Carbon and nitrogen inputs differentially affect priming of soil organic matter in tropical lowland and montane soils

22 Microbial decomposition of soil organic matter (SOM) can be accelerated or reduced by the 23 combined effects of carbon (C) and nutrient inputs through a phenomenon known as 24 ‘priming’. Tropical lowland and montane soils contain large stores of C and may undergo 25 substantial future changes in C and nutrient inputs due to global change, yet how these inputs 26 might interact to influence priming is poorly understood in these ecosystems. We addressed 27 this question using soils from a 3400 m tropical elevation gradient which vary strongly in 28 nitrogen (N) and phosphorus (P) availability. To determine how existing nutrient availability 29 in different tropical soils regulates microbial activity, and whether microbial demand for 30 nutrients leads to priming, soils were amended with simple and more complex 13 C-labelled 31 substrates in combination with inorganic N, P and N+P. Isotopic partitioning ( 13 C in CO 2 and 32 in phospholipid fatty acids; PLFA) was used to identify sources of C (substrate- or SOM- 33 derived) in respiration and in microbial communities. Nutrient treatments did not influence 34 the amount of substrate-respired C for any of the soils, but did affect the direction and 35 magnitude of priming effects. For the upper montane forest and grassland soils, C addition 36 had a relatively minor influence on the turnover of SOM, but N addition (with or without C) 37 reduced SOM mineralisation (negative priming), suggesting reduced microbial N-mining 38 from SOM when N was externally supplied. By contrast, in the lower montane and lowland 39 forest soils, C addition increased SOM mineralisation (positive priming), but the response 40 was unaffected by nutrient additions. The assimilation of 13 C substrates into functionally 41 active microorganisms revealed that C substrate complexity, but not added nutrients, strongly 42 affected C-use within the microbial community: in both lowland and montane forest soils, 43 fungi assimilated a greater proportion of the simple C substrate, while gram-positive bacteria 44 assimilated a greater proportion of the more complex C substrate. Overall, our results have global change.

Microbial decomposition of soil organic matter (SOM) can be accelerated or reduced by the 23 combined effects of carbon (C) and nutrient inputs through a phenomenon known as 24 'priming'. Tropical lowland and montane soils contain large stores of C and may undergo 25 substantial future changes in C and nutrient inputs due to global change, yet how these inputs 26 might interact to influence priming is poorly understood in these ecosystems. We addressed 27 this question using soils from a 3400 m tropical elevation gradient which vary strongly in 28 nitrogen (N) and phosphorus (P) availability. To determine how existing nutrient availability 29 in different tropical soils regulates microbial activity, and whether microbial demand for 30 nutrients leads to priming, soils were amended with simple and more complex 13 C-labelled 31 substrates in combination with inorganic N, P and N+P. Isotopic partitioning ( 13 C in CO 2 and 32 in phospholipid fatty acids; PLFA) was used to identify sources of C (substrate-or SOM-33 derived) in respiration and in microbial communities. Nutrient treatments did not influence 34 the amount of substrate-respired C for any of the soils, but did affect the direction and 35 magnitude of priming effects. For the upper montane forest and grassland soils, C addition 36 had a relatively minor influence on the turnover of SOM, but N addition (with or without C) 37 reduced SOM mineralisation (negative priming), suggesting reduced microbial N-mining 38 from SOM when N was externally supplied. By contrast, in the lower montane and lowland 39 forest soils, C addition increased SOM mineralisation (positive priming), but the response 40 was unaffected by nutrient additions. The assimilation of 13 C substrates into functionally 41 active microorganisms revealed that C substrate complexity, but not added nutrients, strongly 42 affected C-use within the microbial community: in both lowland and montane forest soils, 43 fungi assimilated a greater proportion of the simple C substrate, while gram-positive bacteria 44 assimilated a greater proportion of the more complex C substrate. Overall, our results have 45

52
Tropical soils are a globally important store of terrestrial carbon (C) (Jobbágy and Jackson, 53 2000), with soil microorganisms playing a decisive role in regulating net soil C storage 54 through the mineralisation of plant residues and soil organic matter (SOM). Elevated 55 concentrations of atmospheric carbon dioxide (CO 2 ) are expected to increase plant 56 productivity and, in turn, inputs of plant-derived C to soil (Cusack et al., 2016). However, 57 because C and nutrient cycles are tightly coupled according to microbial metabolic demands 58 (Cleveland and Liptzin, 2007;Finzi et al., 2011), the availability of essential nutrients may 59 strongly influence the response to changes in C input (Dijkstra et al., 2013;Chen et al., 2014;60 Carrillo et al., 2017). In the tropics, external supply of nitrogen (N) is increasing as a 61 consequence of biomass burning (Hietz et al., 2011), with high N deposition reported even 62 across remote Andean systems (Fabian et al., 2005;Boy et al., 2008). Aeolian sources of 63 phosphorus (P), originating from North Africa, will also likely increase the future availability 64 M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 8 measure total C and N concentrations, using a TruSpec CN elemental analyser (LECO,169 USA). Total P concentration was measured using a modified dry ash procedure (Enders and 170 Lehmann, 2012) with sulphuric acid-hydrogen peroxide digestion followed by phosphate 171 detection by automated molybdate colorimetric analysis (Bran Luebbe AA3 Autoanalyser, 172 Germany). 173 Maximum water holding capacity (WHC) was measured using composite samples of 174 soils from each elevation site (homogenised composite of soil from four subplots from each 175 elevation). Soils were saturated with deionised water and left to drain for 6 hours in a fully 176 humid airspace, before drying at 105 °C to constant mass to calculate water content (Öhlinger 177 and Kandeler, 1996). Soil moisture content was standardised across soils to 80 % maximum 178 WHC (by addition of sterile de-ionised water where necessary, allowing for later addition of 179 C and nutrient treatments in 400 µl + 400 µl solution per assay), chosen so that microbial 180 respiration was not constrained by very high or low soil water content. 181 182

Amendment of soils with 13 C-labelled substrates and labile nutrients 183
To determine if soil nutrient availability influenced microbial mineralisation of fresh C 184 substrates and priming of pre-existing SOM, soils were amended with 13 C-labelled xylose or 185 hemicellulose in combination with three inorganic nutrient treatments (N, P, NP), plus 186 controls, in a fully factorial design. Xylose and hemicellulose were selected as two 187 ecologically relevant C substrates which differ in their chemical complexity; xylose 188 represents a simple C substrate (monosaccharide, building block for hemicellulose) and 189 hemicellulose a more complex molecule (polysaccharide, constituent of plant cell walls). 190 13 C-labelled xylose (Sigma-Aldrich, UK) and hemicellulose (IsoLife,The 191 Netherlands) of 10.8 atom% enrichment were created by mixing uniformly labelled 13 C-192 substrates (where 13 C was evenly distributed across the C molecule) with equivalent non-M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 9 labelled substrates. Carbon substrates were added at a concentration of 200 µg C g -1 fwt soil, 194 equivalent to 0.2-0.3 % of total C in the soils (Table 2). This dose rate was chosen based 195 upon previous experimental findings, whereby this amount of added C equated to 53-100 % 196 of initial microbial biomass C (Whitaker et al, 2014b), such that microbial activity and 197 respiration would be stimulated by the added substrate without inducing a significant increase 198 in microbial growth (Blagodatskaya and Kuzyakov, 2008). Xylose and hemicellulose 199 treatments were prepared by dilution in sterile deionised water so that each substrate was 200 added in 400 µl solution per assay. As hemicellulose is insoluble, it was diluted into 201 suspension by sonication and vortexed for 5 seconds prior to addition to soils. 202 Three inorganic nutrient treatments (N, P and NP) were included in the experiment. 203 Nitrogen was added as ammonium nitrate (NH 4 -NO 3 ), phosphorus as monosodium phosphate 204 (NaH 2 PO 4 ) and a combined N+P treatment where ammonium nitrate and monosodium 205 phosphate were added in solution together. The concentration of these nutrient treatments 206 was determined in order to correspond to mean C:N:P stoichiometry of soil microbial 207 biomass (60:7:1) which is reported to be tightly constrained at a global scale (Cleveland and 208 Liptzin, 2007). Hence, soils were amended with N and P treatments in a fixed ratio with the 209 added C substrate (200 µg C g -1 fwt soil), with C: N ratio of 60:7 (equivalent to 20 µg N g -1 210 fwt soil) and C: P ratio of 60:2 (equivalent to 7 µg P g -1 fwt soil). Phosphorus was added in 211 excess of C: P ratio 60:1, to account for P-sorption to clay minerals. Nutrient treatments were 212 prepared by dissolving into sterile deionised water so that each treatment was added in 400 µl 213 solution per assay. 214 Prior to the start of the experiment, soils were pre-incubated in the dark at 16.0 °C 215 (chosen as the average mean annual temperature of the four soils used in the experiment; 216 Table 1) for 24 hours to allow equilibration to the experimental incubation temperature. One 217 common incubation temperature was chosen in order to focus on treatment differences 218 M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT among soils. Aliquots of 8.0 g fwt soil were placed into 160 ml glass Wheaton bottles 219 (Wheaton Science Products, USA). Soils were amended with one of the 13 C-labelled 220 substrates (xylose, hemicellulose or control) in combination with one of the nutrient 221 treatments (N, P, NP, or control), with four replicates per treatment. For controls, sterile 222 deionised water was added in place of C and/or nutrient treatments. The headspace of each 223 bottle was flushed with compressed air for 1 minute to achieve a standard starting atmosphere 224 before bottles were sealed with butyl rubber stoppers and aluminium crimp caps. Bottles 225 were over-pressurised by injecting 20 ml compressed air, to partly compensate for subsequent 226 headspace gas sampling, and incubated at 16.0 °C in the dark for 7 days. In total, 192 soil 227 assays were incubated. As soils were contained in sealed Wheaton bottles, soil moisture was 228 maintained throughout the study, without need for additional adjustment. 229 Samples of compressed air were taken to measure starting gas concentrations, at 230 time 0. To determine the evolution of CO 2 and its 13 C-enrichment, two headspace gas 231 samples were taken from each bottle at 3 time points (24 hours, 48 hours and 168 hours after 232 C substrate-nutrient addition) by taking 5 ml gas samples with an air-tight syringe and 233 injecting into 3 ml evacuated exetainer vials (Labco, UK). After 7 days, at the end of the 234 experiment, soils were frozen at -80 °C and freeze dried for analysis of phospholipid fatty 235 acid (PLFA) biomarkers. Responses measured therefore reflected short-term responses to 236 altered C and nutrient supply. 237 238 2.3.1 CO 2 and 13 C-CO 2 analyses 239 The concentration of CO 2 in gas samples was determined using a PerkinElmer Autosystem 240 Gas Chromatograph (GC; PerkinElmer, USA) fitted with a flame ionization detector 241 containing a methaniser, with results calibrated against certified gas standards (BOC Ltd. 242 Guildford, UK). Carbon dioxide fluxes were calculated using the linear accumulation of CO 2 M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT concentrations in headspace gas samples from 0, 24, 48 and 168 hours, using the approach 244 described by Holland et al. (1999). Linear fluxes best described the data, whereby the 245 Pearson coefficient for accumulation of CO 2 with time was greater than 0.95 for each bottle. 246 The total CO 2 flux was calculated on a soil C mass basis (CO 2 -C µg g −1 soil C), to normalise 247 for differences in C content among soils. 248 δ 13 C values of CO 2 were measured by isotope ratio mass spectrometry (IR-MS) using 249 a Trace-Gas Preconcentrator coupled to an Isoprime isotope ratio mass spectrometer (IRMS,250 Elementar, UK). Between 5 and 60 µl gas samples were manually introduced into the 251 injection port of the pre-concentrator using a gas-tight syringe. Water was eliminated via a 252 perchlorate chemical trap and the CO 2 cryogenically pre-concentrated by liquid nitrogen prior 253 to gas chromatography column separation and introduction to the Isoprime IRMS via an open 254 split. The resultant 13 C/ 12 C isotope ratios were compared to pulses of known reference CO 2 255 (BOC, UK) calibrated against certified reference CO 2 (NIST RM8562 CO 2 -Heavy, RM8563 256 CO 2 -Light and RM8564 CO 2 -Biogenic) to determine δ 13 C values of CO 2 . 257 258

PLFA and 13 C-PLFA analyses 259
Microbial community composition was determined by analysis of PLFA biomarkers 260 extracted from soils after the 7-day experimental period. The number of samples extracted for 261 PLFAs was rationalised due to the time and resource intensity of these measurements. Hence, 262 upper montane (Wayqecha; 3025 m asl) and lowland (Tambopata; 210 m asl) forest soils 263 which had been amended with xylose plus N/P/control and with hemicellulose plus 264 N/P/control were selected, to identify how the addition of N or P affected the assimilation of 265 simple and more complex C substrates by different microbial functional groups. From these 266 assays, three out of four replicate soils were chosen randomly. Untreated control soils (no M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT added C, no added nutrient treatment) were also extracted for microbial PLFAs to determine 268 the natural abundance of 13 C-PLFA in the soils. 269 Phospholipids were extracted as part of the total lipid extract from 0.50 g freeze-dried 270 and ground soil samples by a Bligh-Dyer extraction, consisting of chloroform, methanol and 271 0.15 M citrate buffer (1:2:0.8, v/v/v) (Frostegård et al., 1993). analyser-IRMS (CF-EA-IRMS), to determine its δ 13 C value which was used to back-correct 296 the PLFA δ 13 C values for the addition of the extra C atom introduced to the molecule during 297 methylation (Jones et al., 1991; see 13 C-PLFA calculations). 298 299

Stable isotope calculations 300
Enrichment of 13 C in CO 2 and PLFAs were expressed in delta notation, i.e. δ 13 C (Eq. 1), 301 where R represents the 13 C/ 12 C ratio in the sample relative to the standard (Coleman, 2012). 302

13 C-CO 2 calculations 305
The percentage of respired CO 2 from added C substrates was calculated for treated soils at 24, 306 48 and 168 hours, according to equation 2, derived from a mixing model. δC is the δ 13 C value 307 from the soil source, determined from Keeling plots as the intercept (x = 0) of the linear 308 regression between 1/CO 2 concentration and δ 13 C (for untreated control soils, no added C or 309 nutrient treatment). δL is the δ 13 C value of the labelled C-substrate (10.8 atom% 13 C 310 enrichment, δL = 9774.6), and δT is the δ 13 C value of respired CO 2 from treated soils at each 311 time point. These data were used to calculate cumulative substrate-derived C (µg CO 2 -C g −1 312 soil C) respired after 24, 48 and 168 hours for all soil/C substrate/nutrient treatments. 313 The change in mineralisation of pre-existing SOM following the addition of C substrates and 317 nutrient treatments (hereafter referred to as primed C) was calculated using a mass balance 318 approach. Primed C (µg CO 2 -C g −1 soil C) was estimated from total measured respiration in 319 treated soils, minus substrate-derived respiration, minus basal respiration (from control 320 untreated soils). For soils which were only amended with nutrient treatments (no added C 321 substrate), the change in mineralisation of SOM (hereafter also referred to as primed C) was 322 calculated as total measured respiration in treated soils minus basal respiration (from control 323 soils). Primed C could be positive or negative, where positive priming represented increased 324 mineralisation of pre-existing SOM following the addition of C substrates and/or nutrient 325 treatments and negative priming represented reduced mineralisation of pre-existing SOM 326 following the addition of C substrates and/or nutrient treatments. 327 328

13 C-PLFA calculations 329
Isotopic enrichment of individual PLFAs were expressed as δ 13 C PLFA values after correction 330 for the methyl-group added during methanolysis (Eq. 3), where nPLFA is the number of C-331 atoms of the PLFA molecule, δ 13 C FAME is the δ 13 C value of the FAME after methylation and 332 δ 13 C MeOH is the δ 13 C value of the methanol used for methanolysis (δ 13 C = -54.15 ‰). 333 To determine the assimilation of added C substrates by different microbial functional groups, 337 the percentage of substrate-derived C within individual PLFAs was calculated by a modified 338 version of equation 2, multiplied by the abundance of the specific PLFA (µg PLFA-C g -1 dwt (Nottingham et al., 2009), where δC is the δ 13 C-PLFA value from untreated control 340 soils, δL is the δ 13 C value of the labelled substrate (as above), and δT is the δ 13 C-PLFA value 341 from treated soils. These data were used to calculate the incorporation of the added substrate-342 C into individual PLFAs, from which the total C substrate incorporation into all PLFAs (µg 343 substrate-derived PLFA-C g −1 soil dwt) was determined. The proportional substrate 344 incorporation (% substrate-derived C) into different microbial functional groups (F, GP 345 bacteria, GN bacteria, and unspecified PLFAs) were also calculated, as a percentage of the 346 total C-substrate incorporated into all PLFAs. While it was not possible to determine an 347 absolute value for microbial carbon use efficiency (CUE), here we calculated an index of 348 substrate-CUE as the total substrate-derived C incorporated into PLFAs (used for growth) 349 relative to the total substrate-derived C consumed (substrate-C incorporated in PLFAs + 350 substrate-C respired) (Whitaker et al., 2014a), to screen for differences in the CUE of added 351 substrates between soils and with nutrient treatments. Finally, to examine the incorporation of 352 soil-derived C by different microbial functional groups in response to C and nutrient 353 treatments, concentrations of excess soil-derived C in PLFAs (µg soil C g -1 soil) were 354 determined by a mass balance approach, as the total concentration of PLFA-C in treated soils, 355 minus PLFA-C in untreated control soils, minus substrate-derived PLFA-C (Nottingham et 356 al., 2009). Positive excess soil-derived C in PLFAs represented increased assimilation of C 357 from SOM and negative excess soil-derived C represented reduced assimilation of C from 358 SOM, relative to untreated control soils. 359  Total C, N and P (organic horizon) was lowest in the lowland forest soil ( Table 2). The 378 montane forest soils (1500 m and 3025 m) had the greatest concentration of total N (2.3 %), 379 while the lower montane forest soil had the greatest concentration of total P (0.134 %). The 380 concentration of total N and P in these soils does not however directly correspond to the 381 availability of inorganic N and P. Extractable inorganic PO 4 -P was low in the lowland forest 382 soil, and typically increased with increasing elevation, but was also very low in the montane 383 grassland soil (Nottingham et al., 2015b). Conversely, mineralised N (NH 4 + NO 3 ) 384 determined by in-situ ion-exchange resins was higher in soils from the lowland and lower 385 montane forest, compared to soils from the upper montane forest and montane grassland 386 (Nottingham et al., 2015b; Table 2). Soil pH did not vary markedly among soils from the 387 three forest sites (pH 3.6-3.9) but was higher in the montane grassland soil (pH 4.9). 388 389 M A N U S C R I P T

Microbial mineralisation of 13 C substrates and priming of SOM 390
Basal respiration (CO 2 flux in the absence of added C and nutrient treatments), under 391 controlled temperature and soil moisture, varied 2.7-fold among soils when compared on a 392 soil C mass basis (Fig. 1). Total respired CO 2 varied among soils and C substrate treatments 393 with a significant 'Soil x C Substrate' interaction (Supplementary Table S1). Post-hoc tests 394 revealed that, typically, respired CO 2 was significantly greater following amendment with 395 xylose compared to un-amended (no added C) soils (Supplementary Table S2). The exception 396 was soil from the montane grassland (3644 m), where there was no significant difference 397 among un-amended and xylose-amended soils (Fig. 1). Respired CO 2 from the lowland forest 398 (210 m) soil amended with hemicellulose was also significantly greater compared to un-399 amended soil, but this relationship did not hold for the other soils. 400 Total respired CO 2 -C was partitioned into that derived from added C substrates 401 (substrate C) and from the mineralisation of pre-existing SOM (primed C). The cumulative 402 amount of substrate-derived C respired after 168 hours varied among soils and C substrates, 403 with a significant 'Soil x C Substrate' interaction (Supplementary Table S1). After 168 hours, 404 respired C from xylose (simple C) was significantly greater than from hemicellulose (more 405 complex C). This relationship (xylose > hemicellulose) was consistent for all soils, despite 406 significant differences in the overall amount of substrate C respired among soils (Fig. 2). 407 There was no significant effect of nutrient treatment on substrate-respired C in any treatment 408 combination or soil. 409 The direction and magnitude of primed C after 168 hours varied among soils, C 410 substrates and nutrient treatments, with significant 'Soil x C Substrate' and 'Soil x Nutrient' 411 interactive effects (Supplementary Table S1 the magnitude of which was dependent on nutrient treatments (Fig. 3 a-b). 416 For the upper montane forest (3025 m) and grassland (3644 m) soils, amendment with 417 C substrates had a relatively minor influence on primed C, with no significant difference 418 among responses to control (no added C), hemicellulose and xylose treatments (Fig. 3 a-b). 419 Instead, for these soils, priming responses varied significantly among nutrient treatments, 420 whereby priming induced by N and NP (both alone and in combination with C) was more 421 negative compared to soils without added nutrients. There was no significant difference 422 among responses to N and NP treatments for these soils. Amendment of the montane 423 grassland soil with xylose P enhanced the mineralisation of SOM (positive priming), 424 compared to the negative priming induced by xylose alone (Fig. 3a), a response not evident in 425 the upper montane forest soil. 426 In the lower montane forest (1500 m) and lowland forest (210 m) soils, the magnitude 427 of priming responses after 168 hours varied with C substrate, whereby xylose stimulated 428 significantly greater priming of SOM compared to hemicellulose (Fig. 3 c-d). Greater 429 priming of SOM following amendment with xylose was, however, likely a function of greater 430 mineralisation of the more labile xylose substrate over the 7-day experimental period, rather 431 than greater priming of SOM stimulated by xylose per se, as for these soils the proportion of 432 primed to substrate-derived C did not differ depending on amendment with xylose or 433 hemicellulose ( Supplementary Fig. S1). Amendment of these lower elevation soils with labile 434 nutrients had no significant effect on priming responses, with no significant differences 435 among control, N, P and NP treatments (both when nutrients were added alone and in 436 combination with C substrates). 437 438

19
Total incorporation of C substrates in microbial PLFAs varied significantly among soils but 440 not among C substrates or nutrient treatments (Supplementary Table S3). More substrate-C 441 was incorporated into microbial PLFAs in the upper montane forest soil (3025 m) compared 442 to the lowland forest soil (210 m; Fig. 4). The index of microbial substrate-CUE was also 443 greater in the upper montane forest soil (Fig. 4). The complexity of added C substrates had a 444 marginal effect (p = 0.06) on microbial substrate-CUE, whereby CUE of hemicellulose was 445 marginally greater compared to xylose in the lowland forest soil. In both upper montane and 446 lowland forest soils, fungi assimilated a greater proportion of xylose compared to 447 hemicellulose and GP bacteria assimilated a greater proportion of hemicellulose compared to 448 xylose (Fig. 5). In the upper montane forest soil, GN bacteria assimilated a greater proportion 449 of xylose compared to hemicellulose, whereas in the lowland forest soil assimilation of both 450 C substrates by GN bacteria was low, with no significant difference in the relative proportion 451 of xylose and hemicellulose assimilated (Fig. 5). 452 Nutrient treatments had no significant effect on the total incorporation of C substrates 453 in microbial PLFAs, nor the relative proportion of substrate-C assimilated by different 454 functional groups (Supplementary Table S3). However, nutrient treatments did influence 455 microbial assimilation of excess soil-derived C in some cases (Fig. 6). In the upper montane 456 forest soil, GP bacteria incorporated greater amounts of soil-derived C when amended with 457 xylose alone, with incorporation of soil-derived C by all microbial functional groups reduced 458 when xylose was added in combination with N (Fig. 6a). Microbial incorporation of soil-459 derived C was also reduced following amendment of upper montane soils with hemicellulose, 460 but most strongly in response to hemicellulose alone, and to a lesser degree when 461 hemicellulose was added in combination with N or P (Fig. 6b). In the case of the lowland 462 forest soil, xylose typically had a negligible influence on the incorporation of soil-derived C 463 in PLFAs (Fig. 6c), while amendment with hemicellulose marginally increased assimilation  Substrate respired CO 2 -C was unaffected by nutrient additions for all of the soils (Fig. 2); a 470 finding which was in broad contrast to other studies. While N addition has been shown to 471 suppress the decomposition of lignin-rich plant material (Carreiro et al., 2000;Knorr et al., 472 2005), other studies have often found decomposition of plant material or added C substrates 473 to be greater when inorganic nutrients were added, indicating nutrient constraints to 474 decomposition. This includes patterns of increased decomposition with N addition in some 475 temperate (Knorr et al., 2005;Manning et al., 2008;Vivanco and Austin, 2011) and tropical 476 montane ecosystems (Hobbie, 2000), and with P addition in tropical lowland forests 477 (Cleveland et al., 2002;Nottingham 2012Nottingham , 2015cChen et al., 2016). The differing outcome 478 in our study indicates that the decomposition of added C substrates was not limited by N or P. 479 This lack of effect of N or P on the turnover of added substrates may reflect that substrate 480 metabolism was constrained by the availability of other nutrients (e.g. potassium; Kaspari et 481 al., 2008), or may be a consequence of the low soil pH across all our study sites (Table 2), 482 which can lead to increased sorption of P to minerals (Olander and Vitousek, 2004). 483 Consistent with this, it has been shown for a range of tropical forest soils that the greatest 484 effect of nutrients in enhancing the turnover of added C substrates occurred when soil pH was 485 neutral to mildly acidic (Nottingham et al., 2015c). Alternatively, it is possible that soil 486 microorganisms in our study used the labile substrate-C in preference to more recalcitrant 487 soil-C, without nutrient constraint. Accordingly, even in the absence of added N or P, 488 microorganisms may have been able to acquire sufficient nutrients from the soil to mineralise 489 M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
the relatively low concentration of added C, with nutrient availability instead influencing the 490 mineralisation of native SOM in some cases (Fig. 3 a-b). This dichotomy of different nutrient 491 constraints to added labile C and SOM-derived C has been shown elsewhere in tropical forest 492 (Nottingham et al., 2015c) and temperate grassland soils (Carrillo et al., 2017). 493 Nitrogen addition had a large influence on SOM decomposition in upper montane 494 forest and grassland soils. In these soils, amendment with C substrates alone had a relatively 495 minor influence on priming of SOM, but addition of N (with or without P) significantly 496 reduced the release of SOM-derived CO 2 (Fig. 3 a-b). Negative priming responses to N have 497 also been widely reported elsewhere (Bird et al., 2011;Poeplau et al., 2016;Tian et al, 2016), 498 with reduced mineralisation of SOM under elevated N availability attributed to preferential 499 use of the exogenous N source (Craine et al., 2007;Dijkstra et al., 2013;Chen et al., 2014). 500 Previous studies of this elevation gradient in Peru have identified high microbial demand for 501 N at higher elevations, because N addition increased decomposition (Fisher et al., 2013), and 502 relative investment into N-degrading enzymes increased at higher elevations (Nottingham et 503 al., 2015b). Our findings further suggest that the turnover of organic matter in these montane 504 soils is regulated by microbial demand for N. In these soils, which are high in organic N but 505 low in available NH 4 +NO 3 (Table 2; Nottingham et al., 2015b), there appears to be a high 506 degree of baseline microbial N-mining from SOM, such that in the absence of external 507 nutrients, microbial demand is met by mineralising SOM to acquire N. Consequently, the 508 increased availability of inorganic N following the nutrient addition in our study may have 509 alleviated microbial demand for N and therefore reduced microbial N-mining, resulting in 510 reduced SOM mineralisation (negative priming). 511 Carbon substrate addition in lower montane and lowland forest soils induced positive 512 priming, which was greatest in response to xylose (Fig. 3 c-d). Consistent with this, inputs of 513 labile C were previously found to induce positive priming in tropical lowland soils from Peru 514 M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
22 (Whitaker et al., 2014a) and Panama (Nottingham et al., 2012;2015c). Amendment of 515 lowland soils with C substrates in combination with P and/or N in our study, however, had no 516 significant additional influence (Fig. 3 c-d). This suggests that priming of SOM in these soils 517 was not determined by microbial demand for P (or N), despite the very low availability of 518 inorganic P in the lowland forest soil ( Table 2). The lack of a nutrient effect in reducing 519 priming in lowland soils could be due to rapid immobilisation (high sorption) of added 520 nutrients in low pH soils, limiting their availability to microbes (Olander and Vitousek, 521 2004), with low pH potentially also affecting related enzyme function (Frankenberger and 522 Johanson, 1982;Sinsabaugh et al., 2008) and microbial community composition (Rousk et 523 al., 2010). The lack of a P effect on priming in particular may reflect the numerous pathways 524 for microbial P acquisition which do not destabilise C (Dijkstra et al., 2013), including P 525 mineralisation by biochemical hydrolysis (McGill and Cole, 1981), and P acquisition from 526 inorganic sources (Walker and Syers, 1976). Although there is evidence that priming due to 527 coupled acquisition of C and P can occur (Nottingham et al., 2012;Spohn et al., 2013;Meyer 528 et al., 2018), it may be less common given the opportunities for P acquisition from other 529 sources. For example, in 11 lowland tropical forest soils where C-metabolism was 530 predominantly P-limited, N rather than P was the predominant constraint on priming of SOM 531 (Nottingham et al., 2015c). Similarly, there was no relationship between P availability and 532 priming responses across a substrate-age gradient in tropical forest (Sullivan and Hart, 2013). 533 In our study, therefore, the positive priming responses in lower montane and lowland forest 534 soils following C amendment may have occurred due to a constraint other than N or P, or by 535 a mechanism other than nutrient mining (e.g. due to stimulation of specific groups of 536 recalcitrant C degrading microorganisms; Fontaine et al., 2003;Pascault et al., 2013;Su et 537 al., 2017). 538 The direction of priming responses to C-amendment observed in our study were 539 generally consistent with those identified previously for the same Peruvian gradient 540 (Whitaker et al., 2014a). However, previously, the nitrogenous C-substrate glycine elicited 541 strong positive priming effects in both lowland and montane forest soils. By contrast, 542 addition of C substrates in combination with N (as NH 4 -NO 3 ) here reduced mineralisation of 543 pre-existing SOM in higher elevation soils (Fig. 3 a-b) and had no effect in soils from lower 544 elevations (Fig. 3 c-d). These differences may be attributed to the different sources by which 545 C and N were supplied. Amendment of soils with trace amounts of amino acids such as 546 glycine has been shown to trigger activation of soil microorganisms (De Nobili et al., 2001, 547 Mondini et al., 2006 and, as such, glycine may have stimulated microbial metabolism, 548 resulting in strong priming of SOM (Mason- Jones and Kuzyakov, 2017). Whereas here, 549 microorganisms in the higher-elevation soils appeared to use the exogenous supply of N 550 instead of mineralising SOM to acquire N (resulting in negative priming). Differences in the 551 stoichiometry of C and N inputs may have also influenced the direction of priming responses 552 (Qiao et al., 2016). Thus, the direction of priming may be sensitive to the chemical 553 composition or stoichiometry of N inputs, which merits further investigation. 554 555

Microbial assimilation of 13 C substrates and excess soil-derived C 556
Nutrient treatments had no significant influence on microbial incorporation of C substrates, 557 however the complexity of C substrates was an important determinant of assimilation by 558 different microbial groups (Fig. 5). Gram-negative bacteria are most often associated with the 559 mineralisation of labile C, while gram-positive bacteria are thought to target more complex C 560 sources Kramer and Gleixner, 2008), in broad agreement with our 561 findings (Fig. 5). Although fungi are often associated with the degradation of more complex 562 C-compounds (Cusack et al., 2011a;Müller et al., 2017), some fungal taxa have been shown 563 M A N U S C R I P T

A C C E P T E D ACCEPTED MANUSCRIPT
24 to use more labile C sources (Hanson et al., 2008;Lemanski and Scheu, 2014). This was 564 evident here, where fungi assimilated a greater proportion of the simple C substrate, 565 particularly in the lowland forest soil (Fig. 5). Our results demonstrate that C use within the 566 microbial communities under study is more strongly shaped by the chemical complexity of 567 organic matter and the functional capacity of microbial taxa, rather than the availability of 568 nutrients. Nutrient availability may instead play a role in shaping the composition of the 569 microbial community, as some studies have reported microbial community shifts in response 570 to long-term fertilisation (Güsewell and Gessner, 2009, Liu et al., 2013, Fanin et al., 2015a. 571 While there was no change in microbial composition 7 days after nutrient treatments were 572 added ( Supplementary Fig. S2), the duration of our study may have been too short to observe 573 a response. Alternatively, the nutrient treatments may have been too small relative to the 574 inherent nutrient status of the soil to drive a compositional change, as a recent meta-analysis 575 of N addition studies revealed stronger microbial responses to fertilisation with increasing 576 study duration and nutrient addition-rates (Zhou et al., 2017). 577 In the upper montane forest soil, assimilation of soil-derived C by GP bacteria 578 increased following amendment with xylose, while assimilation of soil-derived C by all 579 microbial groups was reduced when xylose was added in combination with N (Fig. 6a). 580 These findings are consistent with the priming responses identified from the change in CO 2 581 production ( Fig. 3b), further supporting reduced mining of N from SOM when N was 582 externally supplied. Some studies have associated fungi with N-mining (Fontaine et al., 2011;583 Rousk et al., 2016;Soares et al., 2017), while in another study, K-strategists (i.e. 584 microorganisms associated with a slow but more efficient growth strategy) were identified as 585 responsible for priming of SOM to acquire N (Chen et al., 2014). Although the microbial 586 functional divisions used here do not directly correspond to the spectrum of r-K-growth 587 strategies, GP bacteria and fungi are often recognised as K-strategists, with slow growth rates 588 M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 25 and capacity to use more recalcitrant C-sources , Dungait et al., 2013. 589 Results from our study also suggest that, when supplied with labile C, GP bacteria increased 590 mineralisation of SOM, to mine N. Incorporation of excess soil-derived C by fungi in 591 response to xylose alone was, however, negligible, suggesting that fungi were not associated 592 with positive priming of SOM for N-acquisition in these soils. 593 In the lowland forest soil, amendment with hemicellulose tended to increase 594 assimilation of soil-derived C in microbial PLFAs, particularly into GP bacteria (Fig. 6d), 595 consistent with the small positive priming response identified from CO 2 production (Fig 3d). 596 Amendment of these lowland soils with xylose, however, had a negligible influence on 597 excess soil-derived C by all microbial groups (Fig 6c). This may seem surprising, given the 598 magnitude of positive priming induced by the xylose treatment (Fig. 3d). However, the index 599 of microbial substrate-CUE determined for these lowland forest soils was low (Fig. 4b), 600 favouring respiration of mineralised substrate-C rather than assimilation into microbial 601 PLFAs (Manzoni et al., 2012). Given this, it may be expected that primed soil-C mineralised 602 by these microorganisms would be mostly respired or invested in the synthesis of 603 extracellular enzymes, with a very low proportion incorporated into PLFAs for growth, 604 perhaps explaining why no increase in soil-C assimilation was detected here. 605 606

Conclusion 607
The mineralisation of simple and more complex C substrates was not affected by nutrient 608 availability in contrasting fertility soils from a 3400 m tropical elevation gradient. However, 609 the direction and intensity of priming responses varied among soils and was in some cases 610 dependent on the availability of N. Strong negative priming responses in upper montane 611 forest and grassland soils to exogenously supplied N suggest that turnover of SOM in these 612 soils was regulated by microbial demand for N, resulting in reduced mineralisation of SOM M A N U S C R I P T A C C E P T E D ACCEPTED MANUSCRIPT 26 when N was externally supplied. Whereas, microbial activity in lower montane and lowland 614 forest soils was not influenced by N or P additions. While inferences based upon short-term 615 laboratory experiments must be drawn with caution, evidence from long-term experimental 616 studies elsewhere suggest that both positive and negative priming responses to altered C and 617 nutrient supply can persist with time (Martikainen et al., 1989;Sayer et al., 2011). Long-term 618 N fertilisation of tropical montane forests in Puerto Rico (Cusack et al., 2011b), and a young 619 subtropical forest in China (Fan et al., 2014), has been shown to increase soil C stocks. As 620 such, our findings may provide a potential mechansim for this response, whereby inputs of N 621 could constrain the turnover of SOM, due to reduced mining of SOM for N acquisition. 622 Given that N deposition is projected to increase rapidly in the tropics over the coming 623 decades (Galloway et al., 2004), reduced N-mining could have a positive effect on soil C 624 sequestration in high elevation montane systems. By contrast, changes to nutrient supply in 625 tropical lowland forest soils may not influence the turnover of SOM. Instead, increased inputs 626 of labile plant-derived C, for example as a consequence of elevated atmospheric CO 2 (Cusack 627 et al., 2016) Table 1 Summary of site characteristics (Quesada et al., 2010, Rapp and Silman, 2012, Oliveras et al., 2014, Asner et al., 2017 and organic soil properties (reported as mean (1SE) where n = 4 from 4 subplots at each elevation site). Mean annual temperature; MAT, mean annual precipitation; MAP, water holding capacity; WHC.     following amendment with C substrates (xylose (X) and hemicellulose (H)) in combination with nutrient treatments (control, +N, +P, +NP). Data represent mean ± 1SE (n = 4). Note differences in y-axis scales among panels a-d.   • Tropical soils were amended with carbon substrates and inorganic nutrients • Nutrient treatments did not influence the mineralisation of added C substrates • Nitrogen addition reduced SOM mineralisation in higher elevation montane soils • Neither nitrogen or phosphorus influenced SOM mineralisation in lowland soils • C substrate complexity strongly affected C-use within the microbial community