Environmental DNA metabarcoding –

the future of biomonitoring

Jan Pawlowski
Department of Genetic and Evolution, University of Geneva, Switzerland
Institute of Oceanology, PAS, Sopot, Poland
ID-Gene Ecodiagnostics Ltd, Switzerland
DNAqua-net Cost Action: Developing new genetic tools for
bioassessment of aquatic ecosystems in Europe

http://dnaqua.net
WG1 – DNA Barcode
References
WG2 – Biotic Indices &
Metrics
WG3 – Field & Lab
Protocols
WG4 – Data Analysis &
Storage
WG5 – Implementation
Strategy & Legal Issues

Leese et al. 2018

Challenges of
implementing
the DNA-
based
methods in
biomonitoring

Leese et al.
2018
 Hg Cl- Cu

Environmental
Environment COD Physics
NH4

Environmental DNA
Cd
DNA T° Chemistry
PCB pH
NO3

(eDNA) as a new,
complementary, time
and cost-effective tool
for biomonitoring Communities
composition
Evenness Richness
Bioindication
Biotic Index

Biomass Specific tolerance

Status assessment
Environmental DNA:
What’s behind the
term?
eDNA applications
• Species detection
(qPCR, dPCR)
• Biodiversity survey
(metabarcoding)
• Bioassessment
(biotic indices)
eDNA metabarcoding workflow

Millions of DNA
metabarcodes
generated by
high-throughput
sequencing (HTS)
platforms
Benthic monitoring
Assessing ecological quality
status of marine environment
using benthic macro-
invertebrates (annelids, molluscs,
echinoderms, etc)

picture A.Norro RBINS

Traditional benthic monitoring
Sorting macro-invertebrates from sediment
samples and identify them
morphologically.
Next-generation benthic monitoring
Analysing benthic biodiversity using high-
throughput environmental DNA
metabarcoding of sediment samples.
Benthic monitoring applied to marine industries

Assessment of organic enrichment Assessment of chemical pollution

associated with salmon farming associated with oil & gas industry
Experimental approach

Bulk
Morphotaxonomy Morphology-based
sample
Biological Quality
Index

Sediment
Metabarcoding
sample DNA-based Biological
Quality Index
Sampling

• Sediments samples collected following one or

several axes with increasing distance from the
source of potential polution (fish cage / platform)

• Different genetic markers:

• COI – standard macrofauna DNA barcode
• 18S V1V2 – universal eukaryotic marker,
• 18S V9 – universal eukaryotic marker
• 18S 37F – foraminiferal-specific markers

• Comparison of metabarcoding data with

morphology identification.
 eDNA metabarcoding provides an
Results accurate assessment of impacts
associated with marine aquaculture

0m 11m 26m 40m 50m 60m 270m 340m 400m 76m

Taxonomic composition of zoobenthos community identified morphologically (middle

panel), and genetically by RNA (upper panel) and DNA (lower panel)
Lejzerowicz et al. 2015
Results
Results

Different markers amplify

preferentially different
groups of eukaryotes
Results

NMDS illustrating the

level of similarity of
eukaryotes
communities classified
according to the
platform or station
 Results

Agostino-B

Garibaldi-A

6
NMDS plots showing the patterns of divergence between Armida
samples in relation to the distance for all eukaryotes (18S 0

V1V2) in sediment DNA samples N

Challenges
How to transform high-
Scientific & technical throughput sequencing
eDNA data into ecological
status assessment?
 • Benthic macrofauna is under-
Challenges represented in sediment DNA
samples and their sequence
Limitations of sediment DNA abundance data are unreliable.
metabarcoding • Meiofauna and microorganisms
are well represented in sediment
Represented Ecological value
DNA samples but lack taxonomic
Macrofauna
identification and no ecological
values are assigned to them
Meiofauna

Micro-
eukaryotes
Solution 1

Develop de novo
molecular indices
targeting microbial or
meiofauna groups
(e.g. nematodes)

The alpha-diversity indices calculated for nematodes (18S

V1V2) increase with distance
Solution 2

Use machine learning to

predict biotic indices based on
microbial/meiofaunal eDNA
metabarcoding data
 Making
Training a model predictions
Machine learning Macro-invertebrates
Biotic index
New eDNA samples

algorithms are Sieving

trained on
macrofauna-based
ecological status to
eDNA
predict the
ecological status of
eDNA samples eDNA
extraction

Data collection and model training Making predictions

 NSI prediction / Inference assigned metazoan - V1V2

Machine learning significantly improves Aukrasanden

30

100
Beitveitnes

the accuracy of benthic indices

Bjørnsvik

80
Nedre Kvarv

Percentage

60
100 n=23

25
Storvika

1
2

40
R
100 100

n=9

1
Macrofauna

20
2 n=4
R n=2 n=3

0.8
100
80

20 n=1
Calculation

1
NSI prediction / RF2 − V1V2 eukaryotes Metazoa
% of sequences used

0
R Macrofauna

0.8
Molecular
used 80

1
2 Calculation
Macrofauna -2 -1 0 1 2 3
used used

R Aukrasanden

0.8
0.6
80
60

30

15 80 100
Calculation
15

Macrofauna
Beitveitnes Status mismatch
Macro + Meiofauna

0.8
80

2
0.6
Bjørnsvik
60
% of sequences

Calculation

R
SML
Nedre Kvarv n=27

0.6
60

0.2 R 20.4
of sequences
40

Macro + Meiofauna

Percentage

60
2
25
10

Storvika

R
Macro + Meiofauna 11.99
SML

0.6

40
60

0.4

10
40
% of sequences

SML

0.4
n=9
40

520
Macro + Meiofauna
20

n=5

2
NSI prediction / RF − V1V2 eukaryotes All
20

R
4.05
n=1
Eukaryotes
5

SML

0 0
1.71

0.4
40

0.2
%20

SML -0.08
Molecular

0.2
R²=0.474***
20

-1.52 −1
Eukaryotes
Aukrasanden
Eukaryotes
−2 0 1
0 30
0

100
Kappa=0.601***
SML
Beitveitnes

80 -5
15

-5.67
SML

0.2
20

Bjørnsvik Status mismatch

0

0
n=29
Macrofauna 5Macro10
+ Meiofauna Eukaryotes
0

0
0 15 20 25 30 Eukaryotes
Nedre Kvarv

Percentage

4 660
6.35
1025

Calculation SML SML Storvika

SML
Macrofauna
Macrofauna Macro
Macro++Meiofauna
Meiofauna Eukaryotes
MorphologyEukaryotes

−20 0 202 40
0

0
Calculation
Calculation SML
SML SML
SML 2.05
n=7 n=6
0.45 0.41
520

Macrofauna Macro + Meiofauna Eukaryotes −1.2

Calculation SML SML
lecular

R²=0.841*** −1 0 1
−4.91

−6
15

Kappa=0.859***
Machine learning
performance depends
on selected marker
foraminifera 37F bacteria V3V4

30
30
• 18S V1/V2 – Zoobenthos

25
25

20
20
• 18S V9 – Eukaryotes

15
15

10
10
• 18S 37F - Foraminifera

5
5
R =0.82*** R =0.89***
Kappa=0.832*** Kappa=0.9***

Molecular
• 16S V3 - Bacteria

0
0
0 5 10 15 20 25 30 0 5 10 15 20 25 30

eukaryotes V1V2 eukaryotes V9

30
30

25
25
AMBI ISI NSI NQI1 Average

20
20
Forams 0,834 0,651 0,82 0,822 0,78

15
15
Bacteria 0,863 0,785 0,89 0,908 0,86

10
10
Euk V1V2 0,886 0,764 0,914 0,902 0,87

5
5
R =0.914*** R =0.953***

Euk V9 0,906 0,79 0,916 0,938 0,89 Kappa=0.893*** Kappa=0.919***

0
0
0 5 10 15 20 25 30 0 5 10 15 20 25 30

Morphology
Comparison of indices inferred from conventional
macrofauna survey and predicted using machine learning
from microbial and meiofaunal metabarcoding data

Benthic monitoring of salmon farms in Norway (18S V1V2)


Challenges
What are the socio-economic
obstacles in implementation of
ecogenomics in routine
biomonitoring?
• Novelty of environmental
genomics approach
• Expectations that ecogenomics
will provide exactly the same
results as classical approach
• Limited number of ecogenomic
studies and reference datasets
concerning industrial impacts
• Insufficient communication about
the advantages of DNA-based
biomonitoring
 • Develop de novo taxonomy-free
Roadmap for the future pressure indices based on
metabarcoding data
• Develop global metabarcoding
What can be done to help reference database for
implementing the ecogenomic environmental impact
tools in regulatory guidelines? assessments
• Identify new potential bioindicator
metabarcodes
• Establish standardized protocols
for ecogenomic surveys
Take-home message

Why eDNA metabarcoding is the

future of biomonitoring?
• It is non-invasive
• It reveals hidden diversity
• It persists over time
• It allows global monitoring of
environmental impacts across all
domains of life
Thank you for
your attention!