Application of pigment analysis and CHEMTAX to field studies of phytoplankton communities

1
Application of pigment analysis and CHEMTAX to
field studies of phytoplankton communities
Simon Wright Australian Antarctic Division
2
This powerpoint presentation has been cut back
considerably to reduce its size from 29MB to
somewhat closer to the 2MB requested. In doing
so, I have had to exclude all of my antarctic and
shipboard photos (not a great scientific loss),
but also a photo sequence on exactly how we
filter and extract our samples. I am placing
these separately on a Pigment HPLC web site via
the Australian Antarctic Division. I will forward
the address to the PICODIV site. I have also
annotated some of the slides to make them more
stand-alone.
3
The application of pigment analysis to biological
oceanography was largely pioneered by Shirley
Jeffrey. In one of her first post-docs with
George Humphrey, she was given the challenge to

• Find a simple chemical technique for determining
  the abundance of phytoplankton
• This talk will consider how far we have come
  toward that goal.

4
Outline

• Historical perspective development of CHEMTAX
• BROKE 1996 CHEMTAX at work
• Optimising pigment analysis and data
• CHEMTAX problems
• unusual algae
• choice of inputs
• variability of algal pigment content
• Modelling pigments in the underwater light field
• Current directions
• Conclusions

5
How do we measure the abundance of phytoplankton
in the presence of protozoa, bacteria, detritus
and viruses?
6
Many species can be identified by electron
microscopy but cannot be identified by light
microscopy photos omitted Even if they could
be identified by light microscopy, the statistics
of enumeration means that 10000 cells of each
type must be counted to ensure 1 precision.
And. Die numerische Erfassung von
Phytoplankton-Arten gleicht einer Danaiden-Arbeit
die mit einer Zerstoerung von Koerper und Seele
einhergeht Haeckel, 1890 (roughly
). Plankton counting is a task that cannot be
achieved without ruin of body and
soul Chlorophylls and carotenoids are useful
chemical markers that, in the open ocean, are
only found in living phytoplankton. By
chromatographically separating them, we can
determine the composition and abundance of
phytoplankton populations.
7
TLCJeffrey 1974
Chl a
Astaxanthin
Carotenes
Phaeophytin a
Diadinoxanthin Fucoxanthin Neofucoxanthin
Chl b
Peridinin Neoperidinin
Neoxanthin
Chlorophyllide a Chl c
Phaeophorbide a
Origin
8
Jeffrey 1974
In earlier times, we thought in terms of
individual marker pigments indicating particular
algal types or processes.
Pigments Algal types or biological processes
indicated Chl a Chl c Diatoms and / or
chrysomonads Fucoxanthin Diadinoxanthin Chl
b Green algae Neoxanthin Peridinin Dinoflagell
ates Chlorophyllide a Senescent diatoms (due to
chlorophyllase) Phaeophorbide a Faecal pellets
of copepods Phaeophytin a Us. Trace amounts on
all cgrams Astaxanthin Copepods present High
chl ca ratios Senescent phytoplankton or detritus
9
HPLC systems development
Steady improvement in HPLC techniques led to
recognition of many more pigment markers
10
Major marker pigments
Ubiquitous Chl a Unambiguous
Alloxanthin Peridinin Prasinoxanthin
Jeffrey and Vesk (1997)
11
Major marker pigments
Ubiquitous Chl a Unambiguous
Alloxanthin Peridinin Prasinoxanthin Shared
e.g. Fucoxanthin Chl b
Zeaxanthin Violaxanthin
12
Major marker pigments
We can no longer talk in terms of individual
marker pigments. Instead we talk of SUITES of
pigments that may cross conventional taxonomic
boundaries. By the late 80s it became very
apparent that normal interpretation of pigment
data amounted to little more than
guesswork. There was an urgent need for objective
computational methods for determining the
phytoplankton community composition from pigment
data.
13
Computational methods in pigment analysis
14
Computation methods
1. Simple or multiple linear regression e.g.
Gieskes and Kraay 1983 Statistically sound Does
not distinguish algal groups with shared marker
pigments
15
Computation methods
1. Simple or multiple linear regression 2.
Multiple simultaneous equations Everitt et al.
1990 Letelier et al.1993 Peekin 1997 van
Leeuwe et al. 1998
16
Computation methods
1. Simple or multiple linear regression 2.
Multiple simultaneous equations Letelier et al.
1993 ChlaProchl 0.91(Chlb -
2.5Prasino) ChlaCyano 2.1zeax
-0.07(Chlb - 2.5Prasino) ChlaChrys
0.919-but Chrys ChlaPrym 1.319-hex
Prym ChlaBacill 0.8fuco - (0.0219-hex
Prym 0.1419-but Chrys) ChlaDino
1.5perid ChlaPras 2.1prasino
17
Computation methods
1. Simple or multiple linear regression 2.
Multiple simultaneous equations Allowed shared
marker pigments Difficult to set up
18
Computation methods
1. Simple or multiple linear regression 2.
Multiple simultaneous equations 3. Matrix
factorization
19
Computation methods
1. Simple or multiple linear regression 2.
Multiple simultaneous equations 3. Matrix
factorization CHEMTAX (Mackey et al. 1996,
Wright et al. 1996)
20
Matrix factorization

• uses a table of concentration ratios of all
  pigments for each algal group
• Algal Class Pigment
• Chl c3 Peridinin 19-but Fucox 19-hex
  Prasinox
• Diatom - - -
  0.75 - -
• Hapto3 0.045 - -
  - 1.7 -
• Hapto4 0.048 - 0.25 0.58
  0.54 -
• Cryptophyte - - -
  - - -
• Prasinophyte - - -
  - - 0.32
• Chlorophyte - - -
  - - -
• Dinoflagellate - 1.06 -
  - - -
• Cyanobacteria - - -
  - - -
• Each ratio is iteratively modified to minimize
  the difference between observed and calculated
  total pigment concentration

(Half of table only)
21
CHEMTAX software

• Currently based on a MATLAB? platform
• Can distinguish algal groups with qualitatively
  identical pigment compositions using differences
  in pigment ratios (Wright et al, 1996)
• Requires the user to enter the expected mix of
  algal components which the software then
  optimises
• Microscopic examination of the samples is thus
  essential

22
Changes in pigment ratios with depth

• It is essential to split samples into a series
  of depth strata that are computed independently
  (Mackey et al., 1998, Higgins and Mackey, 2000,
  Wright and van den Enden, 2000)

1004 Samples were split into 8 depth layers.
Samples from each layer were computed
independently. Graph at left shows the computed
ratios for type 4 haptophytes (e.g. Phaeocystis
spp.) vs. depth. The smooth change with depth
suggests that CHEMTAX is measuring something real.
23
Phytoplankton community structure and stocks in
the East Antarctic marginal ice zone (BROKE
survey, January - March 1996) determined by
CHEMTAX analysis of HPLC pigment signatures S.
W. Wright and R. L. van den Enden (2000)
Deep-Sea Research II, 47, 2363 - 2400
Does CHEMTAX work?
An example where it worked well to map
phytoplankton communities in the Southern Ocean
24
Study Area
Chlorophyll by satellite
25
Ice shelf
1000m
Ice shelf
26
Antarctic Slope Front
Pycnocline
Tmin
27
ASF
Pycnocline
Tmin
28
ASF
Pycnocline
Tmin
29
ASF
Pycnocline
Tmin
30
ASF
Pycnocline
Tmin
31
ASF
Pycnocline
Tmin
32
ASF
Pycnocline
Tmin
33
ASF
Pycnocline
Tmin
34
ASF
Pycnocline
Tmin
35
ASF
Pycnocline
Tmin
36
BROKE conclusions
1. Effect of Stratification MIXED
STRATIFIED Chl a (µg.L-1)
0.4 2.0 Diatoms
Pycnocline Pycnocline Prasinophytes
Tmin Pycnocline
Hapto4s Tmin Pycnocline 2.
Hole in algal distribution at the ice edge,
except for Cryptophytes 3. Generally uniform to
pycnocline under ice 4. Importance of frontal
features - downwelling tongue from Tmin
layer These observations could not have been
obtained using microscopy or any other method
currently available.
37
Optimizing pigment data
38
Optimising pigment data
Aim Sensitivity maximum peak
height Accuracy Integrity lack of pigment
degradation Reproducibility of retention
times Data reliability These aims require care
at each step of the process
39
Field sampling
It is important to realise that the pigment
composition of the sample starts changing from
the moment it is enclosed in a dark Niskin
bottle. For maximum reproducibility of pigment
ratios, all samples should be subjected to the
same time delay from collection to end of
filtration, and all should be handled in the same
light regime (preferably very dim). For
example, in our cruises, it is normally 40
minutes before we can sample the Niskin bottles
after the physical and chemical oceanographers
have collected their samples. Thus we never see
diatoxanthin. It has all been converted to
diadinoxanthin in the dark.
40
Sample filtration
Dim light, cool lab Fluorescence check (each
sample is measured in a Turner
fluorometer - double checks HPLC
result) Small filter (13mm GF/F, extractable in
1.5 ml solvent) Removal of water
from filter reproducibly Double label cryotubes
(black pen engraver) Freeze in liq. N2
directly into 45 L Dewar
41
Extraction
Sonication in methanol Small volumes
(1.5ml) Internal standard 140 ng
ßapo-8-carotenal (Fluka) Precision Accounts
for volume changes Checks injection
status Straight to refrigerated (-10C)
autoinjector stage
42
More on the internal standard and data
reliability As well as improving analytical
precision, the internal standard provides data
reliability. Thus if you have a sample with no
chlorophyll but a good internal standard peak,
then you know that the injection and the
chromatogram are OK. The filtration may have
been faulty (e.g. holed filter). This is where
you go back to check the fluorescence measurement
you made while filtering. The fluorescence check
has also saved us when fatigued shipboard workers
have labelled two sets of cryotubes with the same
numbers!
43
A photographic sequence of pigment extraction has
been omitted here, to reduce the size of the
file. It will be posted on the Australian
Antarctic Division web site. The address will be
forwarded to the PICODIV site.
44
Our extraction procedure Extraction is performed
in 2.5ml plastic syringes, with a leur lock
tap. Add 1.5 ml cold methanol Add 25 ul internal
standard solution with 140ng apo-8-carotenal
(Fluka) (add these first to avoid delays once
filter is thawed) Remove frozen filter from
cryotube While still frozen, cut 1 x lengthways,
4 x sideways into small pieces, with small
scissors. Pieces fall into the syringe and
thaw. Extract with a probe sonicator (4mm
diameter, 50 W, 60 seconds), moving the syringe
up and down to ensure that no pieces of filter
avoid the sonic beam. The filter is completely
disrupted into a slurry. It gets quite
warm. Immediately put a plunger into the syringe,
attach a 3 mm dia. leur lock filter (0.45 um,
nylon, Advantec MFS Inc) and a needle, and squirt
the extract into an amber autosampler vial. Place
the autosampler vial immediately into a
refrigerated (-10C) autosampler rack. This
process averages 1 min 40 sec from starting
cutting to completion. All syringe parts are
washed with ethanol and dried before reassembly
45
HPLC Analysis
Reproducibility Make up solvents by
weight Column thermostatted in a water bath
(more stable than air
oven) Autoinjection Tubing minimum length
46
RT
BB carot
Chl a
Int Std
diatox
dinox
diad
peridinin
Chl c2
Graph showing reproducibility of retention times
through the day for a series of dinoflagellate
samples
Sample
47
HPLC Analysis
Peak identification Mixed standard every
batch RT table Check column
performance Reference spectral library
48
HPLC Method (1991) Wright et al., 1991

• The SCOR/UNESCO method
• separated 52 pigments in 20 minutes
• C18 monomeric column
• Ternary gradient
• Methanol ammonium acetate
• Acetonitrile
• Ethyl acetate
• Excellent resolution of marker carotenoids
• Inadequate resolution of polar chlorophylls and
  divinyl chlorophylls

49
HPLC Method (2000) Zapata et al. (2000)

• Waters Symmetry C8 (monomeric) column
• Binary gradient
• AMethanolacetonitrilepyridine (pH 5.0)
  (502525 v/v/v)
• B Methanolacetonitrileacetone (206020
  v/v/v)
• 25?C

• Main technique in our lab
• Good resolution of marker carotenoids, esp
  fucoxanthin derivs
• Excellent resolution of polar chlorophylls
• Divinyl chlorophylls resolved from chlorophylls
• Resolution order differs from Wright et al.
  (1991)
• Complementary techniques

50
Peak detection
Diode array detection 436 nm 470 nm 665
nm 400 480 nm (sum) (summing the wavelengths
on the diode array detector provides
about 3 x improvement in signal noise
ratio This greatly improves sensitivity) Fluore
scence detection excitation Broad band blue
(Turner fluorometer filter) emission Long pass
red (gt 630nm)
51
Adjusting HPLC separation

• HPLC separations will not look exactly like
  published methods
• different dead volume between pump and column
• batch variations in columns
• Resolution and retention time may require
  adjustment to suit particular samples
• adjust gradient profile
• adjust solvent composition

52
Data tabulation and checking

• Samples integrated using Waters Millennium
  software
• Peaks matched against library using Millennium
• Data exported to Excel (files concatenated in
  MS-DOS)

53
Data tabulation and checking

• Peaks tabulated and error checked in Excel
• Exported to Prepro files for CHEMTAX
• CHEMTAX output files tabulated using Excel
  macros
• Distributions plotted using Surfer

54
Things may be different in Bermuda (You may
need to modify these methods elsewhere)
55

• Does CHEMTAX have problems?
• Two examples where problems were encountered
  (from Henriksen et al.)
• algae with unusual pigmentation
• inconsistent pigment ratios (?)
• Problems due to variable underwater light field
  follow

56
Exceptional blooms of Chattonella spp.
(Raphidophyceae) and Gymnodinium chlorophorum
(Dinophyceae) in Danish waters

• Peter Henriksen
• National Environmental Research Institute,
  Denmark
• Helene Munk Sørensen
• Århus County, Denmark
• Gert Hansen
• IOC Science and Communication Centre on Harmful
  Algae, Denmark

57
Chattonella spp. in Danish waters 1998
58
Sample dominated by Chattonella spp.(Danish west
coast 12 May 1998)
59
Green dinoflagellate blooming in Danish waters
1999
60
Pigment profiles of strains GH-4 (Århus Bight
1999) and K-539 (Gymnodinium chlorophorum)
chl a
chl b
violax
neox
lutein
61
(No Transcript)
62

• The two dominant organisms were
• A raphidophyte, Chattonella sp., with pigments
  characteristic of pelagophytes and haptophytes
• A dinoflagellate with pigments characteristic of
  chlorophytes
• Message Must use a microscope

63
Species included in2 different pigment
ratio-files
In another expt they compared microscopically
estimated C biomass (itself selective) with
CHEMTAX attribute of chl a, starting from 2
different ratio files (above)
64
Århus Bight 1997-99Dinoflagellates
Gymnodinium chlorophorum
65
Århus Bight 1997-99Diatoms
66
Århus Bight 1997-99Haptophytes
67
I believe these correlations were so bad because
a whole years data (summer and winter) was
included in the analysis. The pigment ratios
would have changed between seasons, contravening
CHEMTAXs assumption that pigment ratios are
constant through the data set.
68
Taxon specific subsurface chlorophyll maxima in
the Southern Ocean south of Tasmania March 1998
S.W.Wright, R. L. van den Enden, F. B.
Griffiths, A. C. Crossley (in prep)
69
Subantarctic zone (SAZ)
The chlorophyll in the SAZ is relatively
uniformly distributed vertically, but CHEMTAX
found the populations to be stratified as follows
Subtropical front
Subantarctic front
Polar front
70
(No Transcript)
71
(No Transcript)
72
(No Transcript)
73
(No Transcript)
74
(No Transcript)
75
(No Transcript)
76
(No Transcript)
77
(No Transcript)
78
Was the observed stratification a result of
splitting the data into depth strata and
computing them independently?
Remember this?
Only the Hapto3 data suggested not - the
subsurface chl maximum deepened between layers
toward the south of the transect, as follows.
79
(No Transcript)
80
The microscopic data had neither the sampling
density nor the statistical precision to
determine whether these patterns were real (hence
the need for CHEMTAX in the first place). The
only comparable data were flow cytometric counts
of cyanobacteria for four stations in the
SAZ. Comparing cyanobacterial counts with
CHEMTAX estimates of cyanobacterial chlorophyll
showed that CHEMTAX consistently underestimated
cyanobacterial abundance near the surface and
overestimated at depth (data for 4 stations
follows)
81
(No Transcript)
82
(No Transcript)
83
(No Transcript)
84
(No Transcript)
85
If you calculate the amount of cyanobacterial
chlorophyll per cell vs depth (CHEMTAX cyano chl
/ flow cytometer counts), you find that it is
relatively constant near the surface, then
increases dramatically at depth. Four stations
follow with cyano chl per cell and cell counts
vs depth. NB. Ignore noisy data where cell
counts approach zero at depth.
86
(No Transcript)
87
(No Transcript)
88
(No Transcript)
89
(No Transcript)
90
Summary
91
Is this real? We compared calculated pigment per
cell with data obtained in our lab over the last
2 years measuring
How does cellular pigment content respond to
irradiance?
92
Variation of pigment content in response to
irradiance

• Several species cultured under a range of
  irradiances
• 10 888 uE m-2 s-1
• marine blue filtered light
• log phase cultures used
• Pigments analysed using Zapata et al. (2000)
• Cells counted by flow cytometry

93

• Species employed
• Phaeodactylum tricornutum - Diatom
• Pavlova gyrans - Haptophyte
• Emiliania huxleyi - Haptophyte
• Dunaliella tertiolecta - Chlorophyte
• Pelagococcus subviridis - Pelagophyte
• Synechococcus sp. - Cyanobacterium
• Amphidinium carterae - Dinoflagellate
• Phaeocystis antarctica x 2 - Haptophyte
• Polarella glacialis - Dinoflagellate
• Homo sapiens
• Lana Pirrone - University of Tasmania
• Suzanne Roy - Université de Québec, Canada
• Peter Henriksen - Danish Environmental
  Research Inst.

94
Light gradient apparatus
95
(No Transcript)
96
(No Transcript)
97
(No Transcript)
98
(No Transcript)
99
(No Transcript)
100
(No Transcript)
101
(No Transcript)
102
We obtained equations for these lines so that we
could model the pigment /cell vs irradiance, then
put those equations into an underwater light
field to model pigment /cell vs depth.
103
(No Transcript)
104
(No Transcript)
105
(No Transcript)
106
Vertical Atten. Coeff. 0.046
107
Combining these models showed that pigment / cell
varied with depth and produced a subsurface chl
maximum in all species.
108
(No Transcript)
109
(No Transcript)
110
(No Transcript)
111
Compare the last graph with the only real data we
have
112
(No Transcript)
113
Changes in pigment ratios with depth

• It seems that CHEMTAX accurately calculated the
  pigment response of cyanobacteria and hence
  presumably the other categories.

114
Message
More data required on pigment content vs light
fields Need to model algal responses to
depth Methods to convert chlorophyll estimates
to cells / L or total carbon? What about
variable light climates?
115
How do we model pigment content?

• Lack of data on pigment composition of
  phytoplankton
• Few species done
• Ideally should know characteristics of major
  species
• But usually dont know what proportion the
  species are in, so cant calculate average
• Refine the limits range
• Normally have to let CHEMTAX calculate average
  ratio
• Nutrient effects?

116
Thought Maybe we can use cells such as
cyanobacteria and cryptophytes (each readily
distinguishable in both flow cytometry and
CHEMTAX) to determine their pigment / cell and
use them as proxies for the underwater light
field.
117
Current directions
New pigments Chl c derivs Non polar chl
cs Gyroxanthin diester 4 keto-acyl
fucoxanthins Un421 Tracking particular species
e.g toxic dinos
118
Use of size fractionation

• Differentiate between nanoplankton and
  microplankton
• ecologically meaningful
• remove contribution of large diatoms from pools
  of fucoxanthin etc.
• Simplifies CHEMTAX interpretation

119
Software improvements
Current workup software (Excel, Surfer
macros) Translate CHEMTAX to Excel? Software
should incorporate changing pigment ratios with
depth or (preferably) light field. Software
should identify changes of oceanic region and/or
gross species composition within a sample
set. (following)
120
After CHEMTAX has finished optimising ratios, it
should look at how each sample responds to a
change in ratio - e.g. increasing diatom
fuco/chl_a ratio may increase total chl_a in some
samples (pink) and decrease others (yellow). By
looking at how individual samples respond to
ratio changes, CHEMTAX may decide the data set
can be spilt and optimized separately.
121
What are the overall conclusions?
122
Power of pigment analysis

• first step in discriminating algal types

123
Power of pigment analysis

• first step in discriminating algal types
• allows hundreds of samples to be analysed

124
Power of pigment analysis

• first step in discriminating algal types
• allows hundreds of samples to be analysed BUT

125
Power of pigment analysis

• first step in discriminating algal types
• allows hundreds of samples to be analysed BUT
• lack of pigment ratios

126
Power of pigment analysis

• first step in discriminating algal types
• allows hundreds of samples to be analysed BUT
• lack of pigment ratios
• more algal cultures must be analysed by the best
  methods

127
Power of pigment analysis

• first step in discriminating algal types
• allows hundreds of samples to be analysed BUT
• lack of pigment ratios
• more algal cultures must be analysed by the best
  methods
• pigmentchl a ratios needed to validate CHEMTAX
  in response to

128
Power of pigment analysis

• first step in discriminating algal types
• allows hundreds of samples to be analysed BUT
• lack of pigment ratios
• more algal cultures must be analysed by the best
  methods
• pigmentchl a ratios needed to validate CHEMTAX
  in response to
• light

129
Power of pigment analysis

• first step in discriminating algal types
• allows hundreds of samples to be analysed BUT
• lack of pigment ratios
• more algal cultures must be analysed by the best
  methods
• pigmentchl a ratios needed to validate CHEMTAX
  in response to
• light
• nutrient regimes

130
Power of pigment analysis

• first step in discriminating algal types
• allows hundreds of samples to be analysed BUT
• lack of pigment ratios
• more algal cultures must be analysed by the best
  methods
• pigmentchl a ratios needed to validate CHEMTAX
  in response to
• light
• nutrient regimes
• need pigment / cell w.r.t. to environment

131
Power of pigment analysis

• first step in discriminating algal types
• allows hundreds of samples to be analysed BUT
• lack of pigment ratios
• more algal cultures must be analysed by the best
  methods
• pigmentchl a ratios needed to validate CHEMTAX
  in response to
• light
• nutrient regimes
• need pigment / cell w.r.t. to environment
• improved computational methods needed