Introduction

Benthic sediments provide much of the phosphorus (P) attenuation observed in streams (Haggard and Sharpley 2007; Hamilton 2012). Sediments contain inorganic particles of varying sizes and geochemical characteristics derived from the parent material, which determine the sites available for P sorption (House 2003). Sediments also contain allochthonous and autochthonous stocks of organic matter (OM; Tank et al. 2010; Kaplan and Cory 2016), providing fuel for the metabolism of sediment microbial biofilms (Battin et al. 2016). These characteristics make the sediment matrix a hotspot of biogeochemical cycling for P as well as for carbon (C) and nitrogen (N). Hence, when P is removed from the water column at baseflow, it is difficult to pinpoint which biotic and abiotic mechanisms are responsible. Such mechanisms, when known, could inform better nutrient pollution modelling (Macintosh et al. 2018) and subsequent strategies to mitigate the effects of P enrichment (Meals et al. 2010; McDowell et al. 2018; Drohan et al. 2019).

The range of abiotic and biotic mechanisms that remove dissolved reactive P (DRP) from solution and into sediments is diverse. For example, reactive surfaces on particles, such as highly sorptive hydrous metal oxides (Dzombak and Morel 1987; Sposito 2004), often constitute much of the abiotic P sorption capacity and P storage in soils and sediments (Small et al. 2016; Audette et al. 2018; Herndon et al. 2019). Meanwhile, microorganisms entrained in the sediment matrix also require P for their growth and likely supply much of the biotic P uptake for stream sediments (Sinsabaugh et al. 2009; Van Horn et al. 2011; Hill et al. 2012). This biotic P uptake is dependent on stoichiometric demand (Cross et al. 2005; Maranger et al. 2018). While microorganisms beneath the benthic zone (primarily heterotrophs due to lack of light; Battin et al. 2016) can vary in stoichiometric P requirements (Cross et al. 2005; Scott et al. 2012), they approximate a molar C:N:P ratio of 60:7:1 (Cleveland and Liptzin 2007; Hill et al. 2010; Sinsabaugh et al. 2012). This ratio is P-rich relative to that of other lotic biota (Cross et al. 2005) and indicates the potential for sediment microorganisms to moderate stream P cycling if sufficient C and N are available. The distinction between abiotic and biotic sediment P removal is important, however, as not only are their rates and capacities different, they also differ in their responses to other processes within the stream (e.g., varying C and N resource supplies or changes in pH).

Sediments can strongly buffer solution DRP towards a concentration termed the equilibrium phosphate concentration at net zero sorption (EPC0; Froelich 1988). The EPC0 is measured from a series of batch sediment incubations with solutions of varying initial phosphate concentrations; the concentration where neither net removal nor release of P by the sediment occurs (i.e., the x-intercept from a P sorption plot) is the EPC0 (Taylor and Kunishi 1971; Froelich 1988). As such, sediment EPC0 relates to the potential of a sediment to buffer variable DRP concentrations in the water column. For example, Ekka et al. (2006) showed increases of several mg P L−1 in sediment EPC0 downstream from point source inputs of DRP relative to sediments immediately upstream, showing how the EPC0 is elevated with greater loadings of P. In large stream survey studies, McDaniel et al. (2009) and McDowell (2015) found that EPC0 not only related to P loading (e.g., relatively greater when influenced by point sources or agricultural land use) but also correlated with particle size and geochemical characteristics. That is, given a similar P loading, more (chemically) sorptive sediments will have a lower EPC0 and thus a greater potential to buffer in-stream DRP concentrations.

EPC0 does not differentiate between abiotic or biotic mechanisms that control sediment P flux. The sediments used for EPC0 likely harbor both abiotic and biotic P exchange processes. So, a recurring question for stream P cycling has been: What is the relative contribution of abiotic and biotic processes towards the EPC0? By sterilizing the sediments via varying methods, EPC0 has been observed to increase dramatically or by very little, with similarly inconsistent changes in P sorption capacity (Meyer 1979; Klotz 1985; Munn and Meyer 1990; Haggard et al. 1999; Lottig and Stanley 2007; McDaniel et al. 2009; Griffiths and Johnson 2018). Unfortunately, such comparisons can suffer from the choice of sterilization method, as several methods cause severe physicochemical changes to the sediment. For example, autoclaving can disrupt soil OM and surface chemistry (Trevors 1996; Buessecker et al. 2019). Measurements of abiotic P sorption will not be representative unless sediment physicochemistry is undisturbed. Further, few EPC0 comparisons have included corrections for cell lysis following sterilization, where microbial biomass P can release into solution and may adsorb onto the sediment surface (e.g., Klotz 1985; McDaniel et al. 2009).

This study seeks to distinguish biotic and abiotic contributions to stream benthic sediment EPC0 while minimally disrupting sediment physicochemistry. We account for P in the microbial biomass and determine if removing C- and N-limitation on biotic P demand affects the EPC0. Our two hypotheses were that (1) abiotic processes determine EPC0 and so EPC0 would be similar for fresh and sterilized sediments if microbial biomass P were negligible and that (2) microbial P demand, and so the biotic contribution to sediment EPC0, would be most pronounced under P-limitation versus C- or N-limitation. For the first experiment, we measured EPC0 in sediments both fresh and following the use of γ-irradiation to sterilize sediments. For the second experiment, we measured sediment EPC0 with and without addition of C and N to remove microbial C- and/or N-limitation.

Methods

Sites and sampling

We sampled 12 streams across Banks Peninsula, New Zealand (Fig. 1), thus targeting streams draining a consistent geology (volcanic-basic; Snelder and Biggs 2002) but a variety of land uses, and therefore, variable C, N, and P inputs (Supplementary Table S1). The streams were 2nd to 4th order, with catchment areas ranging from 3.7 to 51 km2. Benthic substrates were generally gravel to gravel/cobble. Catchment land uses for 2018, derived from the New Zealand Land Cover database (v5.0; https://lris.scinfo.org.nz/), varied from predominantly native forest and shrub (83%) to predominantly intensive grassland (82%).

Fig. 1
figure 1

Study streams on Banks Peninsula, New Zealand. Detailed site information is given in Table S1. Land uses for 2018 were simplified from the LCDB v5.0 database

Full size image

Our sampling was designed to accommodate an immediate delivery of sediments to be γ-irradiated with minimal storage time. Hence, we sampled 7 sites one day prior to shipping and five sites on the day of shipping samples for γ-irradiation.

All streams were sampled at summer baseflow conditions (November 2019) with no recent disturbances. Benthic sediments were collected in actively flowing zones, primarily in riffles (i.e., avoiding depositional pools and lateral storage areas). To collect enough mass and to capture the spatial variability in sediments across the channel, we sampled ~ 10 to 30 m of stream length by sampling in a zig-zag pattern. We collected the top 1–5 cm of substrate with a shovel, wet-sieved the material (< 2 mm), decanted excess water after settling, and kept the composite, fine sediment sample (approximately 0.5 to 1 kg per site). A water sample was collected from the thalweg and filtered (0.45 µm) into vials with minimal headspace. Both sediment and water samples were kept on ice and in the dark until they were refrigerated (4 °C) in the laboratory and a set of subsamples for stream water were frozen (− 20 °C). Additionally, we used a HACH HQ40D meter to measure stream temperature, dissolved oxygen (via an optical probe), pH, and specific conductivity in situ; all probes were calibrated on the day of sampling according to manufacturer instructions.

Sterilization via γ-irradiation

To control for potential microbial P uptake, we sterilized sediment subsamples via γ-irradiation, as this method non-invasively kills or damages cells with minimal effects on sediment physicochemistry (McLaren 1969; Trevors 1996; Buessecker et al. 2019). Subsamples from all 12 sediments were γ-irradiated with a nominal 25 kGy dose from a 60Co source (MSD Animal Health, Upper Hutt, New Zealand). The actual dose delivered was later confirmed to be 26 kGy (Dave Harris, MSD Animal Health, pers. comm.). This dose is comparable to that used in previous studies and represents a favorable tradeoff between potential physicochemical disruption of the sediment and effective sterilization (Meyer 1979; Östlund et al. 1989; Qiu and McComb 1995; McNamara et al. 2003; Buessecker et al. 2019).

We managed the logistics of sampling, sterilization, and analyses so that (1) sediments were most reflective of in situ conditions for EPC0, (2) sterilized sediments had minimal time for microbes to re-proliferate, and (3) both fresh and sterilized sediments were held under similar conditions throughout prior to analyses. Sediment samples were immediately shipped to be γ-irradiated following collection. Samples were kept cool throughout all handling via freeze-packs (with replacement as needed), thus maintaining temperature comparable to that of the fresh sediments (kept refrigerated in the laboratory at 4 °C). Once γ-irradiated, sterilized sediments were returned to the laboratory within 24 h (4 days since sampling). Using aseptic handling for the sterilized sediments, analyses for dehydrogenase activity (DHA; both γ-irradiated and fresh sediments) and for EPC0 (γ-irradiated) began immediately (see below). All batch incubations for γ-irradiated sediment EPC0 were completed within 48 h of sterilization.

Water analyses

Dissolved reactive P was measured on filtered stream samples within 24 h. We used the molybdenum-blue method (Murphy and Riley 1962) with a 5 cm quartz cell (method detection limit of ~ 2 µg P L−1). Replicate measurements and external quality-control P standards established relative error at < 5%. Frozen water samples were thawed and immediately analyzed for dissolved organic C (DOC) via a TOC analyzer and for mineral N via flow-injection analysis. Two samples were below detection for NO3-N (0.1 mg N L−1) while all but one sample (0.13 mg N L−1) were below detection for NH4-N (0.1 mg N L−1).

Sediment physicochemical analyses

Throughout, we refer to wet, un-sterilized sediments as ‘fresh’ and wet, sterilized sediments as γ-irradiated. Except where noted, analyses on these sediments are for wet sediments but all values are given on a dry-weight (d.w.) mass basis.

A subsample of sediment was oven-dried (105 °C) overnight to measure moisture content. Another subsample was freeze-dried for total elemental and particle size analyses (below) as well as for storage. For the following sediment analyses, we used either duplicate or triplicate measurements. Sediment pH was measured on fresh sediments in DI H2O with a 1:5 g g−1 (sediment:solution) after 30 min equilibration. Water extractable P was measured by shaking fresh sediments with DI H2O (1:10 g g−1) for one hour, centrifuging (2400×g for 10 min), filtering the supernatant (0.45 µm) and analyzing for DRP as above. Anion storage capacity (ASC; Saunders 1965) of sediments was measured by shaking 1 g d.w. of fresh sediment in 5 mL of a 1000 mg P L−1 solution (as KH2PO4) in an acetate buffer (pH adjusted to 4.60) for 24 h. The remaining DRP in this extract was measured and ASC was expressed as % of the original concentration removed. Sediment particle sizes (on percent volume basis; Eshel et al. 2004) were determined with laser-diffraction (< 1 mm fraction only) with a Malvern Mastersizer 2000 particle size analyzer according to Sperazza et al. (2004). To meet instrumental constraints for particle size analyses, freeze-dried sediments (un-sterilized only) were sieved to < 1 mm, which represents the majority of the fine sediments collected here. Sediment total C and N were measured on freeze-dried sediments via dry-combustion (Carter and Gregorich 2007) with an Elementar Vario-Max CN elemental analyzer. Sediment total P and metals (Al, Ca, Fe, and others) were measured on freeze-dried sediments via ICP-OES (Varian 720-ES) following a microwave digestion with nitric acid and hydrogen peroxide (USEPA method 3050B; Campisano et al. 2017).

Sediment microbial enzymes and biomass P

We measured dehydrogenase activities (DHA) on both fresh and γ-irradiated sediments as an indicator for (1) microbial metabolic activity in fresh sediments and 92) sterile conditions for γ-irradiated sediments prior to EPC0 incubations. DHA relates to the breakdown of organic compounds during microbial respiration in soils and sediments (Hill et al. 2002, 2012; Prosser et al. 2011). DHA should be minimal for the γ-irradiated sediments since γ-irradiation eliminates or inactivates most microbes and enzymes (McLaren 1969; Tabatabai 1994). However, we note that some enzymes may persist after γ-irradiation (Powlson and Jenkinson 1976; Blankinship et al. 2014) and the DHA method used here (below) may give false positive readings for the sterilized samples since γ-irradiation can reduce minor amounts of redox-sensitive species (Östlund et al. 1989; Buessecker et al. 2019). Similarly, preliminary tests with autoclaved sediments yielded only a ~ 75% decrease in DHA relative to fresh sediment DHA. While not perfect, we expected DHA to be greatly reduced for sterilized sediments as a means to ensure that negligible sediment microbial activity occurred during EPC0 experiments.

For the DHA assay, we used 2,3,5-triphenyltetrazolium chloride (TTC) as the substrate, which is reduced enzymatically to triphenylformazan (TPF) as the product (Tabatabai 1994; Öhlinger and Von Mersi 1996). Two g d.w. of fresh and γ-irradiated sediments were incubated with two mL of a 0.5% (w/v) TTC solution—as recommended for coarse sediments with relatively low organic matter—in a 0.1 M TRIS buffer (pH adjusted to 7.6) for 24 h at 25 °C. The incubation was terminated by adding 10 mL of methanol (AR grade; ≥ 99.8%) and vortexing. After centrifuging, we analyzed the supernatants by measuring absorbances at 485 nm within 1 h with a 1 cm light path. Standards of TPF were prepared in methanol and analyzed in the same fashion. Method blanks were included throughout and sediments were analyzed in triplicate or duplicate (if sample amount was limited). We took care to minimize exposure to light during handling (Öhlinger and Von Mersi 1996). DHA is expressed here as mg TPF kg−1 h−1.

We estimated microbial biomass P in the fresh sediments following the methodology of Brookes et al. (1982) and McLaughlin et al. (1986) as outlined for sediments by McDowell (2003). Notably, we analyzed sediments wet rather than dry, as recommended for soil microbial biomass P analysis by Brookes et al. (1982). Additionally, while clear guidance is lacking on storage times, microbial biomass measurements for soils kept cool are stable for several weeks following initial disturbance effects from sampling (Kouno et al. 1995; Turner and Romero 2010). Hence, our analyses began ~ 3 weeks after sampling. Briefly, 1 g d.w. of fresh sediment was weighed into three centrifuge tubes, in duplicate. To one (Pkilled), 0.5 mL of liquid chloroform stabilized in amylene was added (rather than as vapor; McLaughlin et al. 1986), capped, and mixed while the other two treatments (Pfresh and Pspike) were capped—all tubes then incubated at room temperature for 24 h. Each treatment was then extracted with 0.5 M NaHCO3 (pH adjusted to 8.5) for 30 min, with the Pspike treatment receiving an additional P spike in the extraction equivalent to 25 mg P kg−1 sediment. The reactive P in the extracts was measured with a modified molybdenum-blue method suitable for alkaline extracts (Dick and Tabatabai 1977; He and Honeycutt 2005). We note that the Pfresh treatment is equivalent to Olsen P, an indicator for bioavailable P. Microbial biomass P (mg P kg−1) was calculated by:

$$\frac{({P}_{killed}-{P}_{fresh})}{{K}_{P}}\times \frac{spike}{({P}_{spike}-{P}_{fresh})}$$

where all P values (Pkilled, Pfresh, Pspike, and spike) are given on a mg P kg−1 basis and KP is a coefficient for the recovery of microbial biomass P with chloroform killing. Here, we assumed KP to be 40% (Brookes et al. 1982; McLaughlin et al. 1986; Jenkinson et al. 2004).

Equilibrium phosphate concentrations at net zero sorption (EPC0) and nutrient treatments

The equilibrium phosphate concentration at net zero sorption (EPC0) was determined for γ-irradiated sediments and then fresh sediments in close succession (completed within 48 and 96 h of sterilization, respectively). We weighed 0.5 g d.w. of wet sediment (Simpson et al. 2019) into 15 mL centrifuge tubes and added 10 mL of solution (DI water with background of 3 mM CaCl2; Lucci et al. 2010) adjusted to four P concentrations (as KH2PO4): 0, 30, 100, and 250 µg P L−1. Up to three replicates were used at each concentration depending on the amount of sample available. The tubes were shaken (end over end) for 16 h, centrifuged, and supernatants filtered (0.45 µm). Extracts were then refrigerated until analyzed for DRP within 24 h. Standards for DRP were prepared in the same background matrix (3 mM CaCl2).

For the nutrient addition experiment, we alleviated C- and/or N-limitation during the fresh sediment EPC0 incubations by adding labile C and/or N to the solutions in a factorial experimental design. We targeted a molar C:N:P ratio of 200:20:1 to create an environment enriched in C and N but limiting in P. Nutrient treatments were + C (C:P of 200:1), + N (N:P of 20:1), C + N (C:N:P of 200:20:1), or ‘none’ (only P). We added C as D-(+)-glucose and N as KNO3 to the solutions used for EPC0. We also applied the C + N treatment to the γ-irradiated sediment as a check but these data yielded no discernable difference from the standard EPC0 treatment and so are not discussed for sake of brevity.

Each sediment EPC0 was then determined as the x-intercept from regressing P sorption (mg P kg−1) against initial P concentration (Simpson et al. 2019). We excluded the points from the 250 µg P L−1 treatment, as 100 µg P L−1 already provided suitable adsorption points and greater P concentrations only diminishes the linearity of the observed P sorption near the EPC0. We estimated the uncertainty in the measured EPC0 with a 95% confidence interval (Simpson et al. 2019).

Further, under our alternative hypothesis that microbial biomass P can shift EPC0 following sterilization, we predicted changes in EPC0 post-sterilization assuming that all microbial biomass P would be re-adsorbed upon lysis. To this end, we calculated the linear sorption slopes on an equilibrium P concentration basis and projected these slopes according to the microbial biomass P; i.e., we divided microbial biomass P by the equilibrium sorption slope to get the expected change in EPC0. This result would be the expected increase in EPC0 if 100% of microbial biomass P were adsorbed by the sediments while assuming nearly linear adsorption rates. We compared these predictions to the observed changes in EPC0 following γ-irradiation.

Statistical analyses

Statistical summaries reported here are generally the median (mean ± SD). However, we do not summarize NH4-N data (too many missing values) and summarize NO3-N data (two out of 12 values below detection limit) via methods for censored data (Helsel 2005). For NO3-N summaries, we apply the Kaplan–Meier method as implemented in the ‘NADA’ R package (Lee 2020).

For the sterilization experiment (a paired treatment design), we analyzed the changes in EPC0 (n = 12) with a paired Wilcoxon signed rank test. The null hypothesis is that the distribution of pairs has a median of 0 (Hollander et al. 2013). For the alternative case, we estimated the shift in the median with the pseudo-median and its 95% confidence interval. We also examined changes in EPC0 following γ-irradiation with Spearman correlations versus catchment, sediment, and stream variables.

For the nutrient addition experiment, we sought to test for the effect of nutrients (+ C and + N) and their interaction (C + N) on EPC0 via a two-way layout (n = 48). The facts that (1) EPC0 were not estimated with the same certainty (i.e., varying confidence intervals) and (2) that the EPC0 were dependent upon the sediment analyzed (i.e., there are 12 ‘clusters’ in the data) meant that this information needed to be incorporated in our test. Thus, we analyzed this experiment with a mixed-effects model, where sediment could be a grouping variable (random effect). We used nonlinear mixed-effects (R package “nlme”; Pinheiro et al. 2020) so that we could re-write the linear sorption model to have the x-intercept (the EPC0) as a parameter to be estimated and to test for nutrient effects on EPC0 directly, making greater use of the data. More details on this approach are in the Supplementary Information.

All analyses were conducted in R, ver. 4.0.2 (R Core Team 2020).

Results

Stream and sediment characteristics

At time of sampling, the streams reflected typical summer baseflow conditions (Table 1). While the streams had appreciable DOC (median, 1.73 mg L−1) and DRP concentrations (49 µg L−1), NH4-N concentrations were mostly below detection (one stream, BP9, measured 0.13 mg L−1) and NO3-N concentrations were also low (0.21 mg L−1). Using DOC and NO3–N, stream water molar C:N had a median of 10.4. The fine (< 2 mm) benthic sediments (Table 2) were sandy, but quite sorptive (median ASC of 56% in a range of 0–100%), as expected for sediments with volcanic geology. Sediment organic matter content varied considerably across the sediment from 12 streams, with mean (± SD) sediment total C and total N concentrations of 12.8 (± 7.1) and 0.86 (± 0.51) g kg−1, respectively. Median sediment C:N was 16.9. Correlation matrices and Spearman correlations for select variables are shown in Supplementary Figs. S1-S3. No strong correlations between catchment land use and stream or sediment physicochemical characteristics were noted.

Table 1 Stream water (n = 12) physicochemistry (at time of sampling), dissolved organic C (DOC), nitrate–N (NO3-N), and dissolved reactive P (DRP)
Full size table
Table 2 Benthic sediment (n = 12 streams) characteristics, total elemental concentrations, extractable P, and sorption as anion storage capacity
Full size table

Gamma irradiation experiment

Fresh sediment dehydrogenase activities (DHA) varied widely (Supplementary Fig. S4), with mean activities of 1.89 (± 0.82) mg TPF kg−1 h−1. Fresh sediment DHA was closely associated with sediment total C (ρ = 0.91, p < 0.001) and total N (ρ = 0.93, p < 0.001). However, DHA decreased for all samples following γ-irradiation (62 to 86% decrease), commensurate with previous experimental work for soils (McNamara et al. 2003; Gebremikael et al. 2015). We consider the γ-irradiated samples here to have had negligible microbial activity throughout the EPC0 measurements given that (1) the decreasing trend in DHA with sterilization was consistent and (2) the irradiation dose given (26 kGy) was sufficient to eliminate the majority of microbes and inhibit growth for the timeframe of our experiment (48 h post sterilization; McLaren 1969; Östlund et al. 1989; McNamara et al. 2003).

Fresh sediment EPC0 ranged from 7.6 to 14.8 µg P L−1 (Fig. 2) and correlated with in-stream DRP concentrations (ρ = 0.65, p = 0.026). Sediment microbial biomass P averaged 11.5 (± 5.6) mg P kg−1 and correlated closely with fresh EPC0 (ρ = 0.87, p = 3.1e−4).

Fig. 2
figure 2

Effect of γ-irradiation on sediment equilibrium phosphate concentrations at net zero sorption (EPC0). Bars indicate the 95% confidence interval

Full size image

Following γ-irradiation, EPC0 increased for all sediments studied, though 95% confidence intervals for the calculated EPC0 overlapped for one sediment (BP2; Fig. 2). A paired Wilcoxon signed rank test estimated a median increase in EPC0 of 8.32 (6.38 to 12.6) µg P L−1 following γ-irradiation, an 83% increase relative to the median fresh EPC0. Notably, this consistent increase in EPC0 with sterilization relative to fresh EPC0 was related to microbial biomass P (Fig. 3a; ρ = 0.63, p = 0.032), with a regression slope of 0.635 (0.268 to 1.00) µg P L−1 per mg P kg−1 (the intercept was not different from zero). We note no other clear effect of γ-irradiation on P sorption relative to that of fresh sediments (e.g., changes in sorption slopes or variability in sorption).

Fig. 3
figure 3

The increase in sediment EPC0 due to γ-irradiation (sterilized EPC0 minus fresh EPC0) plotted (a) as a function of sediment microbial biomass P. A linear regression with standard error about the fit is shown. This change in EPC0 with sterilization is also plotted (b) against the predicted change in EPC0 based on the adsorption data and the microbial biomass P assuming 100% adsorption of the lysed P (dashed line is the 1:1 line). Note that these estimates could be negatively biased since we extrapolated a linear sorption curve to equilibrium concentrations past the likely range of linear sorption

Full size image

We examined the relationship between changes in EPC0 post-sterilization and microbial biomass P under the hypothesis that shifts in EPC0 may be due to adsorption of the P flush following cell lysis. The predicted increases in EPC0 based on the adsorption of the microbial biomass P are plotted against actual changes in EPC0 following γ-irradiation in Fig. 3b. These independent estimates were consistent with the measured changes in EPC0 post-sterilization (root mean square error (RMSE) of 5.22 µg P L−1) but were positively biased for larger magnitude changes.

Nutrient amendment experiment

Fresh sediments were analyzed for EPC0 with additions of glucose and/or KNO3 solutions to remove C- and/or N-limitation on microbial growth. When considering the measurement uncertainties within a given sediment, EPC0 did not appear to vary significantly between the nutrient treatments for any one sediment (Fig. 4). However, when modeling the full data set with nonlinear mixed effects, we found a statistically significant decrease in EPC0 with N addition but no significant change with C addition (Table 3; more modeling details in Supplementary Information). The likelihood-ratio tests and Akaike Information Criterion showed that more complex models with an effect for + C (p = 0.07) and the interaction effect (C + N; p = 0.62) on EPC0 did not out-perform the simpler model with only the + N effect. Though statistically significant, the + N effect on EPC0 was relatively small, with a change in EPC0 of − 0.56 (95% CI − 0.81 to − 0.31) µg P L−1. Higher order effects (e.g., + N effect on EPC0 varying per individual sediment) were not warranted by the data, neither were nutrient effects on sorption slopes.

Fig. 4
figure 4

Sediment EPC0 and its 95% confidence interval (via individual linear fits) for sediments from all 12 study streams and for four nutrient treatments: none (i.e., the original EPC0 method, denoted here as −), + C (C:P of 200:1), + N (N:P of 20:1), and C + N (C:N:P of 200:20:1). Note that γ-irradiated sediments were also analyzed for the C + N treatment but this data did not differ from the original EPC0 data

Full size image
Table 3 Model summary for the nonlinear mixed effects fit for the nutrient amendment experiment
Full size table

Discussion

Microbial biomass as a stock of phosphorus: lysed P accounts for changes in EPC0

Upon γ-irradiation, the flush of P from the microbial biomass (median of 10 mg P kg−1) was free to adsorb onto the sediment, increasing EPC0 (Figs. 2, 3). While this result does not speak to the direct comparison of biotic and abiotic contributions towards sediment P uptake, it does show the potential influence of the sediment microbial biomass as a stock of P in streams. Similar to their contribution towards total P cycling in terrestrial ecosystems (Cleveland and Liptzin 2007; Turner et al. 2013), the microbial biomass represents a significant fraction of the total P in stream ecosystems. Importantly, sediment microbial biomass P is likely a very transient P pool, therefore contributing to P dynamics throughout the stream network (Mulholland et al. 1997; McDowell 2003; McDowell and Sharpley 2003; Cross et al. 2005).

Few data exist on microbial biomass P in stream and river sediments, with ranges (using the same method from Brookes et al. (1982)) varying from: 0.5 to 10 mg P kg−1 in a 3rd to 4th order catchment having intensive forestry and pastoral land use (McDowell 2003), 2.4 to 23 mg P kg−1 in the present study, and 10 to 45 mg P kg −1 in a large river subject to considerable agricultural and urban pollution sources (Jaiswal and Pandey 2019). Streams with less P inputs will likely have much less sediment microbial biomass P (e.g., site BP1 in this study) though this also depends partly on the sediment (see below). For comparison, soil microbial biomass P concentrations can be an order of magnitude greater, varying from 3 to 430 mg P kg−1 (Cleveland and Liptzin 2007), likely due to the relatively greater organic matter content (Sinsabaugh et al. 2012). Considering that biomass C:P ratios are often more P-rich in sediment microbes (about 60:1; Cleveland and Liptzin 2007; Sinsabaugh et al. 2009) compared to periphyton (roughly on the order of 102 to 103; Cross et al. 2005), future work could investigate sediment microbial biomass as an important and perhaps variable stock of P in streams.

We assumed all of sediment microbial biomass P would be adsorbed by sediment surfaces and at a linear rate. However, we urge caution in applying this assumption to other sediments, owing to potential nonlinear P sorption (Froelich 1988; House 2003) and that amounts of microbial biomass P likely differ. Nonetheless, the consistency with actual changes in EPC0 following γ-irradiation corroborates the hypothesis that the microbial biomass P stock can re-adsorb onto sediment surfaces, thereby increasing EPC0 following sterilization. Since γ-irradiation is the least likely among sterilization methods to alter sediment physicochemical properties (Eno and Popenoe 1964; Berns et al. 2008; Buessecker et al. 2019), we can reasonably assert that the re-distribution of P—and the resulting increase in EPC0—was primarily due to cell lysis (Meyer 1979; Klotz 1985).

Reconciling past experiments on biotic vs. abiotic sediment P uptake

We suspect that our results may explain two of the inconsistences in past attempts to separate biotic and abiotic P uptake in sediments: sterilization techniques and a lack of microbial biomass P data. Firstly, more disruptive sterilization techniques than γ-irradiation have generally been used, namely, autoclaving and biocides. For example, when analyzing autoclaved sediments, several studies observed impressive differences between P sorption metrics for fresh and autoclaved sediments, concluding that up to ~ 40% of the total sediment sorption capacity was biotic (Haggard et al. 1999; Ryan et al. 2007; McDaniel et al. 2009). However, autoclaving likely changed the physicochemistry of these sediments since, in addition to cell lysis, autoclaving can disrupt sediment organic matter, alter surface chemistry, and increase some extractable elements such as Al, Fe and Mn (Meyer 1979; Wolf and Skipper 1994; Trevors 1996; Buessecker et al. 2019). In two comparison studies, Meyer (1979) and Klotz (1985) assessed potential biotic influence on sediment P uptake by comparing sediments analyzed fresh, autoclaved, and either γ-irradiated or with a phosphorylation inhibitor (cabonyl-cyanide m-chlorophenylhydrazone; CCCP), respectively. They found that autoclaving sediments released more P but also changed P sorption kinetics and capacity. In contrast, γ-irradiated sediments had sorption kinetics comparable to that of fresh sediments but had an increase in EPC0 (attributed to cell lysis), while CCCP-treated sediments had no change in sorption capacity relative to fresh sediments. Similarly, Triska et al. (2006) found autoclaved sediments released more than 10 × more P than fresh sediments in desorption assays and retained ~ 40% less P in sorption assays whereas a CCCP-treated sediment was comparable to fresh sediment for both sorption and desorption.

Aside from autoclaving, various biocides (e.g., HgCl2, azide) have been applied to separate biotic and abiotic P uptake in stream sediments. Results with biocides have ranged from small increases in EPC0 (2 µg P L−1; Griffiths and Johnson 2018) to large or unrealistic changes in EPC0 and sorption capacities (Lottig and Stanley 2007; Stutter et al. 2010). However, little is known about the potential unintended chemical interactions these biocides have with sediment surfaces. For example, HgCl2 can react with organic matter, altering solution pH in the process (Buessecker et al. 2019), as noted by Griffiths and Johnson (2018) in their fresh vs. sterilized EPC0 comparison. Buessecker et al. (2019) compared multiple methods to remove only the biotic mechanisms for soil N2O production and concluded γ-irradiation to be the most suitable, as other methods (autoclaving, azide, HgCl2, Zn, and chloroform) produced more physicochemical changes relative to fresh soil. We argue that such effects also apply to stream sediments. Neither autoclaving nor biocides will likely be accurate in removing solely the biotic reactions responsible for sediment P uptake. Thus, while not always acknowledged, conclusions for biotic and abiotic sediment P uptake reliant on autoclaving or biocides are most likely inaccurate.

In addition to disruptive sterilization methods, a second source for error in past estimates of biotic contributions to sediment P uptake was a lack of microbial biomass P data. As discussed, the microbial biomass can release a substantial flush of P upon lysis (Fig. 3a) that can shift the equilibrium P concentration considerably. If our alternative hypothesis is correct—that the increase in EPC0 with non-disruptive sterilization is due to re-adsorption of the microbial biomass P following cell lysis (Fig. 3b)—then we predict that the effect of sterilization would be even more pronounced for sediments with comparable microbial biomass P but weaker sorption capacities. Owing to their volcanic geology and high metal-oxide content (Dahlgren et al. 2004), our study sediments were relatively reactive and had high P sorption capacities (mean ASC of 54%); for comparison, the mean ASC from a national survey of sediments from 76 large New Zealand rivers was 9% (McDowell 2015). Interestingly, Munn and Meyer (1990) compared fresh and γ-irradiated sediment EPC0 for two streams: one from a volcanic geology and one from a weathered granite geology. While both sediments had similarly low EPC0 (5.4 and 1.0 µg P L−1, respectively), the increases post-sterilization were respectively 3.2 and 9.7 µg P L−1, consistent with a greater P sorption capacity in the volcanic geology. This increase was attributed to a predominance in biotic P uptake with no acknowledgement of possible bias due to the microbial biomass P.

Does carbon and nitrogen addition promote biotic P uptake and lower EPC0?

Microorganisms in sediments demand nutrients according to the elemental stoichiometry of their biomass (Sterner and Elser 2002; Sinsabaugh et al. 2009). In contrast to lotic biofilm communities towards the top of the benthic zone (i.e., where light promotes autotrophic growth; Battin et al. 2016), and unlike for soil, sediment microbial communities (primarily heterotrophic) are more often C- or N-limited rather than P-limited (Hill et al. 2010, 2012; Sinsabaugh et al. 2012). Hence, we would only expect significant sediment biotic P uptake when C- and N-limitation is relieved. For sediments, microbial P-limitation is likely when C:P (molar basis) exceeds 60:1 and when N:P exceeds 7:1 (Cleveland and Liptzin 2007; Sinsabaugh et al. 2012). Under P-limitation, we expect that sediment microbes would sequester P for growth, thus potentially lowering the sediment EPC0 through desorption. In other words, under P-limitation, sediment microbes may deplete P concentrations in solution below the sediment EPC0, promoting P desorption from sediment surfaces and potentially lowering sediment EPC0.

In this study, EPC0 did not respond to C addition but decreased with N addition in both the + N and C + N treatments (− 0.5 µg P L−1; Fig. 4; Table 3). The + N effect here may have been dampened due to the large P (de)sorption capacity of these sediments (Small et al. 2016; Griffiths and Johnson 2018), as sediments display the greatest P buffering when near the EPC0 (Froelich 1988). Indeed, the literature notes varied responses in P uptake to N supply. Using sediments in flume experiments at natural stream temperature, McDowell et al. (2017) observed relatively greater DRP uptake with N additions, a likely biotic effect since there was also a consistent increase in sediment microbial biomass P for the treatments with additional N. In contrast, Griffiths and Johnson (2018) measured stream P uptake rates while varying the background NO3-N concentration: despite evidence for strong N- and P-colimitation, in-stream P uptake was unaffected by background N concentration, suggesting that stream P uptake was predominantly controlled by abiotic sediment P sorption rather than biotic P uptake. In a stream with plentiful N supply (mean of 1.1 mg NO3-N L−1), Oviedo-Vargas et al. (2013) measured P spiraling and sediment EPC0 during an experimental amendment of labile C (acetate) ~ 1 mg C L−1 above ambient conditions to stimulate sediment heterotrophic P demand. Despite greater C uptake during the experiment, P uptake was insensitive to the C addition (as whole-stream P uptake velocities and as sediment EPC0). In their study, Oviedo-Vargas et al. (2013) suggested that the predominantly heterotrophic microorganisms within the sediments were meeting their P demands through organic P mineralization or from sediment P desorption rather than from the water column. Similarly, the sediments in the present study buffered solution P towards the EPC0 through abiotic sorption processes in spite of biotic P uptake.

A reason for the response to N but not C may lie in the fact that labile C can increase microbial respiration (Roberts et al. 2007; Oviedo-Vargas et al. 2013; Demars 2019) but will not directly affect P concentrations. It is microbial production, not respiration, that carries a P demand (Sterner and Elser 2002). Sediment microorganisms here may have had enough C supply available to meet their demands for production, thereby not influencing EPC0 for the + C treatment. For example, sediment biofilms can hold over C in their extracellular polymeric substances (Battin et al. 2016) or enzymatically degrade organic substrates (Findlay et al. 2003; Sinsabaugh et al. 2009; Kaplan and Cory 2016) to meet C requirements. Most of the catchments in our study had some amount of pasture (35 to 82% for sites BP2 through BP12), which may provide an ample source of C when soil pore waters hydrologically connect to the streams (Kaplan and Cory 2016; McNally et al. 2017). Additionally, these pasture-dominated catchments had diminished stream canopy cover, meaning likely greater autochthonous C supplies for the stream ecosystem due to increased primary productivity (Dodds 2007; Finlay 2011). This may have supplied sufficient high-quality C resources (e.g., from algal biomass turnover; Meyer 1994; Kaplan and Cory 2016) to the sediments to prevent any C-limitation before EPC0 measurements.

Our results from labile C and N additions to EPC0 measurements suggest little to no influence of biotic P demand on EPC0 for these streams, with only a small overall effect arising from the removal of potential N-limitation (Table 3). Coupled with the results of the first experiment—that much of the apparent differences in P sorption between fresh and sterilized sediments is due to the microbial biomass P flush—this experiment suggests that sediment biotic P demand may not be a significant part of the EPC0 as it is typically measured, i.e., with only P added. Rather, the benthic sediment EPC0 may primarily reflect the sediment’s abiotic sorption characteristics and prior P loading to that sediment (Froelich 1988; McDaniel et al. 2009; McDowell 2015). Our results come from streams that are likely N-limited and that drain a naturally P-rich geology. For other streams, however, if sediments are poorly sorptive (e.g., little Al/Fe oxides or coarse particle sizes) and P availability is low, then biota may have a relatively stronger influence on P uptake in sediments (Lottig and Stanley 2007). The dynamic between P sorbed to sediments and P immobilized in microbial biomass should be explored in other systems that vary in stream geomorphology, sediment characteristics, and nutrient inputs.

Improving our understanding of the role of microbes in stream P attenuation

The results from this study help to separate some of the processes usually lumped together in sediment P sorption and, at the greater scale, P attenuation in streams. Characterizing these processes more clearly will help target future research on sediment P reactions, for example, in determining the role of microbial biomass as a stock of P and its relationship with the abiotic sediment P compartment. When one measures EPC0, the EPC0 is unlikely to be affected by biotic uptake partly since P is often not the limiting nutrient for sediment heterotrophic microbes. Rather, EPC0 is largely a function of the sediment reactive surfaces and the previous exposure of P to that sediment. Integrating this information and more biota-specific P reactions with hydrological fluxes will improve our understanding of stream and catchment-scale P biogeochemistry (Manzoni and Porporato 2011).

Sediment EPC0 provides information on how sediments subsidize available P for microbial growth and so influence stream P biogeochemistry. Sediments may either maintain or accumulate more sorbed P (DRP ≥ EPC0) but, when local DRP concentrations are depleted through inter alia microbial P uptake (DRP < EPC0), sediments may desorb P so nearby microorganisms can meet their stoichiometric constraints. If sediments drip feed P—according to the EPC0—to ‘P-hungry’ sediment biofilms (particularly heterotrophic bacteria), then we could expect a positive relationship between EPC0 and microbial biomass P. Indeed, the two variables had a strong positive correlation (ρ = 0.87, p = 3.1e−4), suggesting that sediment microorganisms play a role in regulating P desorption from sediments back into the water column. This coincides with what we know on the coupled biogeochemical cycling of C, N, and P in streams (Oviedo-Vargas et al. 2013; Maranger et al. 2018) and has important consequences. For example, the P subsidy provided by stream sediments (as approximated by EPC0) may promote greater C and N processing in streams (Stelzer et al. 2003; Findlay and Sinsabaugh 2006; Tank et al. 2010). Another consequence, however, is that this P supply could support a large sediment biofilm, meaning the biofilm may clog hyporheic flowpaths (Hartwig and Borchardt 2015; Battin et al. 2016) thus limiting the interaction between reactive sediment surfaces and P in the water-column (Boano et al. 2014; Weigelhofer et al. 2018). Further exploring the sediment P buffer will likely aid us in better characterizing stream P biogeochemistry and its consequences (Hamilton 2012).

Overall, considering the dynamic between the sediments and their microbial biofilms will help in understanding the stream sediment P buffer and hence stream P attenuation. This understanding can sharpen our view of stream P biogeochemistry but also help in understanding the effects of our efforts in mitigating nutrient pollution, e.g., from changing wastewater inputs (Scott et al. 2011; Wilcock et al. 2020) or from management on-land (Meals et al. 2010; McDowell et al. 2018). This recognition of both biotic and abiotic mechanisms can help avoid unrealistic expectations by policy agencies in the speed or efficacy of such efforts to mitigate P pollution.