Optimization of viral resuspension methods for carbon-rich soils along a permafrost thaw gradient

Sample collection

Peatland soil cores were collected in July 2013 and July 2014 from palsa (raised hummock underlain by intact permafrost), moss-dominated (Sphagnum spp.) bog, and sedge-dominated (Eriophorum) fen habitats in Stordalen Mire, in Abisko, northern Sweden. Further description of the study site and habitat classification is reported in Hodgkins et al. (2014). Soil was gathered with an 11 cm-diameter custom circular push corer at palsa sites, and with a 10 cm × 10 cm square Wardenaar corer (Eijkelkamp, The Netherlands) at the bog and fen sites. Each core was carefully removed from the coring device, while maintaining the integrity of the core and its layers, and placed on a sterile surface. A measuring tape and camera were used to measure and document the core. Using a sterile knife, the outer 1 cm of the core was removed and then the knife was re-sterilized (cleaned with 70% ethanol after each use) and used to cut the core in five centimeter increments from 1 to 40 cms (if applicable). Intervals were placed in 50 ml conical tubes and treated in one of two ways, per “storage” methods below. Samples were transported to nearby Abisko Naturvetenskapliga Station, and then to the University of Arizona for processing.

Experimental overview

Three sets of experiments evaluated the impact of buffers, physical dispersion, storage conditions, and concentration and purification methods on viral yields. Experiment 1A/B used 2013 samples, pooled from 1–40 cm (evenly) of each soil core that was refrigerated at 4 °C during transportation (∼1–2 week storage while at the field site and then ∼1 week for transportation to Arizona) and then either stored at −80 °C or 4 °C. The 2013 samples were used to test chemical and physical dispersion methods. Experiment 2 used 2014 samples from two depths and tested frozen versus chilled storage methods. Experiment 3 was an extension of Experiment 2, with the same samples, and tested the efficacy of a common concentration and purification method.

The effectiveness of each method was quantified by epifluorescence microscopy (EFM), to determine the viral abundance of each sample. EFM is considered to be a better enumeration technique than plaque assays, flow cytometry, and transmission electron microscopy (TEM), because it is the most reliable, cost effective, and readily available (Hermes & Suttle, 1995; Weinbauer & Suttle, 1997; Noble & Fuhrman, 1998; Ashelford, Day & Fry, 2003; Brum, Schenck & Sullivan, 2013; Williamson et al., 2013). Samples were mixed with SYBR Gold, ascorbic acid anti-fade solution, and beads of a known concentration (as in Cunningham et al., 2015). Counts of virus-like particles and beads were compared to enumerate the viruses (Cunningham et al., 2015). All treatment comparisons of viral yields were tested for significance using two-tailed paired t-tests. Additional viral characterization was obtained by analysis of VLPs via TEM from 10 fields of view for each sample using previously describe preparation techniques (Fig. S1; Brum, Schenck & Sullivan, 2013).

Experiment 1A: optimizing resuspension buffers

The first step in obtaining viruses from soils is to efficiently resuspend them in a buffer that reduces chemical and electrostatic bonds (Dowd et al., 1998; Michen & Graule, 2010), neutralizes pH (Dowd et al., 1998), normalizes moisture content (Williamson, Radosevich & Wommack, 2005), and stabilizes viral capsids (Hurst, 1988). It has been suggested that isoelectric point and hydrophobicity of individual virus capsids are the most important factors influencing how they interact with soils (Dowd et al., 1998; Williamson, Wommack & Radosevich, 2003). We tested a diversity of buffers that were most promising from the literature in other soil systems to maximize viral yields. Specifically, three permafrost-associated peatland soils were used to compare VLP yields after resuspension in each of three buffers (KC, AKC, and PP) with the soils interrogated by a single physical dispersion method (Fig. 2A). We developed buffer AKC to control for changing chemistry (e.g., changing pH and OM chemistry and concentrations) across the Stordalen Mire soils (Hodgkins et al., 2014): we amended KC with 10% PBS (which helps maintain pH and osmotic balance; Sperelakis, 2012), 5 mM EDTA (a ligand and chelating agent that helps prevent environmental interference; Conway et al., 1999), and 150 mM MgSO₄ (used to stabilize virus capsids; Speir et al., 2006).

Overall, the AKC buffer yielded significantly more VLPs g⁻¹ soil than buffers KC or PP (p < 0.05, two-tailed paired t-test; Fig. 2A). For all buffers tested, VLP yields increased from palsa to bog and then to fen, with fen soils having up to 72% more VLPs than palsa soils (Fig. 2A). This is consistent with previous observations of soil VLP abundance correlating with soil moisture content (Williamson, Radosevich & Wommack, 2005; reviewed in Kimura et al., 2008); here palsas are driest, bogs have a fluctuating water table, and fens are fully inundated. Other factors vary progressively along the thaw gradient too (notably soil OM content and lability) and may also be influencing VLP abundances and/or recoveries (Hodgkins et al., 2014). The VLP counts, of ∼5–7 × 10⁸ VLPs g⁻¹ soil, were in the same order of magnitude as or higher than previously reported using KC across wetland, clay, dune, forest, and park soils enumerated with the same epifluorescence microscopy technique (Williamson et al., 2013). (In that study, KC gave higher viral recoveries across soil types than water or sodium deoxycholate as a buffer; Williamson et al., 2013).

Experiment 1B: Physical dispersion methods

With an empirically optimized resuspension buffer, AKC, we next explored how physical dispersion methods (all with AKC) impacted viral yields across our three soil types. These three methods were chosen as follows: the sonication method has been most heavily tested in the literature, the beads in bead-beating might help with desorption of viruses (both reviewed in Kimura et al., 2008), and vortexing might lead to the least tail breakage (Williamson, Helton & Wommack, 2012).

However, in our samples, Experiment 1B revealed no statistically significant difference in VLP yields across the physical dispersion methods tested (Fig. 2B). VLP counts ranged from 4.59 × 10⁸ (±2.75 × 10⁷) VLPs g⁻¹ soil for sonication in palsa to 7.65 × 10⁸ (±8.83 ×10⁷) VLPs g⁻¹ soil for bead-beating in fen. As in Experiment 1A’s results, VLP yields were again lowest in palsa and highest in fen. Although bead-beating was not significantly better, it did have on average higher VLPs than the other physical dispersion methods. We hypothesize shearing of the humic material helped viruses desorb from the peatland soil.

While no significant differences across the methods were detected, we note that there are many variations of each method reported in the literature, which were not tested here. Variants of the sonication method include storing the samples on ice during sonication (Duhamel & Jacquet, 2006; Danovaro & Middelboe, 2010) and varying the sonication times (Danovaro & Middelboe, 2010). Sonication has been the most widely used method across sediment and soil environments and has performed better when combined with a well-paired resuspension buffer such as the KC buffer examined in Experiment 1A (Williamson, Wommack & Radosevich, 2003; Duhamel & Jacquet, 2006; Danovaro & Middelboe, 2010). Bead-beating methods have also been used frequently (Ashelford, Day & Fry, 2003; Williamson et al., 2013), but there is no evidence for its increased efficacy over other methods. However, not all types and sizes of beads have been tested. In our case, vortexing was performed using the same equipment and method as bead-beating, but without beads. Vortexers come with many different speed settings and attachments, and each could potentially influence viral yields. Finally, while vortexing is thought to reduce (by 20%) tail breakage as compared to sonication methods (Williamson, Helton & Wommack, 2012), we did not investigate this further since our goals were largely to obtain VLPs for metagenomic sequencing.

Experiment 2: optimization of sampling conditions

After Experiments 1A and 1B, we sought to better understand how sampling and storage conditions impacted viral yields across multiple depths. We used AKC buffer and vortexing to resuspend the viruses and examined the effect storage temperature had on VLP abundance. Viral abundances in soils may change with depth like microbial biomass (Mackelprang et al., 2011; Frank-Fahle et al., 2014) and our results revealed that the storage temperature had no significant influence on VLP yields either from shallow or deep samples (Fig. 3), and this finding is consistent with results in aquatic systems and sediments (Brum, Schenck & Sullivan, 2013). Soil virus literature to date, suggests viral community structure is not different with depth (based on genetic material; Adriaenssens & Cowan, 2014), nor does abundance change (comparing 5–10 cm and 10–15 cm from soil pore water; Ballaud et al., 2015), but viral abundance was reported to vary seasonally (Ballaud et al., 2015). To this end, we tested storing both shallow and deep soil samples at chilled (4 °C) and frozen temperatures (−80 °C). In this experiment, the same protocol was followed from Experiment 1, but with AKC buffer and vortexing. Since there was not a significant difference in physical dispersion method from Experiment 1B, vortexing was chosen because bead-beating has been shown to release more humics from soil (i.e., ghost particles; Lakay, Botha & Prior, 2007) and the humic fragments might interfere with Amicon filter concentration (humics interact with minicolumns in Harry et al., 1999) and downstream processing of the samples (i.e., DNA extraction and sequencing; Saano et al., 1995; Zhou, Bruns & Tiedje, 1996).

Our results revealed that frozen samples had higher VLPs g⁻¹ soil than chilled samples for shallow (Fig. 3A) and deep samples (Fig. 3B), but this difference was not significant (i.e., the lowest p value was 0.067 in the deep palsa sample). This general finding of no significant difference in VLP recoveries by storage temperate is consistent with results in aquatic systems and sediments (Brum, Schenck & Sullivan, 2013). One caveat is that these samples are Arctic samples, whose in situ habitat temperatures at the time of collection spanned “chilled” temperatures (∼0–15 °C at 36–40 cm depth, although ∼14–20 °C at 1–5 cm; the coldest conditions were in the deep palsa samples). Therefore, storage at 4 °C may allow microbial—and associated viral—communities (especially those at depth) to remain active, and thus drift from in situ composition, even though total viral recovery numbers did not change. In addition, while freeze-thaw cycles are known to broadly decrease infectivity for viruses (Ward et al., 2004), here the primary goal of recovery is enumeration and genomic sequencing, not culturing, and these viruses are also likely less sensitive to freeze-thaw given that it is a characteristic of their habitat. For these reasons, we conclude that frozen storage is the best choice for obtaining an unaltered viral community for these sample types.

The second part of this experiment tested an ecological difference rather than a methodological one: whether virus abundance changed with depth. Our frozen-storage data show significantly higher VLPs in deep than shallow palsa and fen (28% and 27% higher, respectively, p < 0.05; Fig. 3). Conversely, bog had higher virus abundance in the shallow samples (18%), although it was not significant. In the chilled samples, there was no significant difference between shallow and deep (Fig. 3); the trends were the same as for frozen samples for bog and fen, but opposite in palsa (shallow was 14% higher than deep).

In some aquatic systems virus abundance decreases with depth (Cochlan et al., 1993; Noble & Fuhrman, 1998; Parada et al., 2007), correlating with the decrease in microbial abundance with depth (Buitenhuis et al., 2012; Mediterranean Sea data in Mapelli et al., 2013). This is not surprising, as viral and bacterial abundances have been shown to be moderately correlated in at least some systems (Maranger & Bird, 1995; Anesio et al., 2004). If this trend also occurs in terrestrial environments, then we would expect viral abundance to decrease with depth, as it has been shown that microbial abundance decreases with depth (Mackelprang et al., 2011). In this study, this trend (viral abundance decreasing with depth) is supported only at the bog site (for both storage conditions) and at the palsa site with chilled storage. The palsa and bog habitats both have permafrost, with palsa’s being intact and the bog’s being partially thawed. The presence of permafrost would decrease microbial activity and thus viral activity (Frank-Fahle et al., 2014). However, it is contradicted at the palsa site with frozen storage and the fen site (for both storage conditions). At the palsa site, the depth increases in VLPs may be due to a concentration of virus particles and microbial cells due to the permafrost/active layer boundary for palsa samples, as has been observed for microbial cells at the permafrost boundary in other studies (permafrost in palsa starts ∼40 cm at Stordalen mire; reviewed in Jansson & Taş, 2014). At the fen site, which lacks permafrost, the depth increases in VLPs could be due to a temperature-stratification-based water density gradient in the inundated soil column (average fen temperature at 1–5 cm was 15.8 °C, and 13.5 °C at 30–36 cm). The difference in frozen versus chilled results for the palsa site may be due to the similarity of chilled storage to the in situ palsa deep conditions, due to its proximity to the permafrost table. Therefore, the stored chilled samples could be active and therefore likely to give artificial results. The frozen palsa results of higher VLP abundance at depth may represent in situ biological patterns whereas the chilled samples may be artificial and not represent the in situ viral communities. The chilled samples may instead reflect the viral communities in autumn or spring months when soil temperatures average ∼4 °C because that was the temperature at which these viruses were stored. The palsa habitat is not waterlogged (like a majority of bog habitats and all fen habitat at our site) and therefore the viruses may be more resistant to temperature change because soil has a lower heat capacity than water (Bristow, 1998).

Experiment 3: optimization of concentration and purification methods

Having explored how buffer, dispersion, and sampling conditions impact viral yields (and used improvements therein to assess depth differences in viral abundances at the site), we next tested the efficacy of several concentration and purification methods. Specifically, we tested concentration on Amicon filters with and without BSA addition (for “blocking” of the membranes to decrease viral adhesion), followed by purification on a CsCl gradient.

The addition of BSA to Amicon filters quantitatively, though not significantly, increased viral yields (except bog stored frozen), but only significantly in chilled samples (Fig. 4). This is curious, as BSA has previously been shown to increase viral recovery yield by 2-fold with seawater viruses (Deng et al., 2012). We hypothesize that soil viruses across the three habitats in Stordalen mire may be less “sticky” than ocean viruses so that the blocking agent had little effect on yields off the Amicon filters (lack of “stickiness” of soil viruses characterized in Klitzke et al., 2015). This might reflect the highly charged soil environment, relative to seawater, selecting for reduced charges in soil virus tail fibers and structural proteins as compared to those in ocean viruses (Sobsey et al., 1980; Dowd et al., 1998). Additionally, soil viruses may be less “sticky” because they need to be able to attach to their host for infection and viral transport in soils is slow compared to marine environments because in marine environments viruses are transported by currents (Brum et al., 2015).

Comparing samples with or without BSA in the Amicon step showed no significant effect after CsCl (although half had p values around 0.058; Fig. 5). Purification by CsCl gradient caused significant loss (p < 0.05, two-tailed paired t-test; Fig. S2) of VLPs in half of the samples, with 22–67% loss observed (Fig. S2).This loss is not surprising given that non-virus “ghost” particles such as humics and cell debris, which can be mistakenly counted as VLPs by EFM, are typically present in environments with high organic matter (e.g., as discussed in Suttle & Fuhrman, 2010; Forterre et al., 2013) and the CsCl densities (rho = 1.4–1.52 g/ml) used here are targeted to capture dsDNA viruses, while cell debris, humics, and ssDNA or RNA viruses migrate to lighter densities (Thurber et al., 2009). However, several of these assertions where addressed in other recent research (Williamson et al., 2013; Brum & Sullivan, 2015). The use of EFM, TEM, and flow cytometry enumeration techniques have been carefully examined and in each, the use of EFM was suggested (Hermes & Suttle, 1995; Weinbauer & Suttle, 1997; Noble & Fuhrman, 1998; Brum, Schenck & Sullivan, 2013; Williamson et al., 2013). The lowered VLP counts could be due to the CsCl damaging viral particles during purification, which is known to be osmotically challenging for many viral isolates. In seawater samples, where there is less concern for organic matter and ghost particle background, viral loss was observed from CsCl purification (Hurwitz et al., 2013). Also, others have observed viral recovery for isolates after CsCl purification to average between 53–66% with a range of 13–100% recovery (Chen, Estes & Ramig, 1992; Muniesa et al., 2004). Because seawater preparations are much cleaner and isolates are much less prone to VLP counts being augmented by organic matter and ghost particles, we would suggest that the reduced counts in CsCl purified soil virus communities are also likely legitimate VLP yield reductions. In summary, although CsCl gradients typically do represent “a loss step” they are important for purifying viral particles from contaminating nucleic acids or inhibitory substances for downstream processing (Saano et al., 1995; Zhou, Bruns & Tiedje, 1996), and outperform other purification options such as column chromatography or commercial purification kits (Duffy, O’Brien & Strappe, 2005).