INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

In a seminal paper, Pomeroy (47) showed that bacteria play a major role in the cycling of energy and matter in aquatic systems. The development of techniques which allowed quantification of bacterial abundance (31) and production (21) was a milestone in the investigation of the ecology of bacterioplankton. Later, Azam et al. (2) developed the concept of the "microbial loop," in which bacteria recycle organic matter which otherwise would be lost from the food web. These findings have stimulated a large amount of research on the mechanisms which regulate bacterial biomass and processes in aquatic systems.

There is an ongoing debate about whether bacterial production and biomass are regulated by available resources (bottom-up control) or by predators (top-down control). On the basis of a cross-system survey, Billen et al. (8) argued that bacteria are controlled by resources. Similar conclusions were drawn from other cross-system investigations (9, 14), and Pace and Cole (44) found no evidence in experimental studies that protozoa effectively regulate bacterial abundance. Other workers have argued that bacterial mortality is largely due to protist grazing (19, 54), and after reviews of the literature, Sanders et al. (51) and Berninger et al. (6) described a strong relationship between bacterial abundance and heterotrophic nanoflagellate (HNF) abundance and suggested that significant predatory control of bacteria occurs. However, it has also been shown that bacteria and HNFs are not strongly coupled across systems, and, consequently, HNFs do not always control bacterial abundance (25), probably because of predatory control of HNFs by larger zooplankton (e.g., daphnids) (24). Ducklow and Carlson (18) have argued that the control mechanisms may change seasonally. The finding that the range of estimated clearance of bacteria in the water column due to HNF grazing is large, 5 to 250% per day (1), further supports the notion that the effect of grazing on the control of bacterioplankton changes with time and space. Thus, the key problem might be determining where and when protist grazing is important for regulating bacterioplankton.

In the late 1980s it was shown that in marine and limnetic systems viral particles occur in great numbers which usually exceed even the bacterial numbers (5, 48, 59). It was concluded that the majority of viruses are bacterial viruses (bacteriophages) and that viral lysis is a major cause of bacterial mortality. On average, ca. 10 to 20% of the bacterial production is lysed daily by viruses (58). Thus, viral lysis is an additional mechanism which may contribute to the regulation of bacterial production and processes. As viruses cause mortality of bacteria, they are responsible in part for the top-down type of control, as are the protists. The effect of viral lysis on bacterial mortality has been compared with the effect of flagellate grazing in various oceanic systems (22, 28, 55, 66), and the results have shown that the proportion of bacterial production that is removed by viral lysis can be as high as the proportion that is removed by grazing. A major consequence of these findings is that a significant fraction of bacterial carbon is probably not transferred to higher trophic levels, but is cycled in a bacterium-virus-dissolved organic matter (DOM) loop. Thus, this "viral loop" (11) might be a mechanism for controlling bacterial production.

Some information concerning the spatiotemporal distribution of viruses, the mechanisms controlling viral abundance and infectivity, and the role of viral lysis in regulating bacterial production in oceanic systems has been accumulated (12, 20, 23, 58, 61), whereas there have been only a few reports on limnetic systems (30, 37-39). The only two reports which have estimated the contribution of viral lysis to bacterial mortality in freshwater systems indicate that viral lysis can occasionally be an important factor in the control of bacterial production (30, 39). Our knowledge concerning the significance of viral lysis for bacterial production in oxic waters of limnetic systems is still sparse, and there is no information concerning the significance of viral lysis for bacterial production in anoxic waters. Since grazing rates are typically low in anoxic waters, other mechanisms, such as viral lysis, must be responsible for bacterial mortality. Thus, we quantified the roles of viral lysis and flagellate grazing in the control of bacterial production in the oxic and anoxic water layers of a stratified lake. Our data indicate that the rates of removal of bacterial production by flagellate grazing decreased with depth (i.e., with decreasing oxygen concentration), whereas viral lysis was predominant in anoxic waters.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Sampling and study site. The study site was Lake Plußsee (10°23'E, 54°10'N) near Plön in Schleswig-Holstein in northern Germany. Protection against wind exposure, a lack of discharging rivers or creeks, and a comparatively small size result in stratification of the lake which is extremely stable and lasts from spring to fall (35). On 23 September 1996 water samples were collected along a depth profile with a Ruttner sampler from a permanent platform located in the center of the lake. Subsamples were preserved in formaldehyde (final concentration, 2%) and kept at 4°C in the dark. Oxygen and temperature profiles were determined with a microprocessor oximeter (model OX 196; Wissenschaftlich-Technische Werkstätten, Weilheim, Germany) equipped with a model 2BK 90190 oxygen detector and a thermometer.

Microbial cell counts and chlorophyll a concentrations. Bacteria and viruses were stained with 4',6-diamidino-2-phenylindole (DAPI) (final concentration, 1 µg ml¹) and were enumerated by epifluorescence microscopy by using slight modifications of the protocols described by Turley (62) and Suttle (57). Samples (1 ml) examined for bacteria and viruses were stained without DNase treatment for 30 min, filtered onto 0.02-µm-pore-size Anodisc filters (Whatman), and enumerated by epifluorescence microscopy (Axiovert model 135TV microscope; Zeiss) as described by Weinbauer and Suttle (67). Samples (5 ml) examined for HNFs were stained with DAPI, filtered onto 0.2-µm-pore-size black polycarbonate filters (Nuclepore), and enumerated by epifluorescence microscopy as described by Sherr et al. (53). For two samples of every depth layer, bacterial counts obtained by using DAPI staining and epifluorescence microscopy were compared with bacterial counts obtained by transmission electron microscopy (TEM) (see below), and the values obtained by the two methods did not differ significantly (data not shown). Chlorophyll a concentrations were determined spectrophotometrically as described by Parsons et al. (45).

Bacterial production. Bacterial production was estimated by the [³H]thymidine (83.0 Ci mmol¹; Amersham) incorporation method (21). Samples from various depths were placed in 100-ml acid-cleaned conical shoulder reagent bottles and spiked with [³H]thymidine at a final concentration of 20 nM. The rate of incorporation of [³H]thymidine into the trichloracetic acid-insoluble macromolecular fraction is constant for bacterioplankton in Lake Plußsee at concentrations of 15 nM (13). The flasks were mounted on iron racks, protected from sunlight by polyvinyl chloride tubes, and deployed from a fixed platform to the depths from which the samples were collected previously. The samples were deployed for ca. 2 h, and the incorporation of label was stopped with formaldehyde (final concentration, 2%). Samples were deployed in duplicate, and duplicate formaldehyde-killed samples were used as controls. Five-milliliter subsamples were filtered onto cellulose nitrate filters (pore size, 0.22 µm; type GSWP; Millipore), and the [³H]thymidine was extracted by two 10-min incubations with 5% ice-cold trichloroacetic acid (Sigma Chemical Co.). The filters were dissolved with a scintillation cocktail, and radioactivity was determined with a model 8500 Packard Tri-Carb scintillation counter. Conversion factors for relating thymidine incorporation to cell production were obtained from values determined in the fall during a seasonal study performed in Lake Plußsee, and these conversion factors were 1.92 × 10⁶ cells pmol¹ for the euphotic zone and 1.57 × 10⁶ cells pmol¹ for deeper water (13).

Determination of visibly infected bacteria, viral burst size, and cell integrity. A TEM-based method was used to determine the number of visibly infected cells (64). In the modified method used, samples were not incubated with streptomycin, and a lower centrifugation speed and less time were used. Bacteria from 10-ml samples were collected quantitatively onto Formvar-coated, 400-mesh electron microscope grids by centrifugation at 66,000 × g for 20 min in a swinging-bucket rotor (Beckmann type SW-41), stained for 30 s with 1% uranyl acetate, and rinsed three times with deionized distilled water. The time and speed of centrifugation used reduced the disruption of infected bacteria, and, as few viruses were pelleted, phage in bacteria were easily distinguished. Grids were examined for visibly infected cells by using a TEM (model CEM 902; Zeiss) operating at an accelerating voltage of 80 kV. Between 200 and 2,000 cells per sample were examined for mature phages inside the cells in order to obtain at least 10 visibly infected cells, and a minimum of five phage were observed in each visibly infected cell. Viruses inside cells were identified based on structure, size, intensity of staining, and uniformity of structure, size, and staining intensity. The viral burst size (i.e., the number of viruses produced in a cell) was estimated from all of the visibly infected cells in a sample as described previously (64). Total bacterial abundance values were also obtained from the TEM grids (67).

Flagellate grazing. Grazing rates were estimated as described by Steward et al. (55) as the product of the bacterial abundance and the HNF abundance determined along the depth profile and clearance rates obtained from previously published papers. The average clearance rates (± standard error) of HNFs in Lake Plußsee during the entire stratification period were 0.60 ± 0.38 nl cell¹ h¹ in the epilimnion, 0.60 ± 0.40 nl cell¹ h¹ in the metalimnion, and 0.27 ± 0.13 nl cell¹ h¹ in the hypolimnion (calculated from the data in reference 40).

Bacterial mortality and viral production. A model was used to estimate bacterial mortality due to viral lysis (50). This model assumed that the proportion of bacterial mortality due to viral lysis was about equal to the FVIC multiplied by conversion factors (average, 10.84; range, 7.4 to 14.28). Bacterial mortality due to flagellate grazing was calculated by dividing the grazing rate by the bacterial production rate. Summed bacterial mortality was calculated by determining the sum of mortality due to grazing and mortality due to viral lysis. In a steady-state system bacterial mortality due to viral lysis matches the bacterial production which is removed by lysis (58). Thus, multiplying the lysed bacterial production by the burst size yielded viral production. The concentration of carbon released by viral lysis of bacteria was calculated from the viral lysis rate of bacteria, the cell volumes determined in Lake Plußsee along the depth profile on 23 September 1996 (63), and a conversion factor of 350 fg of C µm³ (36) for relating the bacterial biovolume to the carbon content. The release of nitrogen and phosphorus was estimated by using an N/C ratio of 0.26 (10) and a P/C ratio of 0.04 (15). We further assumed that all biomass was released as dissolved compounds.

Contact rates. The rate of contact (R) between viruses and bacteria was calculated by using the following formulae (43): R = (Sh2wD_v)VP, where Sh is the Sherwood number (1.06 for a bacterial community with 10% motile cells [68]), w is the cell diameter (calculated from the mean bacterial cell volume determined at each depth [63], assuming that the cells are spheres), V and P are the abundance of viruses and the abundance of undamaged cells, respectively, and D_v is the diffusivity of viruses, and D_v = kT/3µd_v, where k is the Boltzmann constant (1.38 × 10²³ J K¹), T is the in situ temperature (in degrees Kelvin), µ is the viscosity of water (in pascals per second; µ was calculated from values given by Schwörbel [52] for temperatures in the range from 4 to 15°C), and d_v is the diameter of the viral capsid (73 nm in Lake Plußsee [16]). The contact rate was corrected for bacterial abundance to estimate the number of contacts per cell on a daily basis.

Carbon, phosphorus, and nitrogen concentrations. On 23 September 1996, soluble reactive phosphorus (SRP), NO₃, and NH₄ concentrations were determined along the depth profile by routine limnological methods (27) and were provided by the Max Planck Institute for Limnology, Plön, Germany. Dissolved organic carbon (DOC) concentrations were determined from water samples passed through 0.2-µm-pore-size polycarbonate filters (Nuclepore) that were rinsed with Milli Q water. DOC concentrations were measured by the high-temperature combustion method (56) by using a Shimadzu model TOC-5000 analyzer with a platinum catalyst on quartz and performing regular blank monitoring (4). The contamination of samples by leaching of carbon from the polycarbonate filters was less than 5% of the DOC concentration (data not shown).

Statistical analyses. All data were log transformed for statistical analyses. Analysis of variance and Fisher PLSD post hoc tests (StatView D-4.5 program) were used to determine whether parameters differed significantly between depth layers. A probability of <0.05 was considered significant.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Characterization of study site. Temperature and oxygen profiles showed that the pelagic zone of Lake Plußsee was stratified and separated into three distinct layers, the oxic epilimnion, the thermocline layer (metalimnion), and the anoxic hypolimnion (Fig. 1). As determined by an analysis of variance, the SRP and NH₄ concentrations increased significantly with depth, whereas the DOC and NO₃ concentrations decreased significantly with depth. The DOC concentrations ranged from 9.2 to 14.5 mg liter¹.

View larger version (25K): [in this window] [in a new window]   FIG. 1.   Depth profiles for temperature and oxygen, DOC, and inorganic nitrogen and phosphorus (SRP) concentrations in Lake Plußsee on 23 September 1996.

Bacterial production and abundance of microorganisms. All microbial parameters differed along the depth profile (Fig. 2). Bacterial production was significantly higher in the metalimnion than in the other layers, which is consistent with the high bacterial numbers in this layer. Also, the viral abundance values and chlorophyll a concentrations were significantly higher in the metalimnion than in the epilimnion and the hypolimnion, whereas the HNF abundance values decreased significantly with depth. For one depth of every depth layer thin sections and TEM were used to check for cell integrity. We found that between 70.8 and 79.1% (average, 73.9%) of the cells were intact (Table 1).

View larger version (28K): [in this window] [in a new window]   FIG. 2.   Depth profiles for microbial parameters in Lake Plußsee on 23 September 1996. Chl a, chlorophyll a.

Visibly infected bacteria. For one depth of every depth layer we determined the FVIC by two methods, whole-cell examination and thin-section examination (Table 1). The FVIC obtained in the whole-cell examination were between 74.8 and 85.0% (average, 78.7%) of the FVIC obtained in the thin-section examination. In all samples we observed bacterial cells which were visibly infected with viruses. When corrected for the average percentage of nonintact cells, the FVIC ranged from 0.7 to 9.0% of the total bacterial numbers, and the FVIC increased significantly with depth (Fig. 2), averaging 1.6% in the epilimnion, 3.0% in the metalimnion, and 6.3% in the hypolimnion.

Bacterial mortality and viral production. The bacterial mortality due to viral lysis and flagellate grazing was estimated along the depth profile in Lake Plußsee (Fig. 3) and was expressed as a ratio of the rate of removal (by lysis or grazing) to the bacterial production rate. We estimated that viral lysis removed on average 7.7 to 27.8% of the bacterial production in the epilimnion, 19.6 to 46.8% of the bacterial production in the metalimnion, and 38.4 to 97.3% of the bacterial production in the hypolimnion. Bacterial mortality due to flagellate grazing showed a trend opposite that found for viral lysis. While flagellate grazing removed on average 81.8 to 108.0% of the bacterial production in the epilimnion, the levels of mortality due to grazing were 2.9 to 27.6% in the metalimnion and 5.0 to 8.9% in the hypolimnion. In the epilimnion the summed mortality (calculated by determining the sum of the average mortality due to viral lysis and the average mortality due to grazing) was 66.6 to 128.5%, compared to 22.4 to 56.7% in the metalimnion and 43.4 to 103.3% in the hypolimnion. In the epilimnion 8.4 to 41.8% of the summed mortality could be ascribed to viral lysis, compared to 51.3 to 91.0% in the metalimnion and 88.5 to 94.2% in the hypolimnion. The average ratios of viral lysis to grazing were 0.3 in the epilimnion, 6.0 in the metalimnion, and 10.6 in the hypolimnion (Table 2).

View larger version (26K): [in this window] [in a new window]   FIG. 3.   Depth profile for bacterial mortality due to viral lysis and grazing in Lake Plußsee on 23 September 1996. The summed mortality was calculated by determining the sum of the mortality due to viral lysis and the mortality due to grazing. Error bars indicate the ranges of conversion factors for viral lysis and from the standard error for grazing.

1

1

1

1

1

1

Contact model. In the epilimnion we calculated an average contact rate of 49.0 viral contacts cell¹ day¹, compared to 180.4 viral contacts cell¹ day¹ in the metalimnion and 141.8 viral contacts cell¹ day¹ in the hypolimnion (Fig. 4).

View larger version (19K): [in this window] [in a new window]   FIG. 4.   Viral contact rates in different depth layers of Lake Plußsee on 23 September 1996. Contact rates were estimated from a contact model for viruses and bacteria (43). The values are the means ± standard deviations from four samples.

Organic matter set free during viral lysis of bacteria. The estimated concentrations of organic carbon, nitrogen, and phosphorus released during viral lysis of bacterioplankton were more than 1 order of magnitude lower in the epilimnion than in the two deeper water layers (Table 3). The highest release rates were found in the hypolimnion. We estimated that the contribution of organic carbon released by viral lysis to the standing stock of DOC was <0.1% per day.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

The most significant finding of this study was the shift in the mechanisms controlling bacterial production along the depth profile. Our data indicate that grazing was the main cause of bacterial mortality in the oxic epilimnion, whereas viral lysis predominated in the anoxic hypolimnion. This shift may have major implications for the fate and cycling of organic matter in stratified lakes.

Critical assumptions in estimation of bacterial mortality due to viral lysis. Conversion factors were used to relate the FVIC to bacterial mortality due to viral lysis. These conversion factors were derived from thin sections of isolated virus-host systems (50), whereas we examined whole cells. However, previously we have argued that the conversion factors are also applicable to whole cells (64). A study has shown that the FVIC determined by using whole-cell examination was much lower than the FVIC obtained by using thin sections (22). The high centrifugation speeds and the times usually used for the whole-cell method can disrupt infected cells (unpublished data) and might have resulted in the lower FVIC obtained with this method by Fuhrman and Noble (22). In our study we reduced the centrifugation speed and time, which avoided disruption of visibly infected cells, and we found that the values obtained from the whole-cell examination averaged 78% of the values obtained by using thin sections (Table 1). Thus, the whole-cell method systematically underestimated the FVIC; however, the differences were moderate and similar at all depths. The difference between the two methods might be due to pigmentation of cells, which prevented examination by the whole-cell method, although viruses were detected within pigmented Synechococcus sp. cells in the Gulf of Mexico (60). Also, cells might have lysed during centrifugation, although we reduced the centrifugation speed and time. Moreover, viruses within cells might be more easily detected in thin sections (detection limit, three viruses) than in whole cells (detection limit, five viruses). Overall, our data indicate that the whole-cell examination resulted in a conservative estimate of the effect of viral lysis on bacteria.

Frequency of visibly infected bacteria. FVIC data are available for a variety of freshwater and marine environments. In marine waters, the FVIC ranged from 0 to 7% in free-living and attached bacteria from various environments (48, 49, 55, 66). Along a salinity gradient in solar salterns, the FVIC were between <0.04 and 3.76% (28). In Lake Constance, the FVIC ranged from <0.1% to 1.7% ± 1.2% (30), and the FVIC in a backwater system of the River Danube were between 0 and 4% (39). In Lake Plußsee the FVIC ranged from 0.7 to 9.0% (Fig. 2). The value obtained for the anoxic hypolimnion (9.0%) is the highest FVIC published so far for natural bacterial communities (when the FVIC was not corrected for damaged cells, a maximum value of 6.4% was observed). For an as-yet-unidentified archaeal species in a solar saltern a maximum FVIC of 6.7% was observed (28). Bratbak et al. (12) obtained values of up to 40% by using whole-cell examination after incubation of the entire bacterial community with streptomycin, which supports lysis of the cells. It is not clear why the streptomycin treatment resulted in those high FVIC. However, since streptomycin can induce the lytic cycle in lysogenized bacteria (7), induction of lysogenized cells could have contributed to the high FVIC.

Frequency of intact cells. It has been known for a long time that only a fraction of the bacterioplankton community is active or alive. The data of Zweifel and Hagström (69) indicated that only 2 to 32% of the members of natural marine bacterial communities collected in the Baltic Sea, the North Sea, and the Mediterranean Sea contained a nucleoid, and these authors assumed that the non-nucleoid-containing bacteria (ghosts) (i.e., the majority of bacterioplankton) are inactive or even dead. However, this method resulted in an underestimate of the living fraction of bacterioplankton in samples from the coastal Pacific Ocean off California when the data were compared with data obtained by autoradiography and with universal 16S rRNA-targeted probes (34). Using electron microscopy, Heissenberger et al. (29) demonstrated that on average ca. 25% of bacteria from the Mediterranean Sea were empty, ca. 25% were damaged (i.e., they had a shrunken cytoplasmic membrane or only remnants of plasma), and ca. 50% were intact. We found that between 70 and 80% of the bacterial cells were intact in Lake Plußsee. Although it is not certain that all intact cells were alive, it can be assumed that viruses can replicate only in undamaged bacteria. Thus, the FVIC was calculated by determining the percentage of intact cells. Since previously published data on the FVIC in aquatic systems were not corrected for damaged cells, they are probably underestimates of the effect of viral lysis on bacterial mortality. However, data on cell integrity should be viewed with some caution, since it is not possible to completely exclude preparation artifacts.

Viral production. In Lake Constance the estimated production of viruses ranged from 0.1 × 10⁹ to 2.5 × 10⁹ viruses liter¹ day¹ (30), whereas the estimated production of viruses in a backwater system of the River Danube ranged from 1.0 × 10¹⁰ to 3.0 × 10¹⁰ viruses liter¹ day¹ (calculated from the bacterial lysis rate and burst sizes given by Mathias et al. [39]). The average levels of viral production in the different water layers of Lake Plußsee ranged from 0.2 × 10¹⁰ to 1.9 × 10¹⁰ viruses liter¹ day¹ (Table 3) and thus were higher than the viral production rates in Lake Constance but slightly lower than the viral production rates found in the River Danube backwater system. Although the FVIC detected in our study occasionally exceeded the values determined by Mathias et al. (39), the viral production rates were lower. This can be explained by the fact that the bacterial production rates were higher in the backwater system (range, 1.4 × 10⁹ to 6.2 × 10⁹ cells liter¹ day¹) than in Lake Plußsee (range, 0.2 × 10⁹ to 1.5 × 10⁹ cells liter¹ day¹) and thus sustained higher rates of viral production. Also, differences in the methods used to assess viral production can account for some of the differences between the studies.

Bacterial mortality. When we used the model of Proctor et al. (50) and the average conversion factor, the bacterial mortality due to viral lysis ranged from 7.7 to 27.8% in the epilimnion, from 19.6 to 46.8% in the metalimnion, and from 38.4 to 97.3% in the hypolimnion of Lake Plußsee. When the same approach was used, the bacterial mortality due to viral lysis ranged from less than 0.11 to 18.4% in Lake Constance (calculated from the data in reference 30) and from 10.8 to 43.2% in a backwater system of the River Danube (39). Thus, the metalimnion and hypolimnion of Lake Plußsee exhibited the greatest viral control of bacterial production found so far for limnetic systems.

Organic matter released by viral lysis of bacteria. DOC concentrations decreased slightly with depth, whereas SRP and NH₄ concentrations increased with depth. The high chlorophyll a concentrations in the photic zone (Fig. 3) might sustain high concentrations of organic carbon (e.g., concentrations due to photosynthetic extracellular release of carbon-rich substances). This depth distribution is typical for late summer stratification (42) and could indicate that phosphorus or (less likely) nitrogen is the limiting nutrient in the epilimnion, whereas in the hypolimnion bacteria were limited by the supply of organic carbon. More detailed studies have shown that the DOM in the hypolimnion of Lake Plußsee consists of refractory carbon skeletons depleted of nitrogen and phosphorus (42). The estimated contribution of organic carbon released by viral lysis of bacteria to the DOC pool was small (<0.2% day¹) (Table 3). Since the DOM in Lake Plußsee is depleted with respect to phosphorus and nitrogen compared to bacteria (42), the contribution of organic phosphorus and nitrogen released by viral lysis into the pool of dissolved organic nitrogen and phosphorus is probably greater than the contribution of organic carbon. Overall, viral lysis of bacteria did not seem to contribute significantly to the DOM pool. However, releases of even small amounts of organic phosphorus and nitrogen in the epilimnion and organic carbon in the hypolimnion might stimulate bacterial growth, since bacterial lysis products are very likely readily utilizable by bacteria. Stimulation of bacterial growth by lysis products was also observed in a marine environment (41). Considering the fact that a cell is encountered by ca. 50 to 180 viruses per day, it is also possible that viruses are an important diet for bacteria. Also, viruses might be a diet for HNFs. Support for this hypothesis comes from estimates that viruses can be an important source of carbon, nitrogen, and phosphorus when there are 10⁶ bacteria ml¹ and 10⁷ to 10⁸ viruses ml¹ (26), as in our study (Fig. 2).

Implications. Our data indicate that in oxic waters of Lake Plußsee the majority of the bacterial production was removed by HNFs and thus could be transferred to higher trophic levels of the food web. The high mortality of bacteria due to viruses in anoxic waters indicates that an important fraction of bacterial production remains in the viral loop. This is supported by the finding based on models and experiments that viral lysis and lysis products can strongly stimulate bacterial production and carbon uptake (20, 22, 41). Also, the addition of concentrates of the natural virus community to seawater incubation mixtures resulted in a higher FVIC and in increased concentrations of dissolved amino acids and carbohydrates (65). Thus, a significant portion of organic matter in anoxic water could be cycled several times in the viral loop before it gets mineralized. On the other hand, Proctor and Fuhrman (49) argued that the cell lysis products could act as glue for the formation of organic aggregates, and Peduzzi and Weinbauer (46) found that the addition of the virus size fraction of DOM to seawater resulted in an increase in the size of organic aggregates. Thus, the high lysis rates of bacteria in anoxic water could cause higher rates of sinking of aggregates and higher rates of incorporation of organic matter into the sediment. Although additional studies will have to show whether viral lysis increases the mineralization of DOM or the rates of incorporation of organic matter in the sediments, it is clear that viral lysis is an important mechanism for regulation of organic matter cycling in lakes.