Small-scale distribution of microbes and biogeochemistry in the Great Barrier Reef

Study sites and sampling

Samples were collected at six sites spanning from coast to the outer reef in the Great Barrier Reef (GBR; Fig. 1; Table 1). The Australian Institute of Marine Science (AIMS) provided a permit for this study (G37568.1). Sites 1, 2, 3, and 5 were at coral reefs, site 4 was in the Coral Sea, and site 6 was at the AIMS harbour (Bowling Green Bay; Fig. 1). Site 4 (Coral Sea) was used as a reference non-coral site. Site 6 (Bowling Green Bay) is located in the inner zone of the GBR, has no coral coverage and is dominated by a nearby saltmarsh and small river. All coral reef sites showed clumps of floating Trichodesmium spp. at the surface (Carreira pers. observ.) at the time of sampling. To determine the spatial variability in the small-scale distribution of microbes, viruses and biogeochemical measurements, surface water samples were collected once at sites 1 to 5 (Fig. 1, Table 1; 17 to 22 December 2014). The temporal variability in the small-scale distribution was determined at site 6 with samples collected during high tide every 24 h over four consecutive days (Fig. 1, Table 1; 12 to 15 January 2015). Although 24 h is not a temporal small scale, the objective was only to understand the changes in spatial small scale over time. Furthermore, although it has been shown that there can be differences in bacterial abundances on coral reefs between day and night (Weber & Apprill, 2020), our daily sampling over the 4 days occurred always during daylight between 2 pm and 5.45 pm (Table 1; sunset at 6.55 pm). Niskin bottle samples collected at site 6, at the same time and days as the temporal study (days 1 to 4), were used as controls for the standard sampling method.

Samples were collected with a device built for the purpose of this study consisting of 25 inlets (5 x 5) (Fig. 2). With the help of a lever, all samples were collected manually, at the same time from 0.5 m below the sea surface with the sampling taking about 5 s. As our objective was to understand small-scale heterogeneity in the coral reef system, samples from sites 1, 2, 3, and 5 were taken above the coral reefs, but not in the proximity of a coral as done by Seymour et al. (2005). Each inlet was connected to a 25 mL syringe each separated by seven cm, representing a total sampling area of 784 cm². This distance between the syringes was calculated to account for the volume necessary for all measurements (25 mL) without interfering with neighbouring sampling volumes. In this calculation we assume that the rapid intake of water by the syringes is similar in shape to a turbulent jet (Pope, 2000). The following equations were used for the calculation: (1)${\displaystyle V=\pi \times r^2\times \frac{h}{3}}$ (2)${\displaystyle Tan\theta =\frac{r}{h}}$ (3)${\displaystyle r={\left(\frac{3\times V\times tan\theta }{\pi }\right)}^{\frac{1}{3}}}$ we combined Eqs. (1) and (2) to obtain the minimum distance (r) between syringes (Eq. (3), Fig. S1). Equation (1) calculates the volume of a cone (V), which is the water rapidly sucked up (‘turbulent jet’) by the syringe and Eq. (2) takes into account its shape. This allowed to calculate the distance (r) using a known angle of 11.8° (Eq. (3); Fig. S1) (Pope, 2000; Cushman-Roisin, 2019). This angle is always the same independent of the fluid used (Cushman-Roisin, 2019).

At sites 1 to 5 temperature and salinity were recorded using a conductivity-temperature-depth (CTD) profiles (Seabird SBE19Plus). At site 6 salinity samples were collected and analysed in the laboratory with a Portasal Model 8410A, while temperature was measured manually. Salinity and temperature varied between 32.0 and 35.6, and 28.7 and 32.0 °C, respectively (Table 1). From each syringe, samples were collected for dissolved inorganic nutrients (ammonium - NH₄⁺, nitrate/nitrite - NO₃⁻/NO₂⁻, and phosphate - HPO₄²⁻), dissolved organic carbon (DOC), total dissolved nitrogen (TDN), chlorophyll a (chl a), and bacterial and viral counts. A precombusted (450 °C, 4 h) GF/F filter (13 mm diameter) was used to filter five mL for inorganic nutrients analysis, and 10 mL of seawater for DOC and TDN analysis. DOC and TDN samples were fixed with 50 μL of 25% H₂PO₄ and kept at 4 °C, whereas inorganic nutrients were filtered and kept at −20 °C until analysed. The GF/F filters used for collecting inorganic nutrients, DOC, and TDN samples were snap-frozen in liquid nitrogen and kept at −20 °C for chl a extraction. All syringes, filter-holders and inorganic nutrient sample tubes were acid-washed in 10% HCl for 24 h, and then washed three times with Milli-Q water before use.

For bacterial and viral counts, unfiltered subsamples of one mL, were collected in sterile two mL Eppendorf tubes and fixed with 0.5% glutaraldehyde final concentration (25% EM-grade, Merck) for 15 min at 4 °C, after which samples were snap-frozen in liquid nitrogen and stored at −80  °C until analysis by flow cytometry (FCM).

Samples analysis

Inorganic nutrients (NH₄⁺, NO₃⁻/NO₂⁻ and HPO₄²⁻) were determined by standard segmented flow analysis (SFA) as described in Hansen & Koroleff (1999). As all the determined NH₄⁺ concentrations were below the detection limit of the method (<0.02 µmol L⁻¹) the data is not shown. The detection limit and precisions for the other parameters were: 0.02 µmol L⁻¹ for NO₃⁻/NO₂⁻ and 0.001 µmol L⁻¹ for HPO₄²⁻. Please note that the HPO₄²⁻ concentrations at site 2 were also below the detection limit and therefore the data is not shown. DOC and TDN concentrations were measured using a Shimadzu TOC-L carbon analyser coupled in series with a nitric oxide chemiluminescence detector according to Lønborg et al. (2018). Three to five replicate injections of 150 µL were performed per sample. Concentrations were determined by subtracting a Milli-Q blank and dividing by the slope of a daily standard curve of potassium hydrogen phthalate and glycine. Using the deep ocean reference (Sargasso Sea deep water, 2600 m) we obtained a concentration of 45.6 ± 1.8 µmol L⁻¹ (average ± SD) for DOC and 22.0 ± 1.5 µmol L⁻¹ for TDN. Please note that the TDN measurements for day 2 and 3 of the temporal study are not reported due to problems with the gas supply for the nitric oxide chemiluminescence detector during these specific sample runs. The detection limit for DOC and TDN were 8 µmol L⁻¹ and 0.02 µmol L⁻¹, and the precisions were ± 1 µmol L⁻¹ and ± 0.3 µmol L⁻¹, respectively.

Chl a determinations were made by extracting the GF/F filters in ethanol (96%) for 8 h. Samples were analysed spectrophotometrically according to Strickland & Parsons (1972). The dectection limit and precision for the chl a method were 0.005 µg L⁻¹ and ± 0.05 µg L⁻¹, respectively. Flow cytometric (FCM) enumeration of bacteria and viruses was carried out using a standard bench top Becton-Dickinson FACSVerse FCM, equipped with an air-cooled argon laser (excitation 488 nm, 15 mW power) according to Gasol et al. (1999) and Brussaard (2004) for bacteria and viruses, respectively. Samples were diluted (5-60 times) in TE buffer (Tris 10 mM, EDTA 1 mM, pH 8.0), stained with SYBR Green I (Molecular Probes^®; Invitrogen Inc., Life Technologies™, NY, USA) to a final concentration of 10⁻⁴ of the commercial stock solution. Bacterial samples were incubated at ambient temperature, whereas viral samples were incubated at 80  °C (Brussaard, 2004), both in the dark for 10 min. The trigger was set for green fluorescence and the data was analysed using Flowing Software 2.5.1. (freeware; http://flowingsoftware.btk.fi). The event rate was 300 bacteria s⁻¹ and between 300-800 viruses s⁻¹ to avoid coincidence (Gasol et al., 1999; Brussaard, 2004). We would like to note that recent research (e.g., Forterre et al., 2013) has suggested that viral counts might also include gene transfer agents (GTAs), membrane -derived vesicles (MVs), or even cell debris that might be confused with viruses. However currently there is no method to distinguish between all these particles, therefore, we assumed that the viral counts made by FCM are viruses.

Inorganic nutrients, DOC, TDN, and chl a concentration, and bacterial and viral abundances data were plotted using Surfer 9.0.

Statistical analyses

To measure small-scale heterogeneity within each site/day it was used the coefficient of variation (CV) calculated as the (Standard deviation/Mean) ×100. Values closer to 0% indicate a low variability, whereas values closer to 100% indicate a high variability (Sokal & Rohlf, 1995).

To understand differences in concentrations/abundance between sites/days, boxplots were made using the average values for each site/day. To compare the distributions of the independent samples, the non-parametric Kruskall-Wallis tests were performed for the spatial and temporal data sets because for each variable at least one subgroup (for site or day) failed the normality condition for parametric tests (Agresti, 2007). Multiple comparison tests were also performed to understand which pairs of sites/days had the biggest differences. For these pairwise comparisons two tests were performed: Nemenyi tests with Chi-squared approximation and the Dunn’s tests for multiple comparisons with the Bonferroni adjustment method (Dunn, 1964).

To identify which variables were more linearly correlated, and determine the correspondent magnitude, Pearson correlation coefficients were calculated, considering each site/day and all sites/days combined. Additionally, due to the fact in most cases the data normality was violated, we calculated Spearman correlation (R_s) to measure the strength of a monotonic relationship between paired data.

To understand the relation between the parameters, independent of the site and day, factor analysis was applied to the data. A factor analysis is used to describe an eventual correlation between several observed variables in regard to another group of non-observed variables, of smaller dimension, named factors (Johnson & Wichern, 2007). To perform the factor analysis all variables were considered, regardless of site, as there were no significant correlations between the variables, according to Bartlett’s test of sphericity (p value < 0.001). In order to classify the variables, cluster analyses were performed in the spatial and temporal data. For the temporal data, cluster analysis was tested, but without meaningful results. R (1.1.442) and SPSS (v25) software were used for the statistical analyses.

Small scale variability for each site and day

Using the coefficient of variability (CV) as a measure of heterogeneity, generally, in the spatial and temporal studies, there was a high small-scale heterogeneity (up to 76% for chl a) for chl a, NO₃⁻/NO₂⁻ and HPO₄²⁻ and lower heterogeneity for Dissolved organic carbon (DOC), and bacterial and viral abundances. With the exception of chl a, the chemical variables were more variable, than the biological within each site and day (Tables 2 and 3). Next is given a description of the small-scale variability for each parameters for the sites and days measured.

In the spatial study of NO₃⁻/NO₂⁻, site 5 showed the lowest heterogeneity (10%), while site 1 had the highest (37%; Table 2, Fig. 3). In the temporal study, the highest variability was observed at day 1 (45%) and lowest at day 3 (10%) (Table 3, Fig. 4). Maximum differences observed between two nearby points in the spatial and temporal studies were of 2.6 x and 3.6 x, respectively (Figs. 3 and 4). Also site 1 showed the highest heterogeneity in HPO₄²⁻ concentrations (27%), and sites 4 and 5 the lowest (20%; Table 2, Fig. 3). In the temporal study the heterogeneity of HPO₄²⁻ concentrations was highest at day 4 (26%), and lowest at day 3 (10%; Table 3, Fig. 4). The maximum variability between nearby points was of 2.4 x, both spatially and temporally. DOC concentrations showed the highest heterogeneity at sites 2 and 3 (13%; Table 2, Fig. 3), while the lowest was found at site 5 (6%). DOC showed the lowest heterogeneity of all measured parameters at all sites. The DOC concentrations in the temporal study were higher than in the spatial study, but the heterogeneity was lower. The highest heterogeneity was of 5% at day 3, and the lowest just 4% at all other days (Table 3, Fig. 4). A maximum variability of 1.5 x and 1.2 x between two nearby point was found spatially and temporally. Finally, TDN varied most at site 3 (24%) and least at site 4 (9%; Table 2, Fig. 3). A maximum variability of 3 x was found between points. In the temporal study, TDN was only measured on days 1 and 4, and the heterogeneity was low in those two days measured (9 and 6%).

Chl a showed the highest variability of all parameters, with the highest heterogeneity at site 5 (68%) and the lowest at site 2 (43%; Table 2, Fig. 5). In the temporal study the heterogeneity was higher than found in the spatial study, with a maximum at day 3 (76%). Differences between nearby points were 7.9 x and 25.5 x, spatially and temporally.

Bacterial and viral abundances showed generally similar and low heterogeneity both spatially and temporally, with viral abundances being nearly 1 order of magnitude higher than bacteria. Bacterial abundances showed similarly low heterogeneity at site 2 and day 1 (4% and 5%), and high at site 3 and day 3 (15% and 19%; Tables 2 and 3, Figs. 5 and 6). Viral abundances showed lowest heterogeneity at sites 1 and day 1 (6% and 9%), while highest heterogeneity was found at sites 2 and 3, and day 2 (15% and 13%; Tables 2 and 3, Figs. 5 and 6). Finally, the VBR showed low heterogeneity at sites 4 and 5 (9%) and day 1 (10%), and highest heterogeneity was observed at site 3 (15%) and day 3 (20%; Tables 2 and 3, Figs. 5 and 6). It should be noted that the values measured (except for nitrogen) were one order of magnitude higher than its precision, and the difference between measured values was mostly higher than the precision, thereby allowing to detect real variability between samples.

Overall, although no clear pattern emerged, site 3 (furthest north) and day 3 (high nutrient concentrations) had most parameters with highest heterogeneity, while site 5 (furthest south) and day 1 (low nutrient concentrations) had most parameters with the lowest heterogeneity.

Comparing small-scale variability between sites and days

All sites showed comparable concentrations/abundances overall, with the exception of bacterial and viral abundances at the non-coral site 4 (Coral Sea) that were on average 1.8 x and 2.6 x lower compared to the other sites (Table 2, Fig. 5). Over the four days, nutrient concentrations increased while bacterial and viral abundances decreased, and chl a and VBR showed no differences (Table 3, Fig. 6, Fig. S2). The inorganic nutrients (NO₃⁻/NO₂⁻ and HPO₄²⁻) showed comparable average concentrations between the outer reef and Coral Sea sites (Table 2, Fig. 3), and to site 6 (temporal study; Table 3; Fig. 4), except for day 4 when concentrations increased by 11.0 x and 1.4 x compared to day 1 (lowest concentrations, but still comparable to the sites). DOC was slightly higher at site 6 (average range over the 4 days: 100 ± 4 µmol L⁻¹ to 118 ± 6 µmol L⁻¹; Table 3, Fig. 4), compared to the coral sites (average range over sites 1, 2, 3, and 5: 83 ± 11 µmol L⁻¹ to 90 ± 9 µmol L⁻¹) and Coral Sea site (89 ± 8 µmol L⁻¹; Table 2, Fig. 3). TDN showed slightly higher concentrations at site 6 (particularly day 4) compared to the other sites (Tables 2 and 3, Figs. 3 and 5). Chl a was on average lower at site 6 compared to all other sites (Tables 2 and 3, Figs. 5 and 6). Bacterial and viral abundances were on average higher at site 6 (total average of the 4 days: 18.1 ± 3.1 ×10⁵ mL⁻¹ and 104.8 ± 17.3 ×10⁵ mL⁻¹, respectively; Table 3, Fig. 6), compared to the coral sites (average range: 7.8 ± 0.5 ×10⁵ mL⁻¹ to 12.5 ± 1.9 ×10⁵ mL⁻¹; and 41.3 ± 2.7 ×10⁵ mL⁻¹ to 58.2 ± 8.8 ×10⁵ mL⁻¹ respectively; Table 2, Fig. 5). Overall site 6 (temporal study) showed either similar or slighly higher concentrations/abundances when compared to the other sites (coral sites and non-coral - Coral Sea), but these results should be taken with cautions as there was no temporal follow-up at sites 1 to 5.

Comparing the concentrations and abundances obtained with a Niskin bottles during the temporal study (Table S5) with the range of values for each parameter over the 4 days (Table 3), the values are generally within these ranges, hence also showing the increase in nutrients and decrease in bacterial and viral abundances over the 4 days.

For each variable (NO₃⁻/NO₂⁻, HPO₄²⁻ DOC, TDN, Chl a, and bacterial and viral abundances) pairwaise comparisons between sites and days were performed. The Kolmogorov–Smirnov tests showed that the spatial and temporal data were not normally distributed for each variable. Concerning each variable in the study, the Kruskall-Wallis test revealed that at least one of the samples for each site/day is significantly different from the others. However, the pairwise comparisons using suitable non-parametric tests for multiple comparisons (Nemenyi-tests with Chi-squared approximation and Dunn’s tests) showed statistically significant differences between some sites and days for each variable, but these were ‘random’ and without a clear link between sites/days. The p values are shown in the supplement material (Tables S5 and S6). Overall, statistically significant differences were found, but there were no clear patterns spatially or temporally as determined by non-parametric tests, i.e., no site and day or combinations of sites and days showed a trend or similar behaviour for all the parameters or a subsection of these (Figs. 7 and 8). DOC showed the least statistical differences between sites (Fig. 7, Table S5), while chl a showed the least differences between days (Fig. 8, Table S6).

Pearson and Spearman correlations were determined between all parameters within a site/day and between sites/days without any clear results (Tables S1–S4). Most correlations did not exhibit a strong linear relationship or even a monotonic relationship between the variables in study (Tables S1–S4). However it can be highlighted that spatially, bacterial abundances correlated negatively with HPO₄²⁻ and positively with viruses (n = 25, R² = −0.54 and 0.55, respectively), and temporally bacterial abundances correlated negatively with NO₃⁻/NO₂⁻ and HPO₄²⁻(n = 25, R² = − 0.75 and −0.50). The correlation between all bacterial and all viral abundances (spatial and temporal) showed a positive correlation (n = 325, R² = 0.75; Fig. 9). This correlation showed an intercept not significantly different from zero, indicating a tight link between bacterial and viral abundances.

Although no relations were found between the parameters at the different sites and days, factor analysis was applied to understand the relation between the parameters. The factor analysis showed that the variables can be decomposed into two factors, the chemical (NO₃⁻/NO₂⁻ and DOC) and biological (chl a and bacterial and viral abundances) groups (Fig. S3). It should be noted that HPO₄²⁻ andTDN were excluded from this analysis because the correspondent anti-image matrices values were smaller than 0.5 (0.340 and 0.372, respectively) meaning that we discarded these variables since its partial correlation values are considered too small to apply factorial analysis. Overall these results show: (1) that the chemical variables (NO₃⁻/NO₂⁻ and DOC) are more related to each others than to the biological variables (chl a and bacterial and viral abundances), and likewise for the biological variables, (2) the chemical variables do not explain the bacterial abundances, and (3) given the biological variables are more related to each others, there is a higher likelihood they could explain the bacterial abundances, but the results are insufficient to make a firm conclusion. Cluster analysis showed (Fig. S4A) a clear grouping between the biological (chl a, and bacterial and viral abundances) and the chemical (NO₃⁻/NO₂⁻, HPO₄²⁻, DOC and TDN) between all sites, whereas the classification was less clear between the days (Fig. S4 B). Here bacterial and viral abundances grouped together, while DOC, chl a, NO₃⁻/NO₂⁻ and HPO₄²⁻) clustered. Please note that the TDN data was not included in this analysis as there was no data for days 2 and 3.

The dataset generated for this study can be found at https://figshare.com/articles/Figshare_Carreira_2020Feb_xlsx/11841168.