Elimelech, M.
C. D.; Cremer, P. S.; Hickey, R. J.; Sarles, S. A.; Hou, J.-L.; Aksimentiev, Relating Selectivity and Separation Performance of Lamellar Two-
A.; Kumar, M. Artificial Water Channels Enable Fast and Selective Dimensional Molybdenum Disulfide (MoS2) Membranes to Nano-
Water Permeation through Water-Wire Networks. Nat. Nanotechnol. sheet Stacking Behavior. Environ. Sci. Technol. 2020, 54, 9640−9651.
2020, 15, 73−79. (103) Lang, C.; Ye, D.; Song, W.; Yao, C.; Tu, Y.-m.; Capparelli, C.;
(83) Park, H. B.; Kamcev, J.; Robeson, L. M.; Elimelech, M.; Freeman, LaNasa, J. A.; Hickner, M. A.; Gomez, E. W.; Gomez, E. D.; et al.
B. D. Maximizing the Right Stuff: The Trade-Off between Membrane Biomimetic Separation of Transport and Matrix Functions in Lamellar
Permeability and Selectivity. Science 2017, 356, eaab0530. Block Copolymer Channel-Based Membranes. ACS Nano 2019, 13,
(84) Shen, J.; Ye, R.; Romanies, A.; Roy, A.; Chen, F.; Ren, C.; Liu, Z.; 8292−8302.
Zeng, H. Aquafoldmer-Based Aquaporin-Like Synthetic Water (104) Hladky, S. B.; Haydon, D. A. Ion Transfer across Lipid
Channel. J. Am. Chem. Soc. 2020, 142, 10050−10058. Membranes in the Presence of Gramicidin A: I. Studies of the Unit
(85) Saparov, S. M.; Pfeifer, J. R.; Al-Momani, L.; Portella, G.; de Conductance Channel. Biochim. Biophys. Acta, Biomembr. 1972, 274,
Groot, B. L.; Koert, U.; Pohl, P. Mobility of a One-Dimensional 294−312.
Confined File of Water Molecules as a Function of File Length. Phys. (105) Chan, W.-F.; Chen, H.-y.; Surapathi, A.; Taylor, M. G.; Shao, X.;
Rev. Lett. 2006, 96, 148101. Marand, E.; Johnson, J. K. Zwitterion Functionalized Carbon
(86) Harroun, T. A.; Heller, W. T.; Weiss, T. M.; Yang, L.; Huang, H. Nanotube/Polyamide Nanocomposite Membranes for Water Desali-
W. Experimental Evidence for Hydrophobic Matching and Membrane- nation. ACS Nano 2013, 7, 5308−5319.

10915 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916
ACS Nano www.acsnano.org Review

(106) He, Y.; Hoi, H.; Abraham, S.; Montemagno, C. D. Highly (128) Kedem, O.; Katchalsky, A. Thermodynamic Analysis of the
Permeable Biomimetic Reverse Osmosis Membrane with Amphiphilic Permeability of Biological Membranes to Non-Electrolytes. Biochim.
Peptide Stabilized Aquaporin as Water Filtering Agent. J. Appl. Polym. Biophys. Acta 1958, 27, 229−246.
Sci. 2018, 135, 46169. (129) Levitt, D. G.; Mlekoday, H. J. Reflection Coefficient and
(107) Jeong, B.-H.; Hoek, E. M. V.; Yan, Y.; Subramani, A.; Huang, X.; Permeability of Urea and Ethylene Glycol in the Human Red Cell
Hurwitz, G.; Ghosh, A. K.; Jawor, A. Interfacial Polymerization of Thin Membrane. J. Gen. Physiol. 1983, 81, 239−253.
Film Nanocomposites: A New Concept for Reverse Osmosis (130) Illsley, N. P.; Verkman, A. S. Serial Permeability Barriers to
Membranes. J. Membr. Sci. 2007, 294, 1−7. Water Transport in Human Placental Vesicles. J. Membr. Biol. 1986, 94,
(108) Auerbach, A.; Sachs, F. Patch Clamp Studies of Single Ionic 267−278.
Channels. Annu. Rev. Biophys. Bioeng. 1984, 13, 269−302. (131) Meyer, M. M.; Verkman, A. S. Human Platelet Osmotic Water
(109) Sakmann, B.; Neher, E. Single-Channel Recording, 2nd ed.; and Nonelectrolyte Transport. Am. J. Physiol.: Cell Physiol. 1986, 251,
Springer Science & Business Media: New York, 1995. C549−C557.
(110) Colquhoun, D.; Hawkes, A. G.; Katz, B. Relaxation and (132) Chen, P. Y.; Pearce, D.; Verkman, A. S. Membrane Water and
Fluctuations of Membrane Currents That Flow through Drug- Solute Permeability Determined Quantitatively by Self-Quenching of
Operated Channels. Proc. R. Soc. London, Ser. B 1977, 199, 231−262. an Entrapped Fluorophore. Biochemistry 1988, 27, 5713−5718.
(111) Colquhoun, D.; Hawkes, A. G. On the Stochastic Properties of (133) Harris, H. W., Jr.; Handler, J. S.; Blumenthal, R. Apical
Single Ion Channels. Proc. R. Soc. London, Ser. B 1981, 211, 205−235. Membrane Vesicles of Adh-Stimulated Toad Bladder Are Highly Water
(112) Colquhoun, D.; Hawkes, A. G. On the Stochastic Properties of Permeable. Am. J. Physiol.: Renal Physiol. 1990, 258, F237−F243.
Bursts of Single Ion Channel Openings and of Clusters of Bursts. Philos. (134) Zeidel, M. L.; Albalak, A.; Grossman, E.; Carruthers, A. Role of
Glucose Carrier in Human Erythrocyte Water Permeability.
Trans. R. Soc., B 1982, 300, 1−59.
Biochemistry 1992, 31, 589−596.
(113) Dionne, V. E.; Leibowitz, M. D. Acetylcholine Receptor
(135) Verkman, A. S.; Dix, J. A.; Seifter, J. L. Water and Urea
Kinetics. A Description from Single-Channel Currents at Snake
Transport in Renal Microvillus Membrane Vesicles. Am. J. Physiol.:
Neuromuscular Junctions. Biophys. J. 1982, 39, 253−261.
Renal Physiol. 1985, 248, F650−F655.
(114) Horn, R.; Lange, K. Estimating Kinetic Constants from Single
(136) van Heeswijk, M. P.; van Os, C. H. Osmotic Water
Channel Data. Biophys. J. 1983, 43, 207−223. Permeabilities of Brush Border and Basolateral Membrane Vesicles
(115) Kijima, S.; Kijima, H. Statistical Analysis of Channel Current from Rat Renal Cortex and Small Intestine. J. Membr. Biol. 1986, 92,
from a Membrane Patch Ii. A Stochastic Theory of a Multi-Channel 183−93.
System in the Steady-State. J. Theor. Biol. 1987, 128, 435−455. (137) Hannesschläger, C.; Barta, T.; Siligan, C.; Horner, A.
(116) Yeo, G. F.; Edeson, R. O.; Milne, R. K.; Madsen, B. W. Quantification of Water Flux in Vesicular Systems. Sci. Rep. 2018, 8,
Superposition Properties of Independent Ion Channels. Proc. R. Soc. 8516.
London, Ser. B 1989, 238, 155−170. (138) Paula, S.; Volkov, A. G.; Van Hoek, A. N.; Haines, T. H.;
(117) Pohl, P.; Saparov, S. M.; Antonenko, Y. N. The Effect of a Deamer, D. W. Permeation of Protons, Potassium Ions, and Small Polar
Transmembrane Osmotic Flux on the Ion Concentration Distribution Molecules through Phospholipid Bilayers as a Function of Membrane
in the Immediate Membrane Vicinity Measured by Microelectrodes. Thickness. Biophys. J. 1996, 70, 339−348.
Biophys. J. 1997, 72, 1711−1718. (139) Henis, J. M. S.; Tripodi, M. K. Composite Hollow Fiber
(118) Pohl, P.; Saparov, S. M. Solvent Drag across Gramicidin Membranes for Gas Separation: The Resistance Model Approach. J.
Channels Demonstrated by Microelectrodes. Biophys. J. 2000, 78, Membr. Sci. 1981, 8, 233−246.
2426−2434. (140) Werber, J. R.; Porter, C. J.; Elimelech, M. A Path to
(119) Walter, A.; Gutknecht, J. Monocarboxylic Acid Permeation Ultraselectivity: Support Layer Properties to Maximize Performance
through Lipid Bilayer Membranes. J. Membr. Biol. 1984, 77, 255−264. of Biomimetic Desalination Membranes. Environ. Sci. Technol. 2018, 52,
(120) Zhao, Y.; Vararattanavech, A.; Li, X.; HélixNielsen, C.; Vissing, 10737−10747.
T.; Torres, J.; Wang, R.; Fane, A. G.; Tang, C. Y. Effects of (141) Filmtec Xle-440i Element, Filmtec Sw30xle−440i Element Product
Proteoliposome Composition and Draw Solution Types on Separation Data Sheet. FILMTEC portfolio August, 2019; DuPont: Wilmington,
Performance of Aquaporin-Based Proteoliposomes: Implications for DE, 2019 (accessed 2019-10-21).
Seawater Desalination Using Aquaporin-Based Biomimetic Mem- (142) Iwahashi, H.; Taniguchi, M.; Ito, Y.; Maeda, T.; Fujii, Y.; Linke,
branes. Environ. Sci. Technol. 2013, 47, 1496−1503. P.; Al-Thani, H. A. J.; Albeldawi, M. Advanced Ro System for High
(121) Li, X.; Wang, R.; Wicaksana, F.; Zhao, Y.; Tang, C.; Torres, J.; Temperature and High Concentration Seawater Desalination at the
Fane, A. G. Fusion Behaviour of Aquaporin Z Incorporated Arabian Gulf. Proceedings from the IDA 2015 World Congress, August
Proteoliposomes Investigated by Quartz Crystal Microbalance with 30-September 4, 2015, San Diego, California; Caribbean Desalination
Dissipation (QCM-D). Colloids Surf., B 2013, 111, 446−452. Association: Stuart, FL, 2015.
(122) Grzelakowski, M.; Cherenet, M. F.; Shen, Y.-x.; Kumar, M. A (143) Porter, C. J.; Werber, J. R.; Ritt, C. L.; Guan, Y.-F.; Zhong, M.;
Framework for Accurate Evaluation of the Promise of Aquaporin Based Elimelech, M. Controlled Grafting of Polymer Brush Layers from
Biomimetic Membranes. J. Membr. Sci. 2015, 479, 223−231. Porous Cellulosic Membranes. J. Membr. Sci. 2020, 596, 117719.
(123) Li, X.; Chou, S.; Wang, R.; Shi, L.; Fang, W.; Chaitra, G.; Tang, (144) Kowal, J.Ł.; Kowal, J. K.; Wu, D.; Stahlberg, H.; Palivan, C. G.;
C. Y.; Torres, J.; Hu, X.; Fane, A. G. Nature Gives the Best Solution for Meier, W. P. Functional Surface Engineering by Nucleotide-Modulated
Desalination: Aquaporin-Based Hollow Fiber Composite Membrane Potassium Channel Insertion into Polymer Membranes Attached to
with Superior Performance. J. Membr. Sci. 2015, 494, 68−77. Solid Supports. Biomaterials 2014, 35, 7286−7294.
(124) Verkman, A. S. Optical Methods to Measure Membrane (145) Sagle, A. C.; Van Wagner, E. M.; Ju, H.; McCloskey, B. D.;
Transport Processes. J. Membr. Biol. 1995, 148, 99−110. Freeman, B. D.; Sharma, M. M. Peg-Coated Reverse Osmosis
(125) Gerbeau, P.; Gücļ ü, J.; Ripoche, P.; Maurel, C. Aquaporin Nt- Membranes: Desalination Properties and Fouling Resistance. J.
Tipa Can Account for the High Permeability of Tobacco Cell Vacuolar Membr. Sci. 2009, 340, 92−108.
Membrane to Small Neutral Solutes. Plant J. 1999, 18, 577−587.
(126) Mathai, J. C.; Sprott, G. D.; Zeidel, M. L. Molecular
Mechanisms of Water and Solute Transport across Archaebacterial
Lipid Membranes. J. Biol. Chem. 2001, 276, 27266−27271.
(127) Lande, M. B.; Donovan, J. M.; Zeidel, M. L. The Relationship
between Membrane Fluidity and Permeabilities to Water, Solutes,
Ammonia, and Protons. J. Gen. Physiol. 1995, 106, 67−84.

10916 https://dx.doi.org/10.1021/acsnano.0c05753
ACS Nano 2020, 14, 10894−10916

You might also like