Freshwater Biology (2008) doi:10.1111/j.1365-2427.2008.01992.

Evidence that viral abundance across oceans and lakes

is driven by different biological factors
J E S S I C A L . C L A S E N * , S E A N M . B R I G D E N †, J É R Ô M E P . P A Y E T * A N D C U R T I S A . S U T T L E * , †, ‡
*Department of Earth and Ocean Sciences, University of British Columbia, Vancouver, BC, Canada
†
Department of Botany, University of British Columbia, Vancouver, BC, Canada
‡
Department of Microbiology and Immunology, University of British Columbia, Vancouver, BC, Canada

SUMMARY
1. Samples from 16 lakes in central (n = 145) and western (n = 12) North America, the
coastal northeast Pacific (n = 302) and the western Canadian Arctic Oceans (n = 142) were
collected and analysed for viral, bacterial and cyanobacterial abundances and chlorophyll-a
concentration.
2. Viral abundance was significantly different among the environments. It was highest in
the coastal Pacific Ocean and lowest in the coastal Arctic Ocean. The abundances of
bacteria and cyanobacteria as well as chlorophyll-a concentrations also differed signifi-
cantly among the environments, with both bacterial abundance and chlorophyll-a
concentration highest in lakes. As a consequence, the association of these variables with
viral abundance varied among the environments.
3. Discriminant analyses with the abundance data indicated that the marine and
freshwater environments were predictably different from each other. Multiple-regression
analysis included bacterial and cyanobacterial abundances, and chlorophyll-a concentra-
tion as significant variables in explaining viral abundance in lakes. In regression models
for the coastal Pacific Ocean, bacterial and cyanobacterial abundances were significant
variables, and for the coastal Arctic Ocean viral abundance was predicted by bacterial
abundance and chlorophyll-a concentration.
4. The relationship of viral and bacterial abundance differed between the investigated
freshwater and marine environments, probably because of differences in viral production
and loss rates. However, freshwaters had fewer viruses compared to bacteria, despite
previously documented higher burst sizes and frequencies of infected cells, suggesting that
loss rates may be more important in lakes.
5. Together, these findings suggest that there are different drivers of viral abundance in
different aquatic environments, including lakes and oceans.

Keywords: bacteria, chlorophyll-a, freshwater, marine, viruses

tal gene transfer (see Wommack & Colwell, 2000;

Introduction
Viruses are dynamic and ecologically important
members of aquatic ecosystems that can influence
nutrient cycles, community composition and horizon-
Correspondence: Curtis A. Suttle, Rm 1461 – 6270 University
Blvd, University of British Columbia, Vancouver, BC V6T 1Z4,
Canada. E-mail: csuttle@eos.ubc.ca
 2008 The Authors, Journal compilation  2008 Blackwell Publishing Ltd
Weinbauer, 2004; Suttle, 2005, 2007). The factors
influencing viral abundance and dynamics in aquatic
environments are complex. However, correlation
analysis can be used as a first step to identify
physical, chemical and biological variables that are
associated with changes in viral abundance. Such
analyses have already provided insight into the
factors regulating viral abundance in aquatic
1
2 J. L. Clasen et al.
environments (reviewed in Suttle, 2000; Wommack &
Colwell, 2000). For example, correlation analysis
indicated that high viral abundance is typically
associated with high bacterial abundance and ⁄ or
chlorophyll-a concentration (Cochlan et al., 1993; Paul
et al., 1993; Jiang & Paul, 1994; Steward, Smith &
Azam, 1996; Weinbauer & Suttle, 1997; Kepner,
Wharton & Suttle, 1998). The majority of these studies
focused on marine rather than freshwater environ-
ments, but Maranger & Bird (1995) compared viral
abundances from lakes and marine systems with
bacterial abundance and production, as well as with
concentrations of chlorophyll-a and total dissolved
phosphorus. An important result of this comparison
was the strong correlation found between viral
abundance and chlorophyll-a concentration, but not
with bacterial abundance in lakes, whereas in marine
systems viral abundance was correlated strongly with
bacterial abundance. Based upon these results, the
authors suggested that the factors controlling the
abundance of viruses differ between lakes and oceans.
However, as viral abundance has frequently been
underestimated because of a number of methodolog-
ical problems, including inherent errors in transmis-
sion electron microscopy (TEM) and fixation problems
in many studies that used epifluorescence microscopy
(EFM) (see Hennes & Suttle, 1995; Weinbauer &
Suttle, 1997; Bettarel et al., 2000; Wen, Ortmann &
Suttle, 2004; Suttle, 2005), we re-visited the relation-
ships between the abundances of viruses and poten-
tial hosts as inferred from cell counts of heterotrophic
bacteria and unicellular cyanobacteria, as well as of
chlorophyll-a concentrations (i.e. biological variables).
The purpose was to gain insight into the influence of
biological variables on viral abundances in both fresh
water and marine systems, using several relatively
large data sets from our own laboratory in which we
had confidence in the accuracy, or at least comparabil-
ity, of abundance estimates. We compared viral abun-
dance with several biological variables in fresh water
and marine environments to determine which biolog-
ical factors may drive changes in viral abundance.
Methods
Sampling
Lakes. Samples were collected from lakes in Wiscon-
sin (WIS 1999), British Columbia (BCC 2002), Lake of
 2008 The Authors, Journal compilation  2008 Blackwell Publishing Ltd, Freshwater Biology, doi:10.1111/j.1365-2427.2008.01992.x
Arctic
BFS
(n = 142)
BCC LOW
Pacific ELA
SOG
(n = 302)
Lake WIS
(n = 157)
Fig. 1 Map of the sampling sites used in the present study. Lake
samples were collected from Wisconsin (WIS), British Columbia
(BCC) and Ontario (Lake of the Woods, LOW and the Experi-
mental Lakes Area, ELA). All the coastal Pacific Ocean samples
were collected in or near the Straight of Georgia, British
Columbia (SOG), while coastal Arctic Ocean samples were
collected in the southwestern part of the Beaufort Sea (BFS).
the Woods, Ontario (LOW 2003) and the Experimental
Lakes Areas, Ontario (ELA 2003 and 2004) during the
ice-free season (June–October) (Fig. 1). The 157 fresh-
water samples came from lakes that varied in size,
depth and trophic status (Juday & Birge, 1941; Cleugh
& Hauser, 1971). Samples were collected by either a
Van Dorn bottle, a pump or by carefully submerging
and filling a carboy with subsurface water. Samples
were taken from depths <2 m, except for samples
from Wisconsin (WIS 1999), which were collected at
depths between 0.5 and 32 m. Water samples were
processed immediately for chlorophyll-a concentra-
tions and for viral, bacterial and cyanobacterial
abundances, as described below.
Coastal Pacific Ocean. Coastal Pacific Ocean samples
were collected during four research cruises in and
near the Strait of Georgia, British Columbia, Canada
(Fig. 1). These cruises occurred during August 1996,
June 1997, July 2000 and July 2002 (hereafter, SOG
1996, 1997, 2000 and 2002 respectively). The Strait of
Georgia lies between continental British Columbia
and Vancouver Island. It is c. 28 km wide and 240 km
long and is characterized by areas of high primary

productivity and flushing rates (Harrison et al., 1983).
Water samples were collected from 0.5 to 100 m using
Go-Flo bottles mounted on a rosette equipped with a
CTD and fluorometer. Upon reaching deck, samples
for viruses, bacteria, cyanobacteria and chlorophyll-a
were collected and processed immediately.
Coastal Arctic Ocean. In the Canadian Arctic Ocean,
samples were collected in the southeastern Beaufort
Sea in October 2003 and from May to August 2004
(BFS 2003 & 2004) during the Canadian Arctic Shelf
Exchange Study (CASES) (Fig. 1). Water was collected
at several stations, ranging from stations heavily
influenced by the Mackenzie River to more oceanic
stations in the Amundsen Gulf (Garneau et al., 2006,
2008). All stations were located on the Arctic shelf,
which is an important area of primary production in
the Arctic Ocean (Carmack, MacDonald & Jasper,
2004). Water from depths ranging from 1.0 to 530 m
was obtained with Niskin or Go-Flo bottles mounted
on a rosette. Samples were immediately processed for
viruses, bacteria and chlorophyll-a once onboard.
Enumeration
Viruses. Total viral abundance was determined by
EFM of samples stained with either YO-PRO-1TM
(4-[3-methy-2,3-dihydro-(benzo-1,3-oxazole)-2-methyl
methyledene]-1-(3’-trimethyl-amonium-propyl)-quino-
linium diiodide; Invitrogen, Burlington, ON, Canada;
WIS 1999, SOG 1996 and BFS 2003 & 2004) or SYBR
Green ITM (Invitrogen) (BCC 2002, LOW 2003, ELA
2003 & 2004 and SOG 1997 & 2000).
For samples stained with YO-PRO-1TM, 0.8–1.6 mL
of the sample was gently filtered onto a 0.02 lm pore-
sized Anodisc filter (Whatman, Florham Park, NJ,
U.S.A.) and placed sample side up on an 80-lL drop
of YO-PRO-1TM (50 lM ) (Hennes & Suttle, 1995). The
slides were incubated in the dark for 48 h, rinsed
with filtered-deionized water and then mounted on a
slide with 100% glycerol. Slides were frozen ()20 C)
until enumerated on an epifluorescence microscope
with a wide-blue filter set (450–480 nm). Total viral
abundance (viruses mL)1) was determined by count-
ing a minimum of 300 viruses over 20 random fields
on each slide. The incubation period in the samples
from the Wisconsin lakes (WIS 1999) was reduced to
4 min by microwave radiation (Xenopoulos & Bird,
1997).
 2008 The Authors, Journal compilation  2008 Blackwell Publishing Ltd, Freshwater Biology, doi:10.1111/j.1365-2427.2008.01992.x
Viral abundance in oceans versus lakes 3
When SYBR Green ITM was used (Noble &
Fuhrman, 1998), the sample was fixed with glutaral-
dehyde to a final concentration of 2% and processed
within 1 h of fixation (Wen et al., 2004). A volume of
the fixed sample (0.8–1.6 mL) was filtered onto a
0.02-lm pore-sized Anodisc filters (Whatman) and
placed sample side up on an 80-lL drop of SYBR
Green ITM (diluted 400·) and incubated in the dark for
about 15 min. After incubation, the filter was carefully
blotted dry with a KimwipeTM (Kimberly-Clark,
Irving, TX, U.S.A.) and mounted on a slide using a
mixture of 50% glycerol, 50% phosphate buffered
saline and 0.1% p-phenylenediamine. Slides were
frozen ()20 C) until analysed as described above.
Bacteria. Bacterial abundance was determined by
EFM of cells stained with YO-PRO-1TM (WIS 1999, BFS
2003 & 2004), SYBR Green ITM (SOG 2000), acridine
orange (AO; SOG 1996 & 1997) or 4’6-diamidino-
2-phenylindole (DAPI; BCC 2002, LOW 2003, ELA
2003 & 2004). Bacterial abundance (bacteria mL)1)
was estimated by counting more than 200 bacteria
over 20 random fields on each slide.
Bacteria enumerated with YO-PRO-1TM and SYBR
Green ITM were processed as described for viruses.
When samples were stained with AO (Hobbie, Daley &
Jasper, 1977) or DAPI (Porter & Feig, 1980), 0.8–5 mL of
the sample was incubated for 3 min with AO (2 mL
of 0.1 mg mL)1) or 10 min with DAPI (250 lL of
0.1 lg mL)1) and then filtered onto a black 0.2 lm
pore-sized polycarbonate filter. The filter was mounted
on a slide with low fluorescence immersion oil and
frozen ()20 C) until cells were enumerated by EFM
using either wide blue (AO) or UV (DAPI) filter sets.
Cyanobacteria. For SOG 1996, cyanobacteria abun-
dances were determined from YO-PRO-1TM virus
slides (see above) by counting autofluorescing cyano-
bacteria. For the other samples (LOW 2003, ELA 2003
& 2004 and SOG 1997 & 2000) cyanobacterial abun-
dance was determined by filtering 5–15 mL of water
onto a black 0.2-lm pore-sized polycarbonate filter.
The filter was filtered to dryness, mounted onto a
glass slide with 100% glycerol and frozen ()20 C)
until cells were enumerated. For all the sampling
stations (including SOG 1996), abundance (cyanobac-
terial cells mL)1) was determined by counting >200
autofluorescing cyanobacteria under wide-green exci-
tation (510–550 nm).
4 J. L. Clasen et al.
Chlorophyll-a
The concentration of chlorophyll-a (lg L)1) was deter-
mined by gently filtering 15–100 mL of freshly col-
lected water through a GF ⁄ F filter (Whatman) under
low light (WIS 1999, BCC 2002, LOW 2003, ELA 2003
& 2004, SOG 1997 and BFS 2003 & 2004). The filters
were frozen ()20 C) until extracted in 90% acetone
and the concentration of chlorophyll-a was deter-
mined by fluorometry (Wetzel & Likens, 1991).
Data analysis
ANOVA. Log-transformed biological variables, inclu-
ding viral abundance (viruses mL)1), bacterial abun-
dance (cells mL)1), the virus-to-bacteria ratio,
cyanobacterial abundance (cells mL)1), and chloro-
phyll-a concentrations (lg L)1), were compared
among environments by A N O V A followed by Tukey
post hoc tests (S Y S T A T version 11; Systat Software Inc.,
San Jose, CA, U.S.A.). Viral abundance was also
regressed against the other biological variables.
Discriminant analysis. Discriminant analyses were
used to statistically describe differences among envi-
ronments (S Y S T A T version 11). Viral, bacterial and
cyanobacterial abundances, and chlorophyll-a concen-
trations were used as predictors in these analyses. A
discriminant analysis predicts the likelihood that an
‘unknown’ sample belongs to a particular group
(Manly, 2005). All data were log-transformed to
satisfy the assumptions of this analysis, which include
normal distribution of data and homogeneity of
variance. The assumptions were evaluated, both
initially and again after transformation, by visually
inspecting probability and scatter plots.
Multiple regression analysis. To determine the variables
that statistically explained the most variation in viral
abundance within each environment, the biological
variables were regressed against viral abundance in a
backwards step-wise multiple regression analysis
using S Y S T A T version 11. This analysis systematically
removes independent variables from the model that
contribute little to explaining the dependent variable.
At each removal step, the data are scanned for variables
that should be re-entered into the model because their
contribution may have increased because of inter-
actions with other variables (Sokal & Rohlf, 1995).
Independent variables in the regression models in-
cluded bacterial abundance (cell mL)1), cyanobacterial
abundance (cyanobacteria mL)1) and chlorophyll-a
concentration (lg L)1); however, cyanobacteria data
were not included in the Arctic Ocean regression
model because cyanobacteria were not counted in these
samples. All data were log-transformed prior to anal-
ysis to satisfy the assumptions of regression analyses.
Results
The abundance of viruses was determined in 601
samples, representing three distinct environments
and ranged from 2.2 · 105–3.9 · 108 viruses mL)1
(Table 1). There were significant differences in viral
abundances (Fig. 2a; A N O V A , n = 601, F2,598 = 432.6,
P < 0.0001), and bacterial abundances (Fig. 2a; A N O V A ,
n = 601, F2,598 = 189.3, P £ 0.0001) among environ-
ments. Lakes had significantly higher concentrations
of chlorophyll-a than either the coastal Pacific or Arctic
Oceans (Fig. 2b; A N O V A , n = 397, F2,394 = 322.8,
P < 0.0001), while the coastal Pacific Ocean had slighly
higher abundances of unicellular cyanobacteria than
lakes (Fig. 2b; A N O V A , n = 201, F1,289 = 45.0,
P < 0.0001). Viruses to bacteria ratios were also signif-
icantly different among environments (Fig. 2c; A N O V A ,
n = 601, F2,598 = 435.5, P < 0.0001).
Overall, viral abundance was only weakly related to
bacterial abundance, cyanobacterial abundance and

Table 1 Average values and ranges (in parentheses) of the biological variables determined in each of the three sampled environments

Chlorophyll-a Cyanobacterial
Viral abundance Bacterial abundance concentration abundance Ratio of viruses
Environment (no. mL)1) (cells mL)1) (lg L)1) (cells mL)1) to bacteria

Lakes 9.5 · 106 (1–60 · 106) 4.0 · 106 (0.2–15 · 106) 12.7 (0.3–432) 9.1 · 104 (0.9–29 · 104) 2.9 (0.4–32)
Coastal Pacific Ocean 6.7 · 107 (0.4–39 · 107) 3.0 · 106 (0.01–17 · 106) 0.94 (0–5.4) 1.0 · 105 (3–19 · 105) 40.0 (3.9–2150)
Coastal Arctic Ocean 3.9 · 106 (0.2–10.9 · 106) 5.6 · 105 (0.9–33.7 · 105) 0.26 (0–3.4) nd 11.1 (0.8–50)

nd, not determined.

 2008 The Authors, Journal compilation  2008 Blackwell Publishing Ltd, Freshwater Biology, doi:10.1111/j.1365-2427.2008.01992.x
 Viral abundance in oceans versus lakes 5
(a) 108 chlorophyll-a concentration (bacteria: r2 = 0.22,
b Viruses P < 0.0001, n = 565; chlorophyll-a: r2 = 0.022, P =
Viral or bacterial abundance
(viruses mL–1 or cells mL–1)
Bacteria
0.005, n = 361; cyanobacteria: r2 = 0.02, P = 0.005, n =
a 391); however, when the data were analysed sepa-
107 rately by environment, striking patterns emerged
a* b* c
(Fig. 3). There was little overlap among the clusters
of viral and bacterial abundances generated from each
106 c* environment (Fig. 3a). Similarly, clusters of viral
abundance and chlorophyll-a concentration or cyano-
bacterial abundance were distinctly different between
the lakes and the marine regions, including both the
105
Lakes Pacific Arctic Pacific and Arctic coastal environments (Fig. 3b,c
Ocean Ocean respectively).
(b) To examine the extent to which the data could be
Cyanobacterial abundance (cells mL–1)
Chlorophyll-a concentration (µg L–1) or

separated by environment, a series of discriminant

10 6
b* function analyses were performed. With just viral
a* Chlorophyll a
Cyanobacteria abundance in the model, the data was placed into
10 4 the correct environment about 67% of the time
(Table 2). The accuracy of the model increased when
bacterial abundance was added. Data from the lakes,
10 2
a coastal Pacific and Arctic oceans were correctly
identified 90%, 92% and 88% of the time res-
b pectively (overall average = 90%). The addition of
10 0
chlorophyll-a concentration and cyanobacterial abun-
dance increased the overall accuracy to 94% and
10–2 99% (Table 2).
Lakes Pacific Arctic Backward step-wise multiple regression analyses
Ocean Ocean
(c) were used to determine which variables explained the
60 most variation in viral abundance within each envi-
ronment. The models for lakes and the coastal Pacific
b
Virus-to-bacteria ratio

Ocean regressed viral abundance (viruses mL)1)

against bacterial abundance (cells mL)1), cyanobacte-
40
rial abundance (cells mL)1) and chlorophyll-a concen-
tration (lg L)1), while the Arctic Ocean model
regressed viral abundance (viruses mL)1) against
20 bacterial abundance (cells mL)1) and chlorophyll-a
c concentration (lg L)1). The final models explained
a 39%, 52% and 26% of the variation in lakes, and the
0 coastal Pacific and Arctic Oceans respectively
Lakes Pacific Arctic (Table 3). Bacterial abundance remained in the final
Ocean Ocean model for all three environments, thus always explain-
ing some of the variation. However, the models
Fig. 2 A comparison of the biological variables determined
within each of the three aquatic environments sampled. (a) Viral
differed in their remaining variables. Part of the
abundance and bacterial abundance, (b) chlorophyll-a concen- variation in viral abundance in the coastal Pacific
tration and cyanobacterial abundance and (c) virus-to-bacteria Ocean was explained by the abundance of cyanobac-
ratios. Error bars indicate ±1 SE; when invisible the error was teria; chlorophyll-a concentration explained some var-
less then the width of the bar. The abundance of cyanobacteria
iation in the coastal Arctic ocean, whereas both
was not determined (nd) from the coastal Arctic Ocean samples.
For each variable, the means are significantly different (A N O V A ) cyanobacterial abundance and chlorophyll-a concen-
from each other when marked by different letters. tration contributed to the final model in lakes (Table 3).
 2008 The Authors, Journal compilation  2008 Blackwell Publishing Ltd, Freshwater Biology, doi:10.1111/j.1365-2427.2008.01992.x
6 J. L. Clasen et al.

Lake Pacific Arctic Discussion

(a)
There were several major findings that stemmed from
Viral abundance (viruses x 106 mL–1)
our analyses. The first is that there were clear
100
differences in viral abundance among environments
and the association of viral abundance and the other
biological variables were different in all three envi-
10 ronments. Secondly, different biological variables
were significant predictors of the observed variation
in viral abundance among environments. Thirdly, the
1 relationship between viral abundance and bacterial
abundance was different in lakes and ocean. Finally,
we suggest that differences in viral production and
0.1 1 10 100 loss rates in the environments may contribute to the
Bacterial abundance (cells x 105 mL–1) observed differences between systems. These results
are discussed below.
(b)
Viral abundance (viruses x 106 mL–1)

100 Methodological considerations

Although both chemistry and physics have profound
effects on the biota of aquatic ecosystems, such data
10
were not available for all environments. Therefore and
because we were most interested in comparing our
results with those from other studies, including
1 Maranger & Bird (1995), our analyses only examined
relationships of viruses with other biological vari-
ables. We did not include estimates of viral abun-
0.01 0.1 1 10 100 dance from many other studies because most were
Chlorophyll-a concentration (µg L–1) based on TEM or on EFM using samples that were
(c)
fixed for an indefinite amount of time before being
processed, and these methods can lead to significant
Viral abundance (viruses x 106 mL–1)

100 underestimates of viral abundance (see Hennes &

10
1 the exception of samples from the Wisconsin lakes
0.01 1 10
Cyanobacterial abundance (cell x 103 mL–1)
Fig. 3 Viral abundance as a function of (a) bacterial abundance,
(b) chlorophyll-a concentration or (c) cyanobacterial abundance
within each of the three aquatic environments sampled. The
ovals represent confidence ellipses (P = 0.68) and are centered
on the mean viral and bacterial abundances in each environ-
ment.
Suttle, 1995; Weinbauer & Suttle, 1997; Wen et al.,
2004).
In the present study, both Yo-Pro-1TM and SYBR
Green ITM were used to estimate viral abundance
because the data sets span several years and projects
and involved multiple investigators. However, with
(WIS 1999), which were processed using a modified
Yo-Pro-1TM method (Xenopoulos & Bird, 1997), the
100 1000 samples were handled as outlined in Wen et al. (2004).
Bettarel et al. (2000) compared different methods for
estimating viral abundance in aquatic samples,
including the original (Hennes & Suttle, 1995) and
modified Yo-Pro-1TM (Xenopoulos & Bird, 1997) and
the SYBR Green ITM (Noble & Fuhrman, 1998)
methods. They did not find a significant difference
between viral abundances estimated with the original

 2008 The Authors, Journal compilation  2008 Blackwell Publishing Ltd, Freshwater Biology, doi:10.1111/j.1365-2427.2008.01992.x
 Viral abundance in oceans versus lakes 7
Table 2 Results of discriminant factor
analysis Viruses, bacteria,
Viruses Viruses, Viruses, bacteria, chlorophyll-a,
Environment (%) bacteria (%) chlorophyll-a (%) cyanobacteria (%)

Lakes 49 90 93 98
Coastal Pacific Ocean 79 92 100 100
Coastal Arctic Ocean 63 88 90 nd
Total 67 90 94 99

After initially grouping the data into environments, the percent of randomly re-sampled
data placed in the correct environment was determined to test the robustness of the
grouping criteria. Several discriminant analyses were performed using the biological
variables listed in the top row.
nd, not determined.

Table 3 Backward step-wise multiple regression analyses of viral abundance

Environment Backward step-wise regression model

Lake VA = 4.83 + 0.49 log(BA) ) 0.29 log(CBA) + 0.20 log(chl-a) (r2 = 0.39, P < 0.0001, n = 95)
Coastal Pacific Ocean VA = 4.70 + 0.37 log(BA) + 0.15 log(CBA) (r2 = 0.52, P < 0.0001, n = 295)
Coastal Arctic Ocean VA = 4.56 + 0.36 log(BA) + 0.0.95 log(chl-a) (r2 = 0.26, P < 0.0001, n = 138)

Data included as independent variables in the lake and coastal Pacific Ocean models are the log of bacterial abundance, chlorophyll-a
concentration and cyanobacterial abundance. The coastal Arctic Ocean model contained only the log of bacterial abundance and
chlorophyll-a concentration as independent variables.
VA, viral abundance; BA, bacterial abundance; chl a, chlorophyll-a concentration; CBA, cyanobacterial abundance.

and modified Yo-Pro-1TM method, but reported lower
(by c. 13%) estimates when SYBR Green ITM was used;
Wen et al. (2004) suggested that this difference prob-
ably occurred because the SYBR Green ITM samples
were fixed and stored for up to 1 week before being
processed. However, when water samples are fixed
and processed immediately, as was done in our study,
there is no difference in the accuracy of Yo-Pro-1TM
and SYBR Green ITM (Wen et al., 2004).
Bacterial abundances in the present study were
estimated using four different fluorochromes (AO,
DAPI, Yo-Pro-1TM and SYBR Green I TM). Posch et al.
(2001) reported that estimates of bacterial abundance
made with DAPI were about 10% lower than those
made with AO, while Suzuki, Sherr & Sherr (1993)
reported that DAPI counts in coastal seawater aver-
aged only about 70% of those made with AO. In our
study, DAPI was only used for freshwater samples,
with the exception of the Wisconsin (WIS 1999)
samples which were stained using the modified
Yo-Pro-1TM method. As estimates of bacterial abun-
dance were significantly higher in lakes than in either
the Pacific or Arctic Ocean samples, if bacterial
abundance was underestimated in the lake samples
by using DAPI, it would only exaggerate the differ-
 2008 The Authors, Journal compilation  2008 Blackwell Publishing Ltd, Freshwater Biology, doi:10.1111/j.1365-2427.2008.01992.x
ences reported in this study. Noble & Fuhrman (1998)
determined bacterial abundances with AO and SYBR
Green ITM and found no difference between the two
fluorochromes. Although there are no direct com-
parisons of bacterial abundance estimates made with
Yo-Pro-1TM relative to DAPI or AO, one would
anticipate that if there was a bias, estimates made
with Yo-Pro-1TM (and SYBR Green ITM) would be
higher because of its higher fluorescence yield. How-
ever, there is no evidence for such a bias (see Noble &
Fuhrman, 1998); our bacterial abundances are similar
to those reported previously for each of the environ-
ments (Chrzanowski et al., 1996; Wilhelm, Brigden &
Suttle, 2002; Kirchman et al., 2007).
Finally, determination of viral abundances and
other biological variables by multiple investigators in
the present study also could have introduced a bias,
especially if a given investigator only worked in one
environment. However, a few investigator deter-
mined abundances in different environments and all
were trained by the same individual, reducing the
potential bias associated with the involvement of
multiple investigators. Overall, these arguments
suggest that results of the present study largely
reflect innate differences among environments, and
8 J. L. Clasen et al.
are not rooted in methodological artifacts resulting
from the use of different methods or materials or
the involvement of several investigators in data
acquisition.
Differences across environments
Viral abundance was significantly different among
the environments surveyed in this study (Fig. 2;
Table 1). Additionally, there were clear differences
between environments in the association of viral
abundance and the other biological variables deter-
mined (Fig. 3, Table 3). Together, these results sug-
gest that different factors influence viral abundance
in different aquatic environments. For example, our
data from lakes and the coastal Pacific and Arctic
Oceans combined with data collected within and
outside neutrally buoyant hydrothermal vent
plumes in the deep Pacific ocean (Ortmann & Suttle,
2005) clearly indicate that there are differences
among environments in the relationships between
viral and bacterial abundances (Fig. 4). Data from
Parada et al. (2007) showing very high abundances
of viruses relative to bacteria in the deep waters of
the North Atlantic also support the claim that there
are different controls on viral abundances among
Lake Pacific Arctic
Viral abundance (viruses x 106 mL–1)
100
10
1
0.1
0.1 1 10
Bacterial abundance (cells x 105 mL–1)
Fig. 4 Relationships between viral and bacterial abundances
from the present study and another published data set, the deep
Pacific Ocean (Ortmann & Suttle, 2005). Ovals represent confi-
dence ellipses (P = 0.68) and are centered on the mean viral and
bacterial abundances in each environment. The lines are the
linear regression lines.
 2008 The Authors, Journal compilation  2008 Blackwell Publishing Ltd, Freshwater Biology, doi:10.1111/j.1365-2427.2008.01992.x
1.0
a
0.8
Slope coefficients
b b
0.6
b
0.4
0.2
0.0
Lakes Pacific Arctic Deep
Ocean Ocean Ocean
Fig. 5 Comparison of the slope coefficients generated from
Fig. 4 (viral abundance regressed against bacterial abundance in
lakes and the coastal Pacific and Arctic Oceans and the deep
Pacific Ocean). Error bars indicate ±1 SE around the regression
slope. Means indicated by different letters are significantly
different (A N O V A ) from each other.
environments. Although the data from the different
environments cluster into separate groups (Fig. 4),
the regression slope coefficients from the marine
environments (coastal Pacific, coastal Arctic, deep
Pacific) were remarkably similar to each other, but
were significantly different from the slope coefficient
from the lakes (Fig. 5; A N O V A , F1,3 = 5.71, P < 0.001),
suggesting a fundamental difference between lakes
Deep Pacific and oceans.
Maranger & Bird (1995) compared viral abundance
with other biological variables for 22 Québec lakes
and marine data from the literature and found that
marine and freshwaters differed significantly but the
relationships were not the same as those in our study.
They found a significant relationship between viral
abundance and chlorophyll-a concentration, but not
between viral and bacterial abundances in lakes, as
was found in our study (Fig. 3). However, Maranger
& Bird (1995) fixed and stored their samples at 4 C
for up to a month before determining viral abundance
by TEM; which would have led to inaccurate esti-
mates of viral abundance that cannot be corrected
100
because fixation effects vary substantially (Hennes &
Suttle, 1995; Bettarel et al., 2000; Wen et al., 2004). In
our study, the average lake VBR was lower than those
found in the ocean environments (Fig. 2c, Table 1),
whereas the opposite pattern was reported by
Maranger & Bird (1995). Other studies have also
found high VBRs in lakes (Jacquet et al., 2005; Ram

et al., 2005). The reason for this difference in lake
VBRs is unknown, but may indicate that there is
variability among different lakes types. Yet, the most
interesting comparison is the relationship between
viral and bacterial abundances (Fig. 4), which may not
be entirely captured by VBRs. As an example, while
the VBRs in the Pacific and Arctic Ocean are signif-
icantly different from each other (Fig. 2c, Table 1),
their regression slope coefficients are not (Fig. 5),
suggesting that similar processes are driving the
relationship between these two variables despite
differences in their absolute abundances.
Causes of environmental difference
Our data show a different relationship between viral
and bacterial abundances in lakes and oceans (Figs 4
& 5). Variations in many factors could explain this
difference, as viral abundance is affected by viral
production and loss rates, which can be influenced by
burst size, frequency of infected cells, host diversity,
UV radiation and viral decay rates.
The observed variation in the virus and bacteria
relationships could be caused by differences in the
burst sizes and ⁄ or the percent of infected bacteria. A
recent review by Parada, Herndl & Weinbauer
(2006) reported a higher average burst size and
frequency of visibly infected bacterial cells in fresh
waters, suggesting that viral production rates are
higher in lakes. Therefore, if viral loss rates are
similar in marine and fresh waters, one would
expect this to result in more viruses compared to
bacteria in lakes. This is the opposite of what we
observed (Figs 2–4), suggesting that mechanisms
other than burst size and infection rates are
controlling the relative abundance of viruses and
bacteria in lakes.
Assuming equal viral loss rates in the two systems
may be unrealistic, as unequal loss rates could explain
the difference. Processes that could influence loss
rates include viral adsorption to particles, degradation
by virucidal compounds or UV radiation (Berry &
Noton, 1976; Kapuscinski & Mitchell, 1980; Suttle &
Chen, 1992). However, the potential difference in the
magnitude of viral loss in fresh water and marine
systems is largely unknown. Loss rates could be
higher in lakes because of high concentrations of clays
and chemicals from the terrestrial environment, which
are known to contribute to viral decay (see Brussaard,
 2008 The Authors, Journal compilation  2008 Blackwell Publishing Ltd, Freshwater Biology, doi:10.1111/j.1365-2427.2008.01992.x
Viral abundance in oceans versus lakes 9
2004). Since a difference in loss rates could explain our
observed difference between fresh water and marine
environments, further research is warranted.
Alternatively, the decoupling of viruses and bacte-
ria in lakes compared to the ocean environments
could be due to an increased importance of viruses
infecting other organisms in lakes. Lakes had among
the highest photosynthetic biomass (as measured by
chlorophyll-a concentration and cyanobacterial abun-
dance; Fig. 2b, Table 1) and both cyanobacteria and
chlorophyll-a were significant predictors of viral
abundance in the lake regression model (Table 3),
whereas this was not the case for the marine data. This
suggests that the higher slope coefficient for the lake
data could be related to the higher photosynthetic
biomass; lakes could have a higher contribution of
viruses from photosynthetic hosts, as was originally
suggested by Maranger & Bird (1995) based upon
their lakes survey in Québec.
Potential ecological implications
The ecological implications of the different relation-
ships between viruses and host cells in the different
environments are potentially important. For example,
the observed lower abundance of viruses relative to
bacteria in these lakes suggests that viral-mediated
mortality of bacteria may be less important in lakes
compared to oceans. Of course more extensive sam-
pling and research is necessary to support this
statement. However, if the difference is found to be
true, the role of viruses in biogeochemical cycling
could be affected. For example, nutrients moving
through the viral shunt (Thingstad et al., 1993; Suttle,
2005) could have a less important role in lakes. This
could have ecosystem-wide implications, including
effects on productivity, and would be worth exploring
further.
Acknowledgments
The authors gratefully acknowledge the assistance of
present and past members of the Suttle lab who
collected many of the samples, as well as the crews of
the CCGS Vector and the CCGS Amundsen. A.C.
Ortmann kindly provided her original deep Pacific
abundance data. M.-E. Garneau and W.F. Vincent
from the Canadian Arctic Shelf Exchange Study
(CASES) provided the Arctic Ocean chlorophyll-a
10 J. L. Clasen et al.
data. Collection of some of the lake samples was made Harrison P.J., Fulton J.D., Taylor F.J.R. & Parsons T.R.
possible by the support of the staff and students at the (1983) Review of the biological oceanography
University of Wisconsin’s Trout lake station and of the Straight of Georgia: pelagic environment.
J.J. Elser (IRCEB NSF grant DEB 9977047). The Canadian Journal of Fisheries and Aquatic Sciences, 40,
manuscript was improved by comments from 1064–1094.
Hennes K.P. & Suttle C.A. (1995) Direct counts of viruses
C. Chénard, J. Grainger and A.C. Ortmann, two
in natural waters and laboratory cultures by epifluo-
anonymous reviewers and editors. This research was
rescence microscopy. Limnology and Oceanography, 40,
supported the Natural Science and Engineering 1050–1055.
Research Council of Canada through Discovery and Hobbie J.E., Daley R.J. & Jasper S. (1977) Use of nuclepore
Network (CASES) grants (C.A.S), and a graduate filters for counting bacteria by fluorescence micros-
student fellowship (J.L.C), as well as by an Environ- copy. Applied and Environmental Microbiology, 33, 1225–
mental Research Fellowship from the Lake of the 1228.
Woods Distinct Properties Owners Association Jacquet S., Domaizon I., Personnic S., Sriram A., Ram P.,
(J.L.C). Heldal M., Duhamel S. & Sime-Ngando T. (2005)
Estimates of protozoan-and viral-mediated mortality
of bacterioplankton in Lake Bourget (France). Freshwa-
References ter Biology, 50, 627–645.
Jiang S.C. & Paul J.H. (1994) Seasonal and diel abundance
Berry S.A. & Noton B.G. (1976) Survival of bacterio-
of viruses and occurrence of lysogency ⁄ bacteriocinog-
phages in seawater. Water Research, 10, 323–327.
eny in the marine environment. Marine Ecology Progress
Bettarel Y., Sime-Ngando T., Amblard C. & Laveran H.
Series, 104, 163–172.
(2000) A comparison of methods for counting viruses
Juday C. & Birge E.A. (1941) Hydrography and morpho-
in aquatic systems. Applied and Environmental Microbi-
metry of some northeastern Wisconsin lakes. Transac-
ology, 66, 2283–2289.
tions of the Wisconsin Academy of Science, 33, 21–72.
Brussaard C.P.D. (2004) Viral control of phytoplankton
Kapuscinski R.B. & Mitchell R. (1980) Processes control-
populations – a review. The Journal of Eukaryotic
ling virus inactivation in coastal waters. Water Research,
Microbiology, 51, 125–138.
14, 363–371.
Carmack E.C., MacDonald R.W. & Jasper S. (2004)
Kepner R.L., Wharton R.A. & Suttle C.A. (1998) Viruses
Phytoplankton productivity on the Canadian Shelf of
in Antarctic lakes. Limnology and Oceanography, 43,
the Beaufort Sea. Marine Ecology Progress Series, 277,
1754–1761.
37–50.
Kirchman D.L., Elifantz H., Dittel A.I., Malmstrom R.R.
Chrzanowski T.H., Kyle M., Elser J.J. & Sterner R.W.
& Cottrell M.T. (2007) Standing stocks and activity of
(1996) Element ratios and growth dynamics of bacteria
Archaea and bacteria in the western Arctic Ocean.
in an oligotrophic Canadian Shield lake. Aquatic
Limnology and Oceanography, 52, 495–507.
Microbial Ecology, 11, 119–125.
Manly B.F.J. (2005) Multivariate Statistical Methods: A
Cleugh T.R. & Hauser B.W. (1971) Results of initial
Primer. Chapman and Hall ⁄ CRC, Boca Raton, FL.
survey of experimental lakes area, northwestern
Maranger R. & Bird D.F. (1995) Viral abundance in
Ontario. Journal of the Fisheries Research Board of Canada,
aquatic systems: a comparison between marine and
28, 129–137.
fresh waters. Marine Ecology Progress Series, 121,
Cochlan W.P., Wikner J., Steward G.F., Smith D.C. &
217–226.
Azam F. (1993) Spatial distribution of viruses, bacteria
Noble R.T. & Fuhrman J.A. (1998) Use of SYBR Green I
and chlorophyll a in neritic, oceanic and estuarine
for rapid epifluorescence counts of marine viruses and
environments. Marine Ecology Progress Series, 92, 77–87.
bacteria. Aquatic Microbial Ecology, 14, 113–118.
Garneau M.E., Vincent W.F., Alonso-Stez L., Gratton Y. &
Ortmann A.C. & Suttle C.A. (2005) High abundance of
Lovejoy C. (2006) Prokaryotic community structure
viruses in a deep-sea hydrothermal vent system
and heterotrophic production in a river-influenced
indicates viral mediated microbial mortality. Deep-Sea
coastal arctic ecosystem. Aquatic Microbial Ecology, 42,
Research I, 52, 1515–1527.
27–40.
Parada V., Herndl G.J. & Weinbauer M.G. (2006) Viral
Garneau M.E., Roy S., Lovejoy C., Gratton Y. & Vincent
burst size of heterotrophic prokaryotes in aquatic
W.F. (2008) Seasonal dynamics of bacterial biomass
systems. Journal of the Marine Biological Association of
and production on the Arctic shelf: Franklin Bay,
the United Kingdom, 86, 613–621.
western Canadian Arctic. In press.

 2008 The Authors, Journal compilation  2008 Blackwell Publishing Ltd, Freshwater Biology, doi:10.1111/j.1365-2427.2008.01992.x

Parada V., Sintes E., van Aken H.M., Weinbauer M.G. &
Herndl G.J. (2007) Viral abundance, decay and diver-
sity in the meso- and bathypelagic waters on the North
Atlantic. Applied and Environmental Microbiology, 73,
4429–4438.
Paul J.H., Rose J.B., Jiang S.C., Kellogg C.A. & Dickson L.
(1993) Distribution of viral abundance in the reef
environment of Key Largo, Florida. Applied and Envi-
ronmental Microbiology, 59, 718–724.
Porter K.G. & Feig Y.S. (1980) The use of DAPI for
identifying and counting aquatic microflora. Limnology
and Oceanography, 25, 943–948.
Posch T., Loferer-Krossbacher M., Gao G., Alfreider A.,
Pernthaler J. & Psenner R. (2001) Precision of bacte-
rioplankton biomass determination: a comparison of
two fluorescent dyes, and of allometric and linear
volume-to-carbon conversion factors. Aquatic Microbial
Ecology, 25, 55–63.
Ram A.S.P., Boucher D., Sime-Ngando T., Debroas D. &
Romagoux J.C. (2005) Phage bacteriolysis, protistan
bacterivory potential, and bacterial production in a
freshwater reservoir: coupling with temperature.
Microbial Ecology, 50, 64–72.
Sokal R.R. & Rohlf F.J. (1995) Biometry. W.H. Freeman
and Company, New York, NJ.
Steward G.F., Smith D.C. & Azam F. (1996) Abundance
and production of bacteria and viruses in the Bering
and Chukchi Seas. Marine Ecology Progress Series, 131,
287–300.
Suttle C.A. (2000) Ecological, evolutionary, and geo-
chemical consequences of viral infection of cyanobac-
teria and eukaryotic algae. In: Viral Ecology (Ed. C.J.
Hurst), pp. 247–296. Academic Press, London.
Suttle C.A. (2005) Viruses in the sea. Nature, 437, 356–361.
Suttle C.A. (2007) Marine viruses – major players in the
global ecosystem. Nature Review Microbiology, 5,
801–812.
 2008 The Authors, Journal compilation  2008 Blackwell Publishing Ltd, Freshwater Biology, doi:10.1111/j.1365-2427.2008.01992.x
Viral abundance in oceans versus lakes 11
Suttle C.A. & Chen F. (1992) Mechanisms and rates of
decay of marine viruses in seawater. Applied and
Environmental Microbiology, 58, 3721–3729.
Suzuki M.T., Sherr E.B. & Sherr B.F. (1993) DAPI direct
counting underestimates bacterial abundances and
average cell size compared to AO direct counting.
Limnology and Oceanography, 38, 1566–1570.
Thingstad T.F., Heldal M., Bratbak G. & Dundas I. (1993)
Are viruses important partners in pelagic food webs.
Trends in Ecology and Evolution, 8, 209–212.
Weinbauer M.G. (2004) Ecology of prokaryotic viruses.
FEMS Microbiology Reviews, 28, 127–181.
Weinbauer M.G. & Suttle C.A. (1997) Comparison of
epifluorescence and transmission electron microscopy
for counting viruses in natural marine waters. Aquatic
Microbial Ecology, 13, 225–232.
Wen K., Ortmann A.C. & Suttle C.A. (2004) Accurate
estimation of viral abundance by epifluorescence
microscopy. Applied and Environmental Microbiology,
70, 3862–3867.
Wetzel R.G. & Likens G.E. (1991) Limnological Analysis.
Springer-Verlag, New York, NJ.
Wilhelm S.W., Brigden S.M. & Suttle C.A. (2002) A
dilution technique for the direct measurement of viral
production: a comparison in stratified and tidally
mixed coastal waters. Microbial Ecology, 43, 168–173.
Wommack K.E. & Colwell R.R. (2000) Virioplankton:
viruses in aquatic ecosystems. Microbiology and Molec-
ular Biology Reviews, 64, 69–114.
Xenopoulos M.A. & Bird D.F. (1997) Virus à la sauce
Yo-Pro: microwave-enhanced staining for counting
viruses by epifluorescence microscopy. Limnology and
Oceanography, 42, 1648–1650.
(Manuscript accepted 23 December 2007)