Corrigendum: Archaeal amoA and ureC genes and their transcriptional activity in the Arctic Ocean

Thaumarchaeota and the gene encoding for a subunit of ammonia monooxygenase (amoA) are ubiquitous in Polar Seas, and some Thaumarchaeota also have a gene coding for ureC, diagnostic for urease. Using quantitative PCR we investigated the occurrence of genes and transcripts of ureC and amoA in Arctic samples from winter, spring and summer. AmoA genes, ureC genes and amoA transcripts were always present, but ureC transcripts were rarely detected. Over a 48 h light manipulation experiment amoA transcripts persisted under light and dark conditions, but not ureC transcripts. In addition, maxima for amoA transcript were nearer the surface compared to amoA genes. Clone libraries using DNA template recovered shallow and deep amoA clades but only the shallow clade was recovered from cDNA (from RNA). These results imply environmental control of amoA expression with direct or indirect light effects, and rare ureC expression despite its widespread occurrence in the Arctic Ocean.

photosynthetic rates continuing throughout the Arctic summer, despite low standing stocks of inorganic nitrogen in the euphotic zone 21,22 . Galand and collegues 23 found that MGI and archaeal amoA genes were abundant in Northern Baffin Bay, particularly in the Arctic halocline, leading to the hypothesis that the archaeal community could influence photosynthetic community structure by rapidly recycling inorganic nitrogen in the euphotic zone. Nitrate concentrations in the surface waters of Baffin Bay are often below the detection limits (ca. 0.03 mmol N L 21 ) 24,25 , but ammonium concentrations can be substantial 26 , suggesting bacterial degradation of fixed organic material and possible release of ammonium. The high affinity of MGI Archaea for ammonium 27 could enable MGI to successfully compete with phytoplankton for ammonium, and result in amoA expression nearer the surface. In contrast to NH 4 , there have been few measures of urea in the Arctic, but in general, concentrations are variable and sporadic 28,29 , which leads to the hypothesis that expression of genes coding for urease would also be variable. The presence of archaeal ureC genes in Arctic MGI Archaea is intriguing 18 but to date there is no evidence of ureC being expressed in polar waters.
The main objective of our study was to quantify the amoA and ureC transcripts in three regions of the Arctic Ocean (Amundsen Gulf, Lancaster Sound and Baffin Bay) and to gain an understanding of conditions associated with gene expression. To this end, we investigated amoA and ureC transcripts by qPCR during polar darkness and around the vernal equinox. We then experimentally investigated amoA and ureC expression in natural communities exposed to typical irradiances of the Arctic Ocean euphotic zone. Finally we investigated the depth distribution of genes and transcripts from MGI Thaumarchaeota, archaeal amoA and ureC, along a north to south gradient of the hydrographically complex center of the Northern Baffin Bay 30 , targeting water masses with different bio-physical properties. Phylogenetic differences between the more likely active and more likely dormant Thaumarchaeota with amoA were also examined through cloning and sequencing the amoA gene from both DNA and cDNA (from total RNA).

Results
Effect of light on ureC and amoA transcripts. In the Amundsen Gulf region (Western Canadian Arctic), amoA, transcripts, ureC transcripts and MGI 16S rRNA were quantified on 3 dates as ambient light conditions increased over time (January, March and the beginning of April). In surface waters, amoA and MGI copies mL 21 were greatest in January. At the halocline, more transcripts were recorded in April. There were always more transcripts at the halocline than at surface. UreC transcripts were only detected in January ( Figure 1).
In the initial samples from Lancaster Sound (Eastern Canadian Arctic) used for the light-manipulation experiment, amoA transcripts were detected at T 0 from both 1% (80 m deep) and 25% (25 m shallow) surface irradiance ( Figure 2), with an order of magnitude (ca. 10 5 versus 10 4 ) fewer transcripts near the surface. UreC transcripts were successfully quantified from the freshly collected deeper sample (T 0 ), with 10 copies per mL of seawater, but not in the sample from 25% irradiance. There were no significant changes in MGI 16S rRNA copies mL 21 over the 48 hours, and no effect of light (3 way ANOVA) on MGI 16S rRNA ( Figure 2). There were, however, significantly more copies of MG1 16S rRNA in the deeper sample compared to the 25 m water over the 48 h incubation (P , 0.001).
The amoA transcripts were significantly greater in the dark than in the light overall, but the effect of collection depth was not significant (see below). The multiple pairwise comparisons revealed several interesting trends. After 24 hours, amoA transcripts were always greater in the dark (ANOVA-pairwise multiple comparison, P 5 0.008) compared to bottles exposed to light. After 48 h, the amoA transcripts from the two depths diverged, with a significant increase under both light and dark conditions in the 25 m origin treatments (P , 0.001), conversely in the 80 m origin treatments, amoA transcripts decreased significantly (P , 0.001) ( Figure 2). Transcripts of ureC were not detected in the experimental treatments.   (Supplementary Table S1). Dissolved oxygen concentrations ranged from 279 to 356 mmol kg 21 (Figure 3). At Stn123, there was a low oxygen intrusion between 200 and 300 m ( Figure 3) coinciding with slight increases in salinity and temperature (not shown). Bacteria concentrations were ca. 10 5 cells mL 21 and tended to decrease with depth, with slightly higher concentrations above the pycnocline (Supplementary Table S1). Nitrite concentrations ranged from 0.05 to 0.17 mmol kg 21 , nitrate concentrations ranged from below the detection to 15.34 mmol kg 21 (Supplemenary Table S1). The nitrite maxima were above nitrate maxima and between 100-200 m at the three stations (Fig. 4) Transcript and gene quantification in the water column. MGI 16S rRNA (from cDNA) qPCR values ranged from ca. 175 to 4.5 3 10 5 copies mL 21 of seawater collected, and amoA transcripts from 18 to 1829 copies mL 21 ( Supplementary Fig. S1). In contrast to amoA, where transcripts were detected in all samples tested, ureC transcripts were never detected from the Baffin Bay samples.
Baffin Bay amoA transcripts varied down the water column and between stations. The station (Stn) 129 maximum for amoA transcripts was at 35 m, which was above the nitrite maximum ( Figure 4; note the log scale for depth). For Stn 123, the maximum amoA transcripts were just below the nitrite maximum. Finally for Stn 109, the amoA transcript maximum was at 90 m, which was above the nitrite maximum (Fig. 4).
In Baffin Bay, gene abundance (from DNA) also varied with depth and station. Stn 129 maxima for amoA, ureC and MGI 16S rRNA genes were detected in the 120 m sample (Figure 5a). At Stn. 123, the maxima for MGI, amoA and ureC genes were at 280 m and coincided with the nitrate maximum at the base of a low oxygen intrusion (Figs. 3 and 5a). At station 109, gene copies of MGI, amoA and ureC were also all greatest nearer the nitrate maximum at 195 m (Fig. 5a).
Baffin Bay qPCR values from DNA template ranged from 16 to 3756 copies of MGI (16S rRNA) genes mL 21 , 24 to 5270 amoA gene copies mL 21 , and 45 to 2780 ureC gene copies mL 21 . Overall amoA gene copy numbers were significantly correlated with MGI rRNA genes (Tau, T 5 0.74, p , 0.001) as were ureC gene copies to MGI rRNA genes (T 5 0.66, p , 0.001) (Fig. 5b). UreC gene copy numbers also correlated with amoA gene copies (T 5 0.79, p , 0.001), and genes followed depth related variables, such as silica, phosphate and nitrate ( Supplementary Fig. S2). The ratio of amoA genes to MGI 16S rRNA genes was on average 1.8 ranging from 0.2 to 5.5. There were fewer ureC gene copies than amoA copies with an average ratio of 0.4 ureC genes to MGI 16S rRNA genes ( Figure 5b).
The maximum abundance of amoA transcripts did not correspond to the same depths as the maximum gene copies in any of the water column profiles. Furthermore there was no significant correlation between archaeal amoA transcripts and gene copy numbers estimated from the DNA template, nor between MG1 16S rRNA genes and 16S rRNA copies from cDNA.
Across stations and depths from both sides of the Canadian Arctic we found a significant positive relationship between MG1 16S rRNA and amoA transcripts (analysis on log transformed data with r 2 5 0.80 and p , 0.001) ( Supplementary Fig. S1). The PCA indicated that nutrients, including silica and phosphate, which were associated with the deepest samples, best explained gene concentrations along the X axis. Chl a and ammonium were more associated with amoA transcripts and MGI 16S rRNA ( Supplementary Fig. S2).
Diversity and phylogeny of amoA in Baffin Bay. A total of 423 cloned amoA gene sequences with the correct target size (635 nt) from 8 libraries passed chimera and other quality tests. The two libraries constructed using cDNA as a template were asymptotic. There were five OTUs at the 97% similarilarity level out of 32 clones at Stn 123-200 m and a single OTU from 46 clones at station 109-110 m ( Table 2). The DNA libraries were also asymptotic. At 97% similarity, there were 4 OTUs out of 19 clones from Station 109-110 m. Only 6 clones were recovered from the shallowest library (station 129-35 m) and each clone was unique. At the two stations and depths where both DNA and cDNA were used as templates, the DNA derived communities were more diverse ( Table 2). All closest matches were to uncultured environmental archaeal amoA genes (Fig. 6).
Maximum likelihood (not shown) and Neighbor joining phylogenies ( Figure 6) placed all of the amoA sequences into previously defined marine groups, corresponding to clades A and B defined by Francis et al. 31 , with no sequences within their group C. All amoA sequences from RNA source libraries grouped within clade A, which is thought to represent shallow ecotypes 32 . The most common cDNA sourced amoA sequences grouped together in a subclade that we annotated A.A, which corresponds to the group A of Kalanetra et al. 33 (Figure 6). Our sub-clade A.A sequences were . 99% similar to other polar sequences (not shown). At Stn 123-200 m, amoA clones from two clusters (indicated as A.B. and A.A.a in Figure 6) were most often from cDNA (representing . 50% of cDNA clones). At Stn. 109-110 m, only A.A.a was retrieved from cDNA. A third Arctic sub-cluster in the A clade (A.C. in Figure 6) had no cDNA representatives. DNA sequences fell into both A and B clades 32 . Overall 40% of the sequences from the DNA libraries fell into the B clade.
There was some amoA clone clustering according to depth. The shallow A.A.a group was, for example, was absent from the 400 m sample and one deep B sub-clade was overrepresented in the deepest samples ( Figure 6).

Discussion
Taken together, the lines of evidence suggest that amoA expression was influenced by light but amoA containing MGI persisted in the upper waters of the Arctic Ocean, and were possibly more active when irradiance (PAR) was lower. Our qPCR results indicated that amoA transcripts were more abundant in surface waters during January, the darkest winter month, and later more abundant in the halocline where light levels were lower. In the light manipulation experiment amoA transcript levels were significantly higher in the dark. We found that actual irradiance levels at the time of sample collection likely influenced transcript abundance, with more transcripts detected in samples from the euphotic zone (defined as .1% of surface PAR), when surface irradiance was lower. Light may inhibit amoA activity directly, promote more rapid degradation of transcripts, or have secondary influence on substrate availability, for example phytoplankton may be better able to compete with ammonium oxidizing Archaea for ammonium when light levels are greater. Similarly, in the Pacific, Church et al. 34 reported peak amoA transcript abundance at the bottom of the euphotic zone.
Our results indicated that amoA transcriptional activity probably occurs throughout winter. Compared to other oceanic regions, Arctic and Antarctic MGI 16S rRNA genes are more abundant nearer the surface, especially in winter and early spring 35 , with values lower in summer 36 . Collins et al. 37 also reported MGI 16S rRNA genes in newly formed sea ice in the Beaufort Sea in October when light levels are rapidly decreasing. The lower expression that we observed nearer the surface compared to the halocline in March and April could be due to direct light inhibition or lack of substrate. The net effect was a decline in amoA near surface activity in spring when phytoplankton activity begins 38,39 .
In Baffin Bay, distribution profiles of the amoA transcripts down the water column were closely correlated with the MGI 16S rRNA indicating that the MGI were in a state consistent with being metabolically active and with ammonium oxidation having occurred around the time of sampling. This was notable since the transcript distribution differed from the gene distribution (from DNA) ( Figure 3); suggesting that environmental gene distribution may not accurately reflect activity. In the Pacific, Church et al. 34 reported similar results with relatively more amoA transcripts nearer the surface, despite higher gene numbers at depth. The relationship between the euphotic zone and nitrification activity was reported in a North-South transect of the Atlantic Ocean, ammonia oxidizing activity was greatest at the base of the euphotic zone, and decreased with depth 40 . But this depth dependent distribution, did not occur in the Black Sea and the highest expression of amoA is within a nitrification zone coinciding with low oxygen levels 41 . At Stn 123, where we sampled a lower oxygen layer, amoA transcripts were high compared to other stations sampled at the same depth, but higher still at the base of the euphotic zone. While we did not directly measure nitrification, we speculate that in the Arctic amoA gene maximum activity differed from DNA gene distribution due to substrate availability over relevant time spans, with amoA transcripts associated with Chl a and ammonium concentrations in the PCA analysis ( Supplementary Fig.  S2). In Baffin Bay, dissolved organic matter (DOM) is released by  phytoplankton leakage 21 , and zooplankton and protist grazing 42 . This DOM is quickly broken down by bacterial activity 43 . The higher probable microbial activity at the base of the euphotic zone stands out as the overiding factor associated amoA transcript abundance. In contrast, the surface MGI populations, are most likely maintained by diurnal activity during periods of low incoming solar radiation each day.
In early September, the greatest concentrations of MGI 16S rRNA and amoA genes in Baffin Bay were at the nitrate maximum, similarly in the Pacific, the same two genes tracked nitrate concentrations down the water column 11 . However, our PCA showed that phosphorus and silica, with higher concentrations in deeper waters, had a stronger relationship with gene copy numbers in Northern Baffin Bay, consistent with the oceanography of the region 21 . The abundance of genes, but not transcripts, below the subsurface chlorophyll maximum layer may be linked to loss rates. For example, protists that graze on prokaryotes are also more abundant above the nitracline 44 . Alternatively, the lower numbers of amoA transcripts per MGI 16S RNA gene deeper in the water column could also be due to inherent biological properties of dominant clades in deeper waters 45 .
For ureC, we detected the gene at all depths and stations, and in a parallel study, the gene was reported in Amundsen Gulf in March and February 18 , indicating that MGI with the ureC gene are widely distributed in the Arctic, irrespective of season or depth. In contrast to amoA transcripts, which were detected in all samples, we only recorded ureC transcripts in samples from January, when photosyn-    thetic primary production was not possible because there is no light, and from a sample, collected from the bottom of the euphotic zone in Lancaster Sound, which is a region of high biological activity 25 . The phylogeny of the amoA gene implies the presence of at least two distinct MGI ecotypes. All amoA transcripts grouped within the clade A 31 , which is classed as a shallow group 32 . More specifically, in our clone libraries the most abundant cDNA amoA clones were closely related to each other and within a single sub-clade (A.A). This sub-clade was originally identified by Kalanetra et al 33 from the central Arctic Ocean, with representatives from the Southern Ocean, and may be typical for cold waters. In contrast, amoA sequences from cluster B, which is thought to be more representative of the deep ocean 32 , were never recovered from cDNA clone libraries suggesting this clade was dormant in Baffin Bay at the time of sampling. DNA based and cDNA based clone libraries from the central California Current recovered both A and B clusters and the A cluster dominated in all but the deepest samples 9 . Recently Sintes et al. 45 explored amoA gene diversity using amoA primers designed to detect two distinct clades of the AOA; one associated mostly with less deep waters and high ammonium concentrations (HAC) and the second clade, occurring in deeper waters and with low ammonium concentrations (LAC). Both primer pairs detected high concentrations of amoA HAC and LAC in an Arctic coastal site with co-occurrence down the water column.
The co-occurrence of the two AOA clusters in the Arctic Ocean, even near the surface, is intriguing. Inspecting gene ratios from our qPCR results we note the potential of 2 copies of the amoA gene within ca. 60% of the MGI (with a 1.8 to 1 ratio of amoA genes to 16S rRNA genes), but fewer MGI containing the ureC gene (with a ratio of 0.4). In effect, the persistence of a dormant B clade at similar occurrence rates (40% of our DNA template clones) and the observation of a ureC gene to MGI 16S rRNA gene ratio of 0.4 may be linked.
Apart from its initial discovery in the coral reef endosymbiont 46 , the archaeal ureC gene has been most often reported from deeper waters or winter surface waters where concentrations of fresh organic material would seem to be low 47 . The sporadic amplification of ureC transcripts over a range of samples from the Arctic, is telling. As mentioned earlier, there appeared to be a light effect, but the rarity and weak concentrations of ureC transcripts may also be a result of the ephemeral occurrence of urea itself in the Arctic environment 29 . Zooplankton are the main source of urea in the Arctic, with some copepod species persisting and feeding near the surface and at middepths throughout the Arctic winter 48 . Interestingly, one copepod species, Acartia tonsa, releases different forms of nitrogen depending on diet; with a mixed diet of heterotrophic protists and diatoms resulting in greater release of urea compared to either dietary source alone 49 . Copepod species in the Arctic are mostly opportunists and able to switch to the dominant food source 50 and thus the taxonomy of the protist plankton prey could also result in the variable urea concentrations. The episodic but poorly understood input of urea into the surrounding waters could be sufficient for maintaining Thaumarchaeota with the capacity to scavenge and use urea as a carbon and energy source. While the archaeal ureC gene has been reported in the Arctic and elsewhere 18 , transcriptional activity to date has only been reported in very deep waters 14 and a coral symbiont 4 . The similarity between the two types of environments is not obvious, except possibly the availability of animal generated urea.
Overall we found evidence of two MGI populations, one active nearer the surface, and a second, deeper more dormant population, which may be reliant on sporadic inputs of urea as a source of energy and carbon. Finally, while amoA transcript abundance was greater in the dark, light inhibition was not complete and transcripts persisted under low irradiance. The distribution of AOA in the Arctic suggests multiple links between biological and physical properties in this region.

Methods
Sampling. Samples were collected from Amundsen Gulf during the Circumpolar Flaw Lead System Study (CFL) 51 (Table 1). Samples for both missions were collected from the research icebreaker CCGS Amundsen. Water from the Amundsen Gulf was sampled from 2 depths: surface (ca. 10 m) and at the halocline that is indicative of Pacific origin waters 39 (ca. 50 m) in January, March and April. The three Northern Baffin Bay stations, 109, 123 and 129, were sampled along a south-north transect. Since the cruise was multidisciplinary, with constrained logistics, samples were taken opportunistically over the 24 h period ( Table 1). All seawater samples were collected with 12-L Niskin type bottles (Ocean Test Equipment) mounted on carousel rosette system equipped with a SBE-911 plus conductivity-temperature-depth (CTD) profiler with a dissolved oxygen sensor. Other instruments on the rosette system recorded fluorescence (Seapoint), transmissivity (WetLabs C-Star Transmissometer), photosynthetically active radiation (PAR; Biospherical Instruments), and relative nitrate concentrations from an in situ ultraviolet spectrometer (ISUS) probe (Satlantic). Water sample depths were selected on the downward cast 52 . In particular, we followed the nitrate profile and targeted peak and minimum concentrations indicated from the ISUS trace. Eight depths at each station were selected for DNA and four to six for RNA analysis. DNA and RNA samples were filtered separately, both using a peristaltic pump system. Both filtrations included an in-line 3 mm pre-filtration through a 47 mm polycarbonate filter. The entire system used to filter RNA was treated with RNaseZapH prior to filtration. One to seven L were filtered for RNA at 4uC in the dark, with a maximum filtration time of 3 hours. Samples for RNA were collected on 47 mm diameter 0.2 mm pore size polycarbonate (PC) filters, which were placed in 1.8 mL cryovials with RLT buffer (QIAGEN RNeasy) mixed with 1% beta-mercaptoethanol. The vials were immediately flash frozen in liquid nitrogen and stored at 280uC. For DNA, six L of water were filtered through 0.2 mm Sterivex filter units (Millipore) and stored in buffer at 280uC 23 .
Nutrient concentrations: nitrite (NO 2   2 ), nitrate (NO 3   2 ), silicic acid (Si(OH) 4 ), and soluble reactive phosphate (SRP), were determined on board with a Bran and Luebbe Autoanalyzer III using standard protocols 53 , and detection limit of 0.03 mmol L 21 for the three nutrients. Ammonium was determined manually with the sensitive fluorometric method of Holmes et al. 54 and a detection limit of 0.01 mmol L 21 . Oxygen values from the CTD were calibrated at the Institut National Researche Scientifique in Québec, Canada, using samples taken at discrete depths. Samples for total chlorophyll a (Chl a) were collected onto Whatman GF/F filters. These were stored at 280uC until extraction in 95% ethanol 55 and concentrations determined by spectrofluorometry. Flow cytometry (FCM) samples were fixed with paraformaldehyde, flash frozen in liquid nitrogen and stored at 280C 56,57 . This method does not distinguish Bacteria and Archaea and we use the term bacteria for convenience.
Light manipulation experiment. Samples from two photic depths, 25% and 1% of surface PAR, were collected from near Lancaster Sound, where Arctic water enters Northern Baffin Bay (Stn 301, Table 1). The water from 25 m (25% PAR) and 80 m (1% PAR) was collected into 18 12L PC bottles, and incubated at 4uC, at either in the dark or at 0.60 mmol photons m 22 s 21 in bottles shaded with blue filters, which is typical for September at the 25% photic depth (25% of surface PAR) in the study region. One bottle from each depth was immediately filtered for RNA, as above. After incubations of 24 and 48 h, two replicate bottles for each treatment were filtered as above.
Molecular biological techniques. DNA was extracted using a salt extraction method 58 as modified by Harding et al. 59 and RNA extracted as in Church et al. 60 . RNA extracts were reverse transcribed and PCR amplified (RT-PCR) by using a high capacity cDNA archive kit (Applied Biosystems) as specified by the manufacturer.
Quantitative PCR was performed on a Dyad Disciple thermal cycler with Chromo 4 Real-Time Detector (Bio-Rad) in 96 well opaque qPCR plates with adhesive seals (Bio-Rad). Primers and qPCR conditions, as well as amoA, ureC and MGI positive controls and standard curves were the same as in Alonso-Saez et al. 18 . Primer specificity was verified by running PCR products on gels, and sequencing the resulting single bands from the individual primer sets. The triplicate standard curves and negative controls (without DNA added) were subjected to qPCR along with the samples. Copy number per mL of seawater was calculated as described in Zhu et al. 61 .
Eight clone libraries targeting the archaeal amoA gene and transcripts were constructed using Arch-amoAF and Arch-amoAR primers 31 . Three separate PCR reactions containing 2, 3 or 4 mL of template were amplified to counter primer competition and increase the probability of recovering common gene variants 62,63 . The three PCR products were pooled, and then purified with Qiaquick PCR Purification Kit (Qiagen), and cloned using the TA cloning kit (Invitrogen) according to the manufacturer's instructions. Inserts were amplified with the vectors' M13 primers and resulting amplicons were analyzed by restriction fragment length polymorphism (RFLP) following digestion with RsaI (New England Biolab). Up to 10 clones per pattern were sequenced to cover potential RFLP variants. Sequencing (single read) used the vector's T7p universal primer and was performed at Centre Hospitalier de l'Université Laval (CHUL), Québec, with an ABI 3730xl system (Applied Biosystems).
The sequences were compared to the NCBI GenBank nr/nt database using NCBI BLAST 64 . Non-redundant nucleotide and translated amino acid sequences are www.nature.com/scientificreports SCIENTIFIC REPORTS | 4 : 4661 | DOI: 10.1038/srep04661 deposited under accession numbers HQ713534-HQ713542 for the cDNA clones and JF272615-JF272702 for the DNA clones.
Phylogenetic and statistical analysis. In addition to the 202 archaeal amoA gene sequences from this study, 148 reference sequences were selected from Francis et al. 31 and the resulting 350 sequences were aligned using Clustal X v.1.83, then manually checked and trimmed using GeneDoc, v. 2.6.0.2. After trimming, additional sequences including nearest matches to our sequences, were added for phylogenetic tree construction using the DNAML program from PHYLIP (available at http:// evolution.genetics.washington.edu/phylip.html). The topology of dendrograms using maximum likelihood (ML) and neighbor-joining (NJ) were similar. This larger tree placed all of our sequences fell into Marine Group I (MGI), clades A and B defined in Francis et al. 31 . Non-marine and redundant sequences from our study were then removed from the tree leaving only representative A and B sequences in the final tree where evolutionary history was inferred by NJ, with evolutionary distances computed using the Maximum Composite Likelihood method. Phylogenetic analyses were conducted in MEGA4 65 .
To estimate library coverage using Chao1, sequences were clustered at a similarity level $97% using DOTUR 66 . For the light experiment, significance of the 3 way (depth of collection, light level and time of incubation) ANOVA was tested on log transformed qPCR data following a general linear model after testing for main effects. Pairwise multiple comparisons using the Holm-Sidak method were used to test the significance of interactions.
Kendall's correlation analysis was used to test correlations among all quantitative variables from Northern Baffin Bay. A principal components analysis (PCA) was conducted using environmental and qPCR data, and plotted in a coordinate system, given by the two main components, to explore possible relationships between variables. Missing data were handled by pairwise deletion. PCA was done with PAST 67 . ANOVA, Kendall's correlation and regression analysis on log transformed qPCR data were carried using SigmaPlot version 11 (Systat Software Inc).