RESEARCH ARTICLE

A JH
A biomimetic microfluidic chip to study the circulation and
mechanical retention of red blood cells in the spleen
Julien Picot,1,2,3,4 Papa Alioune Ndour,4,5 Sophie D. Lefevre,1,2,3,4 Wassim El Nemer,1,2,3,4 Harvey Tawfik,6
Julie Galimand,4,7 Lydie Da Costa,3,4,7,8 Jean-Antoine Ribeil,4,9,10,11 Mariane de Montalembert,10,12
Valentine Brousse,1,2,3,4,10,12 Bruno Le Pioufle,6 Pierre Buffet,4,5 Caroline Le Van Kim,1,2,3,4 and Olivier Français6*

Red blood cells (RBCs) are deformable and flow through vessels narrower than their own size. Their
deformability is most stringently challenged when they cross micrometer-wide slits in the spleen. In several
inherited or acquired RBC disorders, blockade of small vessels by stiff RBCs can trigger organ damage, but
a functional spleen is expected to clear these abnormal RBCs from the circulation before they induce such
complications. We analyzed flow behavior of RBCs in a microfluidic chip that replicates the mechanical
constraints imposed on RBCs as they cross the human spleen. Polymer microchannels obtained by soft
lithography with a hydraulic diameter of 25 lm drove flow into mechanical filtering units where RBCs flew
either slowly through 5- to 2-lm-wide slits or rapidly along 10-lm-wide channels, these parallel paths
mimicking the splenic microcirculation. Stiff heated RBCs accumulated in narrow slits seven times more
frequently than normal RBCs infused simultaneously. Stage-dependent retention of Plasmodium falciparum-
infected RBCs was also observed in these slits. We also analyzed RBCs from patients with hereditary
spherocytosis and observed retention for those having the most altered mechanical properties as
determined by ektacytometry. Thus, in keeping with previous observations in vivo and ex vivo, the chip
successfully discriminated poorly deformable RBCs based on their distinct mechanical properties and on
the intensity of the cell alteration. Applications to the exploration of the pathogenesis of malaria, hereditary
spherocytosis, sickle cell disease and other RBC disorders are envisioned.
Am. J. Hematol. 90:339–345, 2015. V
C 2015 Wiley Periodicals, Inc.

䊏 Introduction
Mature red blood cells (RBCs) are non-nucleated, biconcave disc-shaped cells of 7–8 lm in diameter, 2-lm-thick with an average volume of
80–100 lm3 [1]. Normal circulating RBCs are highly elastic and deformable and flow through capillaries narrower than their own size. When
RBC deformability is altered there is a risk of RBC blockade in small capillaries [2]. However, the most stringent mechanical challenge takes place
in the spleen when RBCs cross narrow inter-endothelial slits in the wall of red pulp sinuses [3,4]. In theory, a functional spleen clears the blood
from mechanically altered RBCs before they are trapped in microvessels, where they would create tissue damage. Altered mechanical properties
can be originated by multiple inherited RBC disorders, such as spherocytosis or sickle cell disease [5], or by infectious agents such as Plasmodium
falciparum [6]. Although these diseases result from very different molecular or cellular alterations they share the generic mechanism of poorly
deformable RBCs triggering complications, contributing for example to coma during severe Plasmodium falciparum malaria and anemia in heredi-
tary spherocytosis (HS) [7].
An important technical gap to study the pathophysiological impact of impaired RBC circulation is a device mimicking the dimensions of key
microcirculatory structures [8]. A physiologically relevant device should mimic 4- to 10-lm-wide small vessels (capillaries and small venules), and
0.5- to 3-lm-wide microcirculatory beds in the spleen [3,9]. Inter-endothelial slits in human spleens are typically in the 1-lm-wide range, as
measured by transmission electron microscopy [3]. To mimic the RBC retention rates encountered in the human spleen a device should thus

Additional Supporting Information may be found in the online version of this article.
1
Institut National De La Transfusion Sanguine, Paris, F-75739, France; 2InsermUMR_S1134, Paris, FranceF-75739; 3Universite Paris Diderot, Sorbonne Paris Cite, Paris,
France; 4Laboratory of Excellence GR-Ex, Paris, France; 5Inserm, U1135/Paris 6, Paris, FranceF-75634; 6SATIECNRS UMR8029, Ecole Normale Superieure De Cachan,
Cachan, FranceF-94235; 7AP-HP, Service Hematologie BiologiqueH^ opital RDebre, Paris, FranceF-75935; 8Inserm, U1149, Paris 7, Paris, FranceF-75018; 9Inserm,
UMR1163, Paris, FranceF-75743; 10Universite Paris Descartes, Sorbonne Paris Cite, Paris, France; 11Biotherapy DepartmentH^ opital Universitaire Necker Enfants
MaladesAPHP, Paris, France; 12Reference Centre for Sickle Cell Disease, Pediatric DepartmentH^ opital Universitaire Necker Enfants MaladesAPHP, Paris, France
Conflict of interest: Nothing to report
J.P., P.A.N., S.D.L., and W.E.N. contributed equally to this work.
Julien Picot is currently at GIP Genopole, Evry, F-91030, France
P.A.N. was supported by the laboratory of Excellence GR-Ex.
*Correspondence to: Olivier Français, SATIE, CNRS UMR8029, Ecole Normale Superieure de Cachan, 61 Avenue du President Wilson, 94235 Cachan, France.
E-mail: olivier.francais@satie.ens-cachan.fr
Contract grant sponsor: Laboratory of Excellence GR-Ex; Contract grant number: ANR-11-LABX-0051.
Contract grant sponsor: “Investissements d’avenir” of the French National Research Agency; Contract grant number: ANR-11-IDEX-0005-02.
Contract grant sponsor: Institute d’Alembert (Guidecells Project); LASIPS Labex (SIMULOS Project); DIM MalInf; the Gates Foundation; Follereau
Foundation.
Received for publication: 20 December 2014; Revised: 6 January 2015; Accepted: 8 January 2015
Am. J. Hematol. 90:339–345, 2015.
Published online: 12 January 2015 in Wiley Online Library (wileyonlinelibrary.com).
DOI: 10.1002/ajh.23941

C 2015 Wiley Periodicals, Inc.

V

doi:10.1002/ajh.23941 American Journal of Hematology, Vol. 90, No. 4, April 2015 339
Picot et al. RESEARCH ARTICLE

Figure 1. Design (A), process flow (B), manufacture (C) of the spleen-like chip. Each filtering unit comprises of fifty-three 2-mm-wide slits between the 15-mm
pillars that form the boundaries of the lattice. The inner pillar lines comprise 12, 13, and 14 slits, the width of which is 5, 4, and 3 mm, from left to right. The U-
shaped filter is 275-mm-wide (including pillars) included a single 320 microscopic field captured by the camera sensor (18.6 3 6.7 mm2) Side channels are
10-mm-wide. The filtering unit is 5-mm-high. Microfluidic channels connecting the eight filtering units are 25-mm-high to minimize hydraulic resistance
between filtering units. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

contain narrow spaces of 2 lm. To replicate the specific shape
deformation of RBCs as they cross narrow splenic slits, in vitro
obstacles should also be <10-lm long. With recent advances in
microtechnology and in particular in microfluidics [10–12], it is now
possible to envision “in vitro” experiments mimicking the microvas-
cular flow in vital organs [13]. Microchannels with dimensions of sev-
eral micrometers molded in PDMS have explored how RBC
deformability may impact the pathogenesis of capillary obstruction
[14], and malaria [15]. Microfluidic devices with wider microchannels
(>30 lm) have opened the way to the real-time analysis of vaso-
occlusion in sickle cell disease [16,17]. Most microfluidic approaches
have quantified RBC deformability based on velocity through devices
that mimic microcirculatory structures [15]. An experimental system
that quantifies retention rather than velocity may better reflect patho-
physiological processes occurring in malaria or other RBC diseases
that are associated with marked clearance of altered RBCs in the
spleen, a process that contributes to the enlargement of the spleen.
A microfiltration device has been recently developed that uses
microspheres to mimic the mechanical sensing of RBCs by the
human spleen [18]. This device provides relevant retention rates for
poorly deformable RBCs [19,20] but does not allow a direct observa-
tion of RBCs as they cross narrow slits. By contrast, most microflui-
dic devices allow a direct morphological and dynamic exploration of
RBCs as they navigate in the chip. RBC deformability is a complex
function involving membrane stiffness, cytoplasmic viscosity and
surface-to-volume ratio [21]. The deformability of P. falciparum-
infected RBCs is even more complex [22]. Fine analysis of the event
as it occurs is likely the most powerful way to decipher the complex
cellular processes with a filtering capacity large enough to allow
robust quantification of subpopulation enrichment. We present here
the design, set-up and validation of a new biomimetic microfluidic
device where the retention of poorly deformable human RBCs can be
robustly quantified and observed.
䊏 Methods
Patients. Small (0.5–1 ml) venous blood samples from 4 patients with hereditary
spherocytosis were retrieved with patient approval from EDTA tubes within 2 hr of
collection in the context of normal medical care. The « Ile de France VI » Institu-
tional Review Board has approved this approach as a non-biomedical research pro-
cess (Article L1121-1 of the French Code for Public Health) that does not affect the
medical care. Venous blood samples from three controls were also obtained using a
similar approach.
Microfluidic device fabrication. The spleen-like chip includes polymer micro-
channels with an average hydraulic diameter of 25 lm, including local mechanical
filtering units with an equivalent behavior of apertures from 5 to 2 lm (Fig. 1A).
Side flow, which is observable in the spleen, is rendered possible in the device, with
side channels added in the filter zone (Fig. 1A). The microfluidic shape that defines
the filtering functions of the device is casted in polydimethylsiloxane (PDMS, Syl-
gard), a silicone elastomer [23]. The procedure to design the device is based on
standard microfabrication techniques associated to molding techniques [24]. We
first fabricated a master using SU-8 photoresist (Microchem, Newton, MA) to
obtain a negative mold on a 4 in. Silicon substrate (Fig. 1B). A two-step phase of
photolithography with UV exposure had been developed. The first photolithogra-
phy defines the microchannels with a uniform thickness of 5 lm and a geometry
corresponding to the repeated mechanical solicitation and filter design. The second
photolithography gives the fluidic accesses with a uniform thickness of 25 lm in
order to reduce the access fluidic resistance. From the SU-8 mold, the casting of
PDMS (ratio 10:1) channels is realized. The chip obtained is then irreversibly
bounded to a microscope cover slide using plasma 02 activation (30 W, 300 mT,
20 s) on both surfaces. Before sealing, access holes were done in the PDMS with a
homemade punch giving female Luer (TM) connection compatibility.
The design is based on eight mechanical filtering units associated in parallel and
connected together with a microchannel network as seen on the mold (Fig. 1C). To
reduce the hydraulic resistance of the full design, the microchannel network is 25-
lm-high compared to the 5-lm height of each filtering unit. The obtained

340 American Journal of Hematology, Vol. 90, No. 4, April 2015 doi:10.1002/ajh.23941
RESEARCH ARTICLE
microfluidic device is fully transparent which makes possible the use of conven-
tional inverted microscopes.
Red blood cell preparation. Cell suspensions: Packed RBCs from the French
blood bank (Etablissement Français du Sang, Rungis, France) were washed twice
with 4 volumes of RMPI 1640 (Life technologies, France) and used for further
preparation.
Heated RBCs: Washed RBCs were suspended in PBS 1% AlbuMAX II (Life
Technologies, France) and then heated for 15 min at 50  C in a water bath fol-
lowed by two washes with PBS 1% AlbuMAX II (Life technologies, France).
PKH labeling: RBCs were stained with PKH 67 fluorescent Cell Linker Kit
(heated or HS RBCs) or PKH 26 fluorescent Cell Linker Kit (control non-heated
RBCs) according to the manufacturer’s instructions (Sigma–Aldrich, France).
Infected RBCs: FUP line Plasmodium falciparum parasites (FCR3) was used for
all experiments. Parasites were cultured following Worldwide Antimalarial Resist-
ance Network (WARN) procedures in 5% hematocrit, at 37  C, with gas mixture:
5% CO2, 10% O2 and 85% N2, and RPMI 1640 media containing Hepes 25
mmol l21, NaHCO3 25 mmol l21, Glutamine 0.3 g l21, Gentamycin 10 mg l21,
and Serum 10%. Cultures were synchronized by sorbitol method (Lambros C, Van-
derberg JP. 1979) and then by VarioMACS Separator (Miltenyi Biotec, France) to
obtain different, tightly synchronized stages of maturation.
Parasites staining: Synchronized ring-infected RBCs (8–16 hr) were stained with
Hoechst dye while trophozoites (26-34h) were stained with SYBR Green (Hoechst
and SYBR Green stain the nucleic acid of DNA, Life technologies, France). RBC
suspensions were prepared with the following composition: 5% of labeled ring-
infected RBCs, 5% of labeled trophozoite-infected RBCs and 90% of control unla-
beled RBCs. A 0.1% hematocrit RBC suspension in PBS 1% AlbuMAX II was pre-
pared with this cell mixture before being perfused in the microfluidic device.
The microfluidic system setup. The microfluidic system consists of the microflui-
dic device, tubing, fluid connectors, syringe and syringe pump (Syramed SP6000,
Arcomed’Ag, Regensdorf, Switzerland) or MFCSTM-EZ microfluidic flow control
system (Fluigent, France). Tubes and fluid connectors were used to connect the
syringe to the chip. The syringe was filled with RBC suspensions and RBCs were
perfused through the microchannel network at a specified flow rate (0.2 ml h21
and a mean of 100 mBar). The device was placed in an inverted Zeiss AxioOb-
server Z1 microscope for monitoring RBC movements. Pictures of the eight filter-
ing units were taken sequentially and at regular time intervals with a high
resolution AxioCam MRm Rev.3 camera and AxioVision 4 analysis software (Carl
Zeiss). Pictures of the whole filtering unit were taken using the 203/0.45 Plan-
Apochromate objective; smaller zones are imaged at higher magnification using the
633/1.40-Oil Plan-Apochromate objective (Carl Zeiss). Blue, green and red fluores-
cent cells were visualized using the Colibri LED 365, 470, and 555 nm modules
(Carl Zeiss), respectively.
Electronic microscopy of a human spleen. For the sizing of the microfluidic chip,
samples of human spleen were processed as described [25] then examined and photo-
graphed with a JEOL JSM 6700F field emission scanning electron microscope operat-
ing at 5 or 7 kV (images were acquired with the upper SE detector on the lower
secondary sector), or with a JEOL 1200 EX electron microscope operating at 80 kV.
Ektacytometry analysis. The ektacytometer [26] (LoRRca MaxSis,
R
MechatronicsV, Hoorn, Netherlands) is a laser diffraction viscometer, in which the
deformation of red cells suspended in a viscous PVP solution at defined values of
applied shear stress of 30 Pa and at a constant temperature measurement of 37  C
are monitored as a continuous function of suspending medium osmolarity. Blood
samples from 4 HS patients and three healthy controls were exposed to increasing
shear stress and an osmotic gradient. Three distinct features of the osmotic gradient
ektacytometry profiles are the Omin, the DImax and the O’hyper points. The
Omin point corresponds to the osmolarity at the minimal deformability in hypoos-
molar area or at the osmolarity when 50% of the red cells hemolyzed during the
regular osmotic resistance test. It reflects the surface to area ratio. DImax corre-
sponds to the maximal deformability index or elongation index (EI). The hyper
point or O’ corresponds to the osmolarity at half of the DImax and reflects the
hydration state of the cells. In case of HS [27,28], the constant and characteristic
features are a decrease in the DImax, in conjunction with a shift of the Omin point
to the right (reduced surface to volume ratio) and a shift of the O’hyper point to
left (increased dehydration of the red cells). The severity is correlated to the level of
the DImax.
䊏 Results
Size of human spleen microcirculation
The dimensions of an interendothelial slit were determined from a
transmission electron microscopy picture (Supporting Information Fig.
S1A). The length and width of this slit were graphically estimated as 1.8–
2 lm and 4.5 lm (Supporting Information Fig. S1B), respectively, i.e.,
similar to previous measurements [3,4]. We used these values for the
width and length of the narrowest slits in the filtering part of the chip.
doi:10.1002/ajh.23941
Red blood cells deformability in a spleen-like-microfluidic-chip
From in silico to in vitro validation of physiological
flow through the chip
The whole dimensions of the microfluidic biomimetic filter were
optimized using 3D numerical calculation based on finite element
analysis (ComsolV C ), with the goal to obtain 80% of the cell flow pass-
ing in the slit-shaped central zone that mimics the slow open micro-
circulation of the human spleen; the 10-mm-wide side channels
mimic the closed fast microcirculatory pathways (Supporting Infor-
mation Fig. S2A).
The microfilter hydraulic resistance, calculated from the numerical
simulation, was 6.13 3 1014 Pa m23 s21. Such a high resistance value
affects considerably the fluidic properties within the device. Indeed, a
cell solution flowing at 1 ll mn21 within the device will induce 100
mbar pressure drop, while the average cell velocity will be 1.1
cm s21. The complete microfluidic structure, including eight filtering
units set in parallel and access channel, presents 2.4 3 1014
Pa m23 s21 as hydraulic resistance value [29].
On Supporting Information Fig. S2B, the velocity flow behavior
inside the chip is represented, close to the filter. It had been obtained
with an applied pressure of 100 Pa. The fluid velocity inside the filtering
unit is different in the inner slit-shaped zone and in the side channels.
The velocity profile (perpendicular to the flow, Supporting Informa-
tion Fig. S2C) obtained in the filter highlights the difference of velocity
value between the side channels and the inner zone of the filtering unit.
The average ratio between these velocities was calculated to be 4.5 with
the chosen geometry, which was further confirmed by following experi-
ments. Such velocity ratio between fast and slow circulations is reminis-
cent of observations in isolated-perfused spleens [3,4].
Mechanical retention of rigid heated red blood cells
The capacity of the microfluidic chip to retain poorly deformable
RBCs was tested. RBC rigidity was induced by heating the cells at 50

C for 15 min. This reduces their size, induces their sphericity and
increases their irreversible stiffness [30]. Heated and control RBCs
were then fluorescently labeled with PKH67 (green) or PKH26 (red),
respectively. A RBC suspension at 0.1% hematocrit, containing equal
amounts of heated and control RBCs (Fig. 2A), was perfused through
the device. RBC retention was monitored over time in the eight filter-
ing units. Pictures were taken at different pre-set time points during
the experiment. For each filtering unit images were captured in the
green, red and brightfield channels and merged. A picture was also
taken before perfusing the RBC suspension in order to confirm that
the filtering units where not blocked by debris (Fig. 2B-T0). Early pic-
tures showed that stiff heated RBCs (green-labeled) but not normal
(red-labeled) RBCs were retained preferentially in the 2 lm-wide slits
(Fig. 2B-T1; T2; T3; T4 and C). The number of stiff heated RBCs
retained in slits increased over perfusion time, with the vast majority
of the 2-lm-wide slits being clogged by these RBCs at the end of the
assay (Fig. 2B-T4).
At the experiment endpoint, retained RBCs present in each zone
were counted and sorted depending on their fluorescence (Fig. 2D).
At the end of the assay, poorly deformable heated RBCs were seven
times more abundant in the filtering unit than control RBCs (Fig.
2D, right panel) and often had a quasi-spherical aspect (Fig. 2C). Fil-
tering units could thus discriminate between deformable and non-
deformable RBCs, and could reveal the presence of subpopulations of
rigid RBCs in a suspension.
Retention of red blood cells parasited by plasmodium
falciparum
The ability of the chips to retain poorly deformable RBCs was fur-
ther tested using RBCs infected with Plasmodium falciparum.
American Journal of Hematology, Vol. 90, No. 4, April 2015 341
Picot et al. RESEARCH ARTICLE

Figure 2. Preferential retention of poorly deformable heated RBCs in slits of the filtering unit: (A) Equal concentrations of PKH26-stained normal (red) and
PKH67-stained poorly deformable heated RBCs (green) in the RBC suspension perfused through the chip. (B) T1!T4 progressive accumulation of poorly
deformable heated RBCs in slits: 4 (T1), 8 (T2), 12 (T3), 16 (T4) minutes after initiation of RBC circulation through the filtering unit (flow is from left to right).
(C) Ellipsoid or quasi spherical aspect of heated RBCs retained in 2-mm-wide slits (flow is from left to right, red arrows). (D) Number of poorly deformable
heated RBCs (green) and normal RBCs (red) retained in narrow slits of each of the eight filtering units in a chip, and mean values (right panel). [Color figure
can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

P. falciparum proteins (RESA, KAHRP, PfEMP3) exported at the
membrane of infected RBCs contribute, among other factors [22,25]
to their stage-dependent decrease in deformability [31,32]. To test the
differential retention of infected RBCs, we perfused the biochip with
a cell suspension at 0.1% hematocrit containing 3 different RBC pop-
ulations: 5% of blue-labeled ring-infected RBCs, 5% of green-labeled
trophozoite-infected RBCs and 90% of control non-infected unlabeled
RBCs (Fig. 3A).
RBC retention was monitored every 4 min in the eight filtering
units of the chip. For each filtering unit images were captured in the
blue, green and brightfield channels and merged. They showed that
infected RBCs (blue- and green-labeled RBCs) were preferentially
retained in the slits as compared to non-infected control RBCs (unla-
beled) (Fig. 3B,C). We analyzed the retained infected RBC sub-
populations and found that the mature trophozoites (26–34 hr of
maturation) accounted for a higher proportion of retained RBCs than
young ring-infected RBCs (8–16 hr of maturation after reinvasion).
Indeed, the retained RBCs were composed of 85% (215/253) of
green-labeled trophozoites and 15% (38/253) of blue-labeled ring-
infected RBCs (Fig. 3D). Furthermore, during the first minutes of
perfusion, the ring-infected RBCs were preferentially retained in the
2-mm slits while trophozoites were retained in slits of all sizes (from
2 to 5 mm), consistent with the differences in the RBC mechanical
properties between these two infection stages.
Thus, despite the relatively low proportion of infected RBCs in the
initial RBC suspension, the spleen-like microfluidic device successfully
discriminated infected RBCs from non-infected RBCs, as well as
young-infected from the late-infected RBCs, based on their distinct
mechanical properties and on the intensity of the mechanical
alteration.
Retention of red blood cells from patients with
hereditary spherocytosis
We next explored another pathophysiological model using RBCs
from patients with hereditary spherocytosis (HS). In HS, RBCs are
poorly deformable because of mutations in RBC membrane proteins

342 American Journal of Hematology, Vol. 90, No. 4, April 2015 doi:10.1002/ajh.23941
RESEARCH ARTICLE Red blood cells deformability in a spleen-like-microfluidic-chip

Figure 3. Stage-dependent retention of Plasmodium falciparum-infected RBCs in narrow slits. (A) Image capture of RBC suspension at 0.1% hematocrit con-
taining subpopulations of young ring-infected RBCs (rings) and late trophozoite-infected RBC, each at 5% parasitemia. Ring-infected RBCs were stained
with Hoechst dye (blue) and triphozoites were stained with SYBR Green (green). (B) Retention pattern of infected RBCs in the chip at the end of the experi-
ment. (C) High magnification of infected RBCs trapped in 2-mm-slits (flow in the chip is from bottom to top, red arrows). (D) Number of rings- or trophozoite-
infected RBC trapped according to the size of the slits at the end of experiment. [Color figure can be viewed in the online issue, which is available at
wileyonlinelibrary.com.]

that affect the vertical interactions between the lipid bilayer and the
membrane skeleton, negatively impacting membrane cohesion and
leading to surface loss.
We analyzed RBC samples from 4 HS patients. HS and control
RBCs were fluorescently labeled with PKH67 (green) and PKH26
(red), respectively, and perfused in the microfluidic chip as described
above for heated RBCs. RBCs of two patients (P1 and P2) were
retained in the chip, preferentially in 2-mm slits (Fig. 4A), while those
from the 2 other patients (P3 and P4) were not. We hypothesized
that this difference resulted from the known heterogeneity among HS
patients. As a matter of fact, HS patients exhibit heterogeneous phe-
notypes depending on their mutated protein(s), the disease severity
being primarily dependent on the extent of membrane surface area
loss [33,34]. We evaluated the deformability and the surface area loss
of the 4 HS RBC samples by osmotic gradient ektacytometry profiles,
measuring three parameters: [1] the maximal deformability index
(DImax) value, which represents the maximal cellular deformability
of the RBC population and [2] the Omin point, which corresponds to
the osmolarity at which 50% of the RBCs are lysed in the classical
osmotic fragility test and reflects the reduction in surface area to vol-
ume ratio of the RBC population being analyzed (a shift to the right
of the Omin corresponding to a reduced surface-to-volume ratio),
and [3] the O’ or the hyper point, which reflects the hydration state
of the red cells. As expected, in conjunction with the presence of
spherocytes on the blood smears, all HS RBC samples exhibited the
classical features of HS when compared with control RBCs (Fig. 4B).
In addition, the two RBC samples from P1 and P2, that were retained
in the microfluidic chip, showed lower DImax and higher Omin than
those from P3 and P4 (Fig. 4B), indicating that these RBCs were less
deformable and more spherical (reduced surface-to-volume ratio)
than the two others and that the microfluidic chip only retained HS
RBCs displaying the greatest phenotypic alterations.
䊏 Discussion
We developed a simple microfluidic device to study and quantify
the mechanical ability of RBCs to cross the human spleen. We
favored a simple chip design that would reproduce key microcircula-
tory structures of the human spleen and display a biomimetic behav-
ior. These include unique, narrow filtering slits localized at the exit of
a slow circulatory pathway, and a wider fast circulatory pathway into
which RBCs are derived when slits are saturated. This fluidic architec-
ture was initially designed with numerical simulations using 3D finite
element analysis. Soft-lithography molding techniques were precise
enough to generate PDMS pillars delineating slits as narrow as 2 mm,
the channel height being 5 mm. Relatively large slits (3- to 5-lm
wide) are predicted to exert mechanical constraints on RBCs similar
to those generated by the tortuous structures of the cords in the sple-
nic red pulp (also called reticular meshwork [35]). The narrowest 2
mm-wide slits mimic the last and most stringent mechanical obstacle
that RBCs must cross on their way from the cords back to the venous
circulation [18]. The width of side channels mimicking the fast circu-
lation were set at 10 mm to allow RBC circulation with minimal
mechanical constraints. A biomimetic spleen-like device has been
recently reported, beautifully mimicking the closed-fast and the open-
slow microcirculations, the latter ending by a row of a dozen of paral-
lel 2-lm constrictions resembling the inter-endothelial slits. Dynamic
crossing could be observed but the number of slits was too limited to
robustly quantify enrichment of poorly deformable subpopulations
[36]. Our device includes eight filtering units per chip, each

doi:10.1002/ajh.23941 American Journal of Hematology, Vol. 90, No. 4, April 2015 343
Picot et al.
Figure 4. Microfluidic and ektacytometry analyses of RBCs from heredi-
tary sperocytosis patients. (A) RBC suspensions with equal concentrations
of PKH26-stained control (red) and PKH67-stained HS RBCs (green, left
panels) were perfused through the chip. HSRBCs from patients P1 and P2
were preferentially retained in the filtering units (right panel). (B) Ektacy-
tometry curves of four HS patients (P1, P2, P3, and P4) and three control
(C1, C2, and C3) and a table showing the values of Omin, Dlmax, and
O’Hyper for each RBC sample. [Color figure can be viewed in the online
issue, which is available at wileyonlinelibrary.com.]
containing more than fifty 2-lm-wide slits. This offers 400 slits into
which mechanical retention can be quantified with statistical robust-
ness. Wider slits upstream from these slits allow an estimate of the
intensity of RBC stiffness; mildly altered RBCs are retained only in 2-
lm slits while markedly altered RBCs accumulate in slits of all sizes.
Using hydrogel micro-particles that mimic the mechanical behavior
of RBCs [37] a more precise determination of retention thresholds
should be achievable.
The pathophysiological relevance of this spleen-like microfluidic
device was established using several subpopulations of poorly deform-
able RBCs. A marked accumulation in very stiff heated RBCs was
observed in slits, a result perfectly consistent with the previous obser-
vation of their clearance from the peripheral blood and accumulation
in the spleen of patients [38]. Retention of RBCs parasitized by early
and late asexual stages of P. falciparum was then analyzed. A selective
retention of both parasitized subpopulations was observed, albeit with
a stage-dependent variation in intensity. While RBCs parasitized by
mature trophozoite stages were, as expected, retained in slits of all
sizes (2–5 mm), RBCs parasitized by ring stages were retained almost
344 American Journal of Hematology, Vol. 90, No. 4, April 2015
RESEARCH ARTICLE
exclusively in the narrowest 2-lm-wide slits that mimic inter-
endothelial splenic slits in human spleens. In full agreement with
these results, complete and partial retention of trophozoite- and ring-
infected RBCs (respectively) had been previously observed in an ex
vivo human spleen perfusion [25]. Interestingly, ring-infected RBCs—
the deformability of which is mildly altered [39]—were retained only
partially in human spleens [29] and accumulated selectively upstream
from interendothelial slits. Behavior of poorly deformable RBCs from
different subpopulations in the chip thus corresponded to expecta-
tions from previous observations in vivo and ex vivo. This biomimetic
behavior was further confirmed by the ability of the chip to accumu-
late poorly deformable RBCs from HS patients. In this genetic heredi-
tary disease patient RBCs exhibit a wide spectrum of altered
mechanical properties because of mutations in proteins that sustain
the interactions between the lipid bilayer and the skeleton [33,34].
The direct consequence of such alterations is a surface area loss of
variable extent with a direct impact on the RBC surface-to-volume
ratio that was shown to play a critical role in RBC splenic entrap-
ment. This had been demonstrated using a model of normal human
spleens perfused ex vivo with RBCs displaying various degrees of sur-
face area loss after treatment with increasing concentrations of lyso-
phosphatidylcholine [40]. The study showed splenic entrapment for
RBCs with > 18% of surface area loss (>27% reduced surface area-to-
volume ratio) and identified an adaptive volume regulation mecha-
nism that allows spherocytic RBCs to prolong their life span in circu-
lation by loosing volume thus escaping splenic retention. Partial
retention of RBCs from HS patients had also been observed in this
human spleen model [18]. In agreement with these observations our
chip discriminated between HS RBC samples by trapping those with
the lowest surface-to-area ratio (corresponding to the highest Omin)
confirming its spleen-like function and emphasizing its potential use
as a sensor of poorly deformable RBCs in the blood circulation. Fur-
ther technical improvement is envisioned whereby high-speed video
will analyze the fate of the first few hundred RBCs entering the chip.
This complementary approach will allow expression of results as
retention rates, in a chip not yet saturated by retained poorly deform-
able RBCs.
The spleen-like microfluidic device mimicking the RBC filtering
function is a promising tool for the exploration of RBC disorders
associated with mechanical alterations. Moreover it provides precious
real-time information of the RBC mechanical behavior in highly con-
strained microfluidic. Indeed, the RBC mechanical rigidity and mem-
brane abnormalities contribute to occlusion of small vessels and
splenic microcirculatory beds [41]. In RBC disorders such as sickle
cell disease, pathological interactions between blood cells, endothelial
cells and proteins of the extracellular matrix lead to vaso-occlusion
events and splenic sequestration crises. To date, very few reports have
designed chip models to investigate microcirculatory behavior of
RBCs in relevant geometrical contexts. Successful experiments in
three complementary pathophysiological situations establish the rele-
vance of the device and show the robustness of a quantified approach
of RBC retention. This opens the way for future fruitful experiments
and applications. Another advantage of the spleen-like chip resides in
its ability to be functionalized with purified proteins or endothelial
cells and adapted to fit with specific pathological disorders in which
RBCs exhibit abnormal deformability.
Finally, in addition to applications in the RBC field our device
opens new perspectives to investigate nuclear rheology and the traf-
ficking of nucleated cells across microporous barriers. Nuclear viscoe-
lasticity was shown to be dependent on the expression level and the
ratio between lamins A and B and might account for the retention of
hematopoietic cells, such as erythroblasts and megakaryocytes in the
bone marrow [42]. Furthermore, the entrapment of mesenchymal
stem cells in the lungs following their intravenous administration [43]
doi:10.1002/ajh.23941
RESEARCH ARTICLE
might partly result from mechanical constraints and this could be
evaluated and measured using our microfluidic device.
䊏 References
1. Bessis M. Red cell shapes. An illustrated classifi-
cation and its rationale. Nouv Rev Fr Hematol
1972;12:721–745.
2. Chien S. Red cell deformability and its relevance
to blood flow. Annu Rev Physiol 1987;49:177–
192.
3. Buffet PA, Safeukui I, Deplaine G, et al. The
pathogenesis of plasmodium falciparum malaria
in humans: Insights from splenic physiology.
Blood 2011;117:381–392.
4. MacDonald IC, Schmidt EE, Groom AC. The
high splenic hematocrit: A rheological conse-
quence of red cell flow through the reticular
meshwork. Microvasc Res 1991;42:60–76.
5. Ballas SK, Larner J, Smith ED, et al. Rheologic
predictors of the severity of the painful sickle
cell crisis. Blood 1988;72:1216–1223.
6. Suwanarusk R, Cooke BM, Dondorp AM, et al.
The deformability of red blood cells parasitized
by plasmodium falciparum and P. vivax. J Infect
Dis 2004;189:190–194.
7. Dondorp AM, Kager PA, Vreeken J, et al.
Abnormal blood flow and red blood cell
deformability in severe malaria. Parasitol Today
2000;16:228–232.
8. Rosenbluth MJ, Lam WA, Fletcher DA. Analyz-
ing cell mechanics in hematologic diseases with
microfluidic biophysical flow cytometry. Lab
Chip 2008;8:1062–1070.
9. Chen LT, Weiss L. The role of the sinus wall in
the passage of erythrocytes through the spleen.
Blood 1973;41:529–537.
10. Abgrall P, Gue A-M. Lab-on-chip technologies:
Making a microfluidic network and coupling it
into a complete microsystem—A review.
J Micromech Microeng 2007;17:R15–R49.
11. Chin CD, Linder V, Sia SK. Commercialization
of microfluidic point-of-care diagnostic devices.
Lab Chip 2012;12:2118–2134.
12. Domachuk P, Tsioris K, Omenetto FG, et al.
Bio-microfluidics: Biomaterials and biomimetic
designs. Adv Mater 2010;22:249–260.
13. Huh D, Torisawa Y-S, Hamilton GA, et al.
Microengineered physiological biomimicry:
Organs-on-chips. Lab Chip 2012;12:2156–2164.
14. Shelby JP, White J, Ganesan K, et al. A micro-
fluidic model for single-cell capillary obstruction
by plasmodium falciparum-infected erythrocytes.
Proc Natl Acad Sci USA 2003;100:14618–14622.
15. Bow H, Pivkin IV, Diez-Silva M, et al. A micro-
fabricated deformability-based flow cytometer
with application to malaria. Lab Chip 2011;11:
1065–1073.
16. Higgins JM, Eddington DT, Bhatia SN, et al.
Sickle cell vasoocclusion and rescue in a micro-
fluidic device. Proc Natl Acad Sci USA 2007;
104:20496–20500.
doi:10.1002/ajh.23941
Red blood cells deformability in a spleen-like-microfluidic-chip
䊏 Acknowledgments
The authors thank Selin Topçu for her help during the fabrication
phase of the biodevices.
17. Tsai M, Kita A, Leach J, et al. In vitro modeling
of the microvascular occlusion and thrombosis
that occur in hematologic diseases using micro-
fluidic technology. J Clin Invest 2012;122:408–
418.
18. Deplaine G, Safeukui I, Jeddi F, et al. The sens-
ing of poorly deformable red blood cells by the
human spleen can be mimicked in vitro. Blood
2011;117:e88–e95.
19. Sanyal S, Egee S, Bouyer G, et al. Plasmodium
falciparum STEVOR proteins impact erythrocyte
mechanical properties. Blood 2012;119:e1–e8.
20. Tiburcio M, Niang M, Deplaine G, et al. A
switch in infected erythrocyte deformability at
the maturation and blood circulation of plasmo-
dium falciparum transmission stages. Blood
2012;199:e172–e180.
21. Mohandas N, Groner W. Cell membrane and
volume changes during red cell development
and aging. Ann N Y Acad Sci 1989;554:217–
224.
22. Safeukui I, Buffet PA, Perrot S, et al. Surface
area loss and increased sphericity account for
the splenic entrapment of subpopulations of
plasmodium falciparum ring-infected erythro-
cytes. PLoS One 2013;8:e60150.
23. McDonald JC, Whitesides GM. Poly (dimethyl-
siloxane) as a material for fabricating microflui-
dic devices. Acc Chem Res 2002;35:491–499.
24. Eddings M, Johnson M, Gale BK. Determining
the optimal PDMS–PDMS bonding technique
for microfluidic devices. J Micromech Microeng
2008;18:067001.
25. Safeukui I, Correas JM, Brousse V, et al. Reten-
tion of plasmodium falciparum ring-infected
erythrocytes in the slow, open microcirculation
of the human spleen. Blood 2008;112:2520–
2528.
26. Mohandas N, Clark MR, Jacobs MS, et al. Ekta-
cytometric analysis of factors regulating red cell
deformability. Blood Cells 1980;6:329–334.
27. Cynober T, Mohandas N, Tchernia G. Red cell
abnormalities in hereditary spherocytosis: Rele-
vance to diagnosis and understanding of the
variable expression of clinical severity. J Lab
Clin Med 1996;128:259–269.
28. Da Costa L, Mohandas N, Sorette M, et al.
Temporal differences in membrane loss lead to
distinct reticulocyte features in hereditary spher-
ocytosis and in immune hemolytic anemia.
Blood 2001;98:2894–2899.
29. Oh KW, Lee K, Ahn B, et al. Design of
pressure-driven microfluidic networks using
electric circuit analogy. Lab Chip 2012;12:515–
545.
30. Mohandas N, Greenquist AC, Shohet SB. Effects
of heat and metabolic depletion on erythrocyte
deformability, spectrin extractability and phos-
American Journal of Hematology, Vol. 90, No. 4, April 2015
phorylation. Prog Clin Biol Res 1978;21:453–
477.
31. Glenister FK. Contribution of parasite proteins
to altered mechanical properties of malaria-
infected red blood cells. Blood 2002;99:1060–
1063.
32. Mills JP, Diez-Silva M, Quinn DJ, et al. Effect of
plasmodial RESA protein on deformability of
human red blood cells harboring plasmodium
falciparum. Proc Natl Acad Sci USA 2007;104:
9213–9217.
33. Da Costa L, Galim J, Fenneteau O, et al. Heredi-
tary spherocytosis, elliptocytosis, and other red
cell membrane disorders. Blood Rev 2013;27:
167–178.
34. Perrotta S, Gallagher PG, Mohandas N. Heredi-
tary spherocytosis. Lancet 2008;372:1411–1426.
35. Groom AC, Schmidt EE, MacDonald IC. Micro-
circulatory pathways and blood flow in spleen:
New insights from washout kinetics, corrosion
casts, and quantitative intravital videomicro-
scopy. Scanning Microsc 1991;5:154–159.
36. Rigat-Brugarolas LG, Elizalde-Torrent A,
Bernabeu M, et al. A functional microengi-
neered model of the human splenon-on-a-chip.
Lab Chip 2014;14:1715–1724.
37. Merkel TJ, Jones SW, Herlihy KP, et al. Using
mechanobiological mimicry of red blood cells to
extend circulation times of hydrogel micropar-
ticles. Proc Natl Acad Sci USA 2011;108:586–
591.
38. Looareesuwan S, Merry AH, Phillips RE, et al.
Reduced erythrocyte survival following clearance
of malarial parasitaemia in Thai patients. Br J
Haematol 1987;67:473–478.
39. Cranston HA, Boylan CW, Carroll GL, et al.
Plasmodium falciparum maturation abolishes
physiologic red cell deformability. Science 1984;
223:400–403.
40. Safeukui I, Buffet PA, Deplaine G, et al. Quanti-
tative assessment of sensing and sequestration of
spherocytic erythrocytes by the human spleen.
Blood 2012;120:424–430.
41. Brousse V, Buffet P, Rees D. The spleen and
sickle cell disease: The sick(led) spleen. Br J
Haematol 2014;166:165–176.
42. Shin JW, Spinler KR, Swift J, et al. Lamins regu-
late cell trafficking and lineage maturation of
adult human hematopoietic cells. Proc Natl
Acad Sci USA 2013;110:18892–18897.
43. Eggenhofer E, Luk F, Dahlke MH, et al. The life
and fate of mesenchymal stem cells. Front
Immunol 2014;5:148.
345