Introduction

The activity of soil microbial communities plays a crucial role in the nutrient cycling of tropical lowland forests. Seasonality and subsequent variation in litter input affect soil microbes who are both consumers and suppliers, i.e., sink and source, of available nutrients in soil and ecosystem carbon (C), nitrogen (N) and phosphorus (P) cycling (Singh et al. 1989; Cavicchioli et al. 2019). Heterotrophic soil microbes depend on the supply of both labile and complex organic substrates from plants as their main energy (C) source (Soong et al. 2020). Nevertheless, soil microbial communities also depend on N and P for the synthesis of essential components—e.g., N for protein synthesis and P for DNA and energy transport and storage. Both plants and microbes need available forms of nutrients for uptake, which are largely provided through the conversion of more complex organic substrates to bioavailable products by breaking down larger polymers, a process catalyzed by extracellular enzymes (EE) (Skujiņš and Burns 1976; Baldrian 2009; Burns et al. 2013; Luo et al. 2017). Depolymerization of larger molecules by EE has often been considered the rate-limiting step in organic matter decomposition (Sinsabaugh and Follstad Shah 2012), and thus an important determinant of C and nutrient cycling potential in soil. Increased insight into the effects of seasonality on nutrient cycling may inform biogeochemical models.

Soil enzyme assays provide potential activities of enzymes, generally acting on the chain ends of polysaccharides, chitin and organic P, each specific substrate responsible for the rate limiting step in C, N and P decomposition. Commonly assayed enzymes include β-glucosidase, N-acetyl glucosamidase and phosphatase (liberating C, N, and P, respectively) (German et al. 2011). EE production depends on nutrient availability, and follows principles of resource or substrate supply and demand (Allison et al. 2011). Plant litter and soil nutrient contents can be used to characterize organic matter quality and to predict its respective turnover, with higher quality material (i.e. higher nutrient contents, lower molecular complexity) being turned over faster compared to more complex or lower quality organic matter (Zechmeister-Boltenstern et al. 2015). Stoichiometry of EE is used to assess nutrient limitations to microbial requirements (Sinsabaugh and Follstad Shah 2012; Moorhead et al. 2016). While N is considered to be the main limiting nutrient at higher latitudes, P-limitation is a prevalent characteristic in highly weathered tropical soils (Camenzind et al. 2018). Consequently, when compared to temperate ecosystems, tropical soil EE stoichiometries show high investments in phosphatases relative to enzymes targeting C and N (Waring et al. 2014).

Temperature and soil moisture may affect the activities of enzymes directly through reaction rates of EE (Nottingham et al. 2016) and EE and substrate diffusion within the soil. Temperature and moisture may also indirectly affect EE activities by affecting soil microbial community composition (Malik and Bouskill 2022). As a consequence, seasonal fluctuations in abiotic factors and litter input may affect soil microbial communities and their ability to take up or mineralize nutrients. There is evidence that litterfall and soil microbial biomass vary asynchronously, in association with seasonal shifts in nutrient availability in the wet tropics (Ruan et al. 2004). Seasonal fluctuations in various tropical ecosystem processes, such as the production of plant litter (Wu et al. 2016) and fine roots (Cordeiro et al. 2020) indicate seasonality in the cycling of C, N and P. Increasing our understanding of associated EE dynamics may provide insight into essential processes for sustained ecosystem functioning under future climatic conditions.

In this study we studied the effect of precipitation, temperature, litterfall and soil water content on microbial EE activities over the course of a seasonal cycle. Furthermore, we investigated seasonal dynamics of EE activities associated to C, N and P cycles in a tropical forest soil and used them as a proxy for soil microbial activity and nutrient demand. We hypothesized that: (1) litter inputs are the main driver of enzymatic activities, as opposed to temperature, precipitation, or soil water content; (2) total soil C and available C, N and P are negatively related to BG, NAG and AP activities, respectively (through increased microbial investments driven by low nutrient availability); and, (3) the P-related EE activities are higher than C and N related EE, due to low P availability in tropical soils.

Methods

Site description and sampling strategy

The study was carried out at the AmazonFACE experimental site (2°35′40′′S 60°12′29′′W) in Central Amazonia (more info on https://amazonface.inpa.gov.br/), approximately 70 km north of Manaus, Brazil, in the “Cuieiras” experimental reserve (Estação Experimental de Silvicultura Tropical (EEST), see also Pereira et al. 2019), which is also the base for the LBA-K34 tower and several experimental observation stations. The area is characterized by pristine old-growth tropical forests locally known as “Terra Firme” forests. These forests are situated on plateaus covered with nutrient poor and clay-rich soils classified as Geric Ferralsols. Soil texture consists on average of 68% clay, 20% sand and 12% silt and soil pH is on average 3.94 (Quesada et al. 2010). The mildly seasonal climate is defined by average annual rainfall of about 2400 mm, with a relatively drier period from June to November (months with at least 40% of days with < 3 mm precipitation), while the average temperature fluctuates from 25.8 °C in April to 27.9 °C in September (Araújo et al. 2002).

Sample collection and processing

Soils were collected from 18 sampling points. At 6 locations along a 400 m north-south transect (every 80 m), we sampled three points in the east-west direction, with 10 m distance between the three sampling points. The sampling scheme was adopted to consistently sample soils close to the AmazonFACE plots (for details, see Lapola and Norby 2014), without disturbing soil within the plots. Soils were sampled monthly between February 2016 and January 2017, using a custom-made steel soil corer (ø 10 cm). Soils were sampled at 0–5 cm and 5–15 cm depth and transported to the lab for sieving (2 mm), root and detritus removal and further processing.

Part of the samples were stored after drying at 65 °C for 48 h until further analysis, while fresh soil was used for selected measurements within 3 days of sampling. Soil enzymes were analyzed monthly for each of the sampling locations in fresh soil. Total soil P, extractable organic carbon, extractable nitrogen, and microbial biomass were analyzed every 3 months. Total soil C and N contents were determined monthly in composite samples consisting of the three east-west samples collected at each location along the north-south transect. Apart from the total C and N contents, all analyses were performed at the LTSP laboratório temático de solos e plantas (LTSP) laboratory at Instituto Nacional de Pesquisas da Amazônia (INPA) in Manaus, Brazil, nationally certified by Embrapa Soils (2016 Fertility Laboratory Quality Analysis Program, PAQLF, https://www.embrapa.br/en/solos/paqlf) and by the PIATV (Esalq/USP) inter-laboratorial program of vegetation tissue analysis (Grade A, http://piatv.com.br/). Litterfall was collected biweekly at two of the AmazonFACE plots located along the transect (used in this study) starting in August 2015. Litter traps (0.5 × 0.5 m, n = 24) were installed 1 m above the ground, 12 traps per plot in a circular pattern. The total litter was dried, separated into leaf litter and other litter fractions, and weighed.

Total C, N and P

Total soil C and N were determined in milled dry aliquots by an EA (IRMS). Total P was determined in dry (unmilled) 0.5 g aliquots with the molybdate blue method (Murphy and Riley 1962) after acid digestion using concentrated sulphuric acid solution (H2SO4, 18 M) followed by H2O2 (Quesada et al. 2010; see also Schaap et al. 2021).

Extractable C, N and P

Extractable organic carbon (eoC) and extractable nitrogen (eN) were obtained from extracts of 2 g of fresh soil in 20 ml 1 M KCl solution, shaken for one hour and subsequently filtered. The filtered extract was then analyzed in a TOC/TN analyzer (TOC-V CPH E200V⁄TNM-1 220 V; Shimadzu, Vienna, Austria). Extractable P (Olsen P, Olsen et al. 1954) was determined from extractants of 2 g of soil in 20 ml 0.5 M bicarbonate solution (NaHCO3, pH 8.5), shaken for one hour and filtered. Extractant was analyzed following the photometrical Murphy-Riley molybdate blue method (712 nm) (Murphy and Riley 1962). All analyses were accompanied by method blanks (no soil) to account for contamination or background signal, and lab variation was accounted for by analyzing standards during each batch of photometric extract reading.

Potential soil extracellular enzyme activities

Potential EE activities of three common hydrolytic enzymes relevant to C, N and P cycling were assayed using a fluorescence method based on Marx et al. (2001) and German et al. (2011). 4-Methylumbelliferyl β-D-glucopyranoside (M3633 Sigma, substrate concentration 200 µM), 4-Methylumbelliferyl N-acetyl-β-D-glucosaminide (M2133, substrate concentration 200 µM) and 4-methylumbelliferyl phosphate disodium salt (M8168 Sigma, substrate concentration 1 mM) were used as substrates for β-glucosidase (BG), N-acetyl glucosaminidase (NAG) and acid phosphatase (AP), respectively. All are widely used in soil enzyme assays as they can be considered a proxy for microbial demand of C, N and P. Substrate concentrations were established in preliminary experiments to ensure reaction rates at substrate saturation (and thus Vmax). 4-methylumbelliferone standards (M1381 Sigma) were used and substrate controls, sample controls and blanks were measured to control any background signal. All enzymes were assayed in soil slurries of 0.5 g of fresh soil dissolved in 50 ml sodium acetate buffer (pH 5.5) and vortexed for one minute before pipetting aliquots in a microplate (96 well polystyrene, black flat bottom). 200 µl soil slurry was used with 50 µl substrate. Microplates were incubated for 1 h at 23 °C, then fluorescence measurements were performed with an Infinite F200 Pro plate reader (Tecan Austria GMBH, Grödig, Austria), with fluorescence intensity measured from the top (λexcitation = 360 and λemission = 440 nm).

Quantitative analyses

Extracellular enzymatic stoichiometry (EES) and vectors were calculated according to Moorhead et al. (2016). Enzyme activity ratios and proportional activities were calculated using the natural logarithm. The enzyme and nutrient ratios for C:N, C:P and N:P were calculated in each sample as with ln transformed ratios (e.g., ln(C:N), ln(BG:NAG), etc.), while proportional ratios were calculated as

$${C:N}_{proportional}=\text{ln}\frac{BG}{BG+NAG}$$

and

$${C:P}_{proportional}=\text{ln}\frac{BG}{BG+AP}$$

Vectors were calculated using both of those proportional ratios, their length as

$$Vector length= \sqrt{{{C:P}_{proportional}}^{2}+{{C:N}_{proportional}}^{2}}$$

and their angle in degrees as

$$Vector angle= {\text{tan}}^{-1}\left(\frac{{C:N}_{proportional}}{{C:P}_{proportional}}\right)$$

Means were calculated per sampling date (n = 18, at two depths) according to the stoichiometric mean recommended by Isles (2020) as the mean of each natural logarithm, all values are reported ± their standard error. Data processing and statistical tests were performed in R 4.2.1 (R Core Team 2022).

We used linear regression models to assess direct relations between precipitation, temperature and litterfall. Linear mixed-effect models were applied to assess relations between enzymes and the other variables using the “lme” function from the “nlme” package (version 3.1–157, Pinheiro et al. 2022); sampling location was included as random effect, with data ln transformed for normality where indicated. For the linear mixed-effect models shown in the graphs, only one fixed effect was included per model, for the reported mixed-effect models in the table the depth was included as a fixed effect as well. Additionally, for the models shown in the table the “varIdent” variance structure was used to allow for different variances per stratum (sampling depth) and additionally the model residuals were checked for autocorrelation (no significant temporal autocorrelation was found). All models’ residuals were checked for homogeneity and normality. Conditional R2 values for the linear mixed-effect models shown in graphs were obtained with the “r.squaredGLMM” function from the “MuMIn” package (version 1.46.0, Bartoń 2022) (Fig. 1).

Fig. 1
figure 1

a daily average air temperature (at 34.6 m, above forest canopy), and b rainfall at the AmazonFACE plots, daily average (line and dots ± SE) and the monthly sum (bars). Both temperature and precipitation were measured every 30 min (n = 48 per day) and calculated per day. The “Drier season” bracket indicates which months are treated as the drier months of the year in the manuscript, defined as the months with at least 40% of days with < 3 mm precipitation

Full size image

Results

Precipitation showed a distinct drier period (at least 40% of days with < 3 mm, Fig. 1b) between July and November, during which it was also a few degrees warmer, but average temperature varied little and stayed within a 24.5–27.5 °C range (daily average) (Figs. 1a, S1a). Annual leaf litterfall amounted to 5565 ± 55 kg ha− 1 year− 1, with a distinct peak during the drier months (Fig. 2a). Leaf litterfall was significantly correlated to the average monthly temperature (F(1;9) = 5.3, p = 0.047, Fig. S1b) and showed a negative relation with the average rainfall (F(1;9) = 42, p < 0.001, Fig. 2b) indicating higher leaf litterfall during the drier months. Soil water content showed limited variation (Fig. S2), and had no significant relation to either precipitation, temperature or litterfall.

Fig. 2
figure 2

a leaf litter collected (biweekly) at the AmazonFACE plots, recalculated for average daily litter quantity (average per month, each observation represents a different littertrap, n = 24. Boxplot shows median and quartiles), and b the relation between average daily leaf litterfall and average daily precipitation (natural logarithm) per month. Drier season indicates the drier months as established in Fig. 1

Full size image

Total soil C at 0–5 cm was on average 5.53 ± 0.02%, with the lowest value of 4.19 ± 0.09% in January to highest value of 7.28 ± 0.27% in May (Fig. S3). Total soil N was 0.35 ± 0.00% on average, following roughly the same pattern as total C. Total P averaged 156.39 ± 0.69 mg kg− 1, ranging from 141.8 ± 2.13 mg kg− 1 in August to 204.52 ± 3.46 mg kg− 1 in February, with the note that the measurement frequency of total P was lower than for C and N (Fig. S3). For 5–15 cm, the average total C (2.84 ± 0.01%), total N (0.21 ± 0.00%), and P contents (118.22 ± 0.52 mg kg− 1) were lower as compared to the top 5 cm but followed the same temporal trend as in the top 5 cm (see Fig. S3). The eoC, eN and Olsen P in the top 5 cm of soil were 1034.0 ± 6.2 µg C g− 1 dry soil, 101.41 ± 0.34 µg N g− 1 dry soil and 2.08 ± 0.01 µg P g− 1 dry soil. At 5–15 cm those values were lower with 916.86 ± 6.41 mg C kg− 1 soil, 76.78 ± 0.24 mg N kg− 1 and 1.19 ± 0.00 mg P kg− 1 dry soil, respectively (Fig. S4). At both soil depths, we found the highest average values for eoC in May (1806.6 ± 31.3 mg kg− 1 and 1727.6 ± 31.6 mg kg− 1 respectively), while the lowest average was measured in August (594.7 ± 6.0 mg kg− 1 and 455.7 ± 2.8 mg kg− 1 respectively). In contrast, the eN values were lowest in February at both depths (82.1 ± 1.54 mg kg− 1 and 58.24 ± 0.74 mg kg− 1 respectively), while reaching their highest values in August (135.3 ± 1.19 mg kg− 1 and 95.92 ± 0.58 mg kg− 1 respectively. Olsen P peaked in March at both depths, while in the top 5 cm showed the lowest concentration in April (1.16 ± 0.02 mg kg− 1), while in the lower soil increment the lowest value was reached in January (0.52 ± 0.02 mg kg− 1).

Fig. 3
figure 3

 C, N and P related extracellular enzyme activities (BG, NAG and AP per soil C) from February 2016 till January 2017, and their relation to the average monthly leaf litterfall. Boxplots are showing the median, the lower and upper hinges correspond to the first and third quartiles (plots a, c and e). The text in plots b, d and f shows the relation between enzymes and litterfall estimated with a linear mixed effects model, with the sampling location as a random effect

Full size image

Average EE activities (as expressed per gram soil C; for values per dry soil see Fig. S5) were 0.21 ± 0.00 µmol g C− 1 day− 1 for BG, 0.87 ± 0.00 µmol g C− 1 day− 1 for NAG and 20.21 ± 0.04 µmol g C− 1 day− 1 for AP, while at 5–15 cm those activities were 0.23 ± 0.00, 0.63 ± 0.00, and 26.26 ± 0.08 µmol g soil C− 1 day− 1 for BG, NAG and AP respectively (Fig. 3a, c, e). In the top 5 cm EE activity rates peaked just before and during drier season and were lowest in the wetter season, with BG showing highest rates in August (0.34 ± 0.02 µmol g C− 1 day− 1), and NAG and AP peaking in September (1.22 ± 0.06 µmol g C− 1 day− 1 and 44.61 ± 0.90 µmol g C− 1 day− 1 respectively) in January for BG and NAG (0.12 ± 0.00 µmol g C− 1 day− 1 and 0.20 ± 0.01 µmol g C− 1 day− 1 respectively), and in June for AP (15.52 ± 0.35 µmol g C− 1 day− 1). This pattern was reflected at 5–15 cm, but BG and NAG peaked just before the drier season (in June, 0.31 ± 0.01 µmol g C− 1 day− 1 and 1.37 ± 0.04 µmol g C− 1 day− 1 respectively) while AP peaked in September (31.91 ± 0.57 µmol g C− 1 day− 1). The lowest EE activities at 5–15 cm depth were all in January (BG 0.13 ± 0.00 µmol g C− 1 day− 1, NAG 0.40 ± 0.01 µmol g C− 1 day− 1 and AP 12.80 ± 0.19 µmol g C− 1 day− 1).

We applied linear mixed effect models to assess relationships between climatic factors (temperature, moisture), leaf litter inputs and soil enzyme activities, (Table 1, Fig. 3). We found that BG activities at both soil depths were significantly positively related to litterfall inputs (Fig. 3b), but not to temperature (Fig. S6a), and to precipitation only in the top 5 cm (Fig. S6b), while potential NAG activity rates were significantly related to only the litterfall (Figs. 3d, S6c, d). In contrast, potential AP activity was significantly related to all studied drivers, with litter and temperature having a positive, and precipitation a negative relationship (Figs. 3f, S6e, f). Based on linear mixed effect models including sampling depth as a fixed factor, we identified leaf litterfall as the strongest driver for potential EE rates while we found no significant effects of precipitation, temperature, and soil water content (except for NAG), respectively (Table 1).

Table 1 Analysis of variance F statistics, with p values in parentheses, for enzyme responses to either litterfall, precipitation, temperature, or soil water content, combined with sampling depth. Sampling location included as a random factor in all models. Asterisks (*) indicate a significant improvement over the null models AIC
Full size table

Although the EE activities were higher overall in the drier months from July to November, total soil nutrient contents did not seem to show a clear seasonal pattern (Fig. S3). To assess if nutrients provided ecological constraints on the microbial activity, we investigated relations between nutrients and enzymes. Although the total soil C, N, and P contents provided limited significant relations with EE (Fig. S7), the relations of enzymes with the extractable C, N, and P were dominated by the significant positive relations that eN and the enzyme activities showed overall (especially in the top 5 cm), while also showing significant negative relations between the AP activity and eoC and Olsen P (Fig. 4).

Fig. 4
figure 4

Relations between EE and extractable soil nutrients at 0–5 cm and 5–15 cm. Conditional R2, F and p values for linear mixed effect models with sampling location as a random effect. a, b, and c β-glucosidase as related to eoC (natural logarithm), eN and Olsen P (natural logarithm); d, e and f N-acetyl glucosamidase as related to eoC (natural logarithm), eN and Olsen P (natural logarithm); g, h, and i Phosphatase as related to eoC (natural logarithm), eN and Olsen P (natural logarithm)

Full size image

While the EE activities showed an observable drier-season effect when considered separately (Fig. 3), their activity ratios and proportional activity ratios showed less distinct patterns (Table S1). From these proportional activity ratios we calculated enzyme vectors, useful for distinguishing relative nutrient demand (Fig. 5). Although at a first glance the vectors indicate a persistent high demand of P, the vector angles decreased after the drier months, indicating a relative shift from P acquisition towards N acquisition enzymes during the rainy season; especially in June, just before the onset of the drier season, the relative N-demand was high. Moreover, the vector lengths peaked in June, indicating an increased C demand. The angles and the lengths of the vectors showed a weak yet significant negative relationship (Fig. S8), indicating a weak relation between relative C and N demand.

Fig. 5
figure 5

Average monthly vectors of proportional enzyme activities at a 0–5 cm and b 5–15 cm, and average vector properties c length (unitless), and d angle (in degrees) of the monthly average vectors. The error bar in c and d represents the standard error)

Full size image

Discussion

In this study, we report on the dynamics of EE in tropical soils, which highlights seasonal fluctuations in soil microbial nutrient demand and relative investments in respective EE. We found that EE activities follow the seasonal signal of leaf litterfall. Furthermore, we found a significant positive relationship between eN and most enzyme activities. BG and NAG activities were relatively low compared to AP, which may indicate a strong demand for P in these highly weathered tropical forest soils. Our results suggest that microbial resource limitation shifts from relatively more N-demand at the end of the rainy season, to increased P-demand during the drier season, indicating seasonal changes in the relative microbial investment in EE.

The reported leaf litterfall seems to be comparable to earlier studies in the area (Lucas et al. 1993; Luizao et al. 2004; Wu et al. 2016), with slightly lower annual litter production. Possibly, the lower observed litterfall was a consequence of relatively higher litterfall in association with an El Niño event observed in the preceding year (e.g. Hilker et al. 2014). Aboveground phenology and litterfall is well established to be seasonal in the tropics (Chave et al. 2010; Wu et al. 2017), and evidence is emerging that these patterns are reflected also in soil microbial communities, with more decomposers and anaerobic saprophytes present in the wetter season (Buscardo et al. 2018). The positive relationship between enzyme activity and litter inputs therefore also indicates a synchronization and a link between new substrate, microbial community changes and changing investments of microbes in enzymes.

The pattern of increased potential EE activity during the drier season was evident for all studied enzymes, and we hypothesized this was mainly driven by increases in litter inputs in these months. Litterfall showed significant (positive) relations to all enzyme activities, while precipitation showed weaker or insignificant (negative) relations towards EE, and temperature only showed consistent (positive) relations with AP activity. The pattern of increased potential EE activity in drier months has been observed by others as well; Smith et al. (2015) found increased EE activity during the dry season in a Puerto Rican subtropical forest, as did Singh et al. (2020) in a dry tropical ecosystem. They attributed these dry season increases in EE to reduced access of microbes to resources—reducing microbial assimilation and triggering enzyme production—and decreased enzyme turnover and clay-mineral interactions causing immobilization, all attributed to reduced soil moisture. We reported a similar pattern, yet soil water content did not seem to be a big constraint in our studied soil system, even in the drier season. Similarly, the observed temperature range was limited at our site, which could explain why we observed relations between temperature and AP activity, but not consistently to the other EE activities.

Even though our hypothesis of litter driving EE investments is supported by our analysis, it remains a challenge to untangle the effects of litter, precipitation, and to a lesser extent, temperature. Precipitation was strongly related to litterfall, indicating that indirectly or directly, precipitation drives changes in belowground biochemistry. Most likely, they are both determining soil EE expression through different mechanisms—with decreases in precipitation stimulating litterfall and thus substrate (mainly C), while soil moisture limits enzymatic mobility to some extent. Both could be reflected in increased potential activity; increased labile substrate would stimulate microbial enzyme production through increased return of investments from produced enzymes, yet decreased moisture would increase immobilization—the last one being more of a methodological artefact than something which would be reflected in in-situ turnover.

We hypothesized total soil nutrient contents to be negatively related to EE, however, we found weak relations between total nutrients and enzymes. Moreover, we expected the relation between the available nutrient contents and the related enzyme activity to be negative, since low nutrient availability would stimulate investments in acquisition, yet relations between eoC, eN, Olsen P, respectively and EE activities in the soil were not always significant. BG and eoC were not related, while eN and NAG even showed a positive relation in the top 5 cm. Only Olsen P and AP activity showed the expected relation, albeit weak, suggesting some degree of demand driven AP investments by microbes (Sinsabaugh and Follstad Shah 2012) with possible contributions from roots (Guilbeault-Mayers et al. 2020). However, as an alternative driver of EE activity, we found significant positive relations of eN with most enzyme activities, which suggests that soil microbial communities depend on a supply of eN to maintain EE production.

Fertilization studies in tropical areas indicate that N addition can stimulate organic matter turnover and EE activities (Marklein and Houlton 2012; Wang et al. 2018), although others did not find such effect (Turner and Wright 2014). Nitrogen fixation can be linked to the acquisition of P by AP production (Allison et al. 2006; Nasto et al. 2014). Our study indicates that this stimulation of EE by available N can be observed in short timespans as well, suggesting the production of enzymes is dependent on the available N-supply. This is of interest to the functioning of P-limited tropical forests, i.e. the link between N availability and AP production would imply N limitation on P acquisition. This suggests further investigation into the seasonal dynamics of the tropical N-cycle, especially in relation to precipitation, would be important to improve our understanding of microbial functioning and the P-cycle.

The EE activity rates at our site were in the same range as compared to enzyme activities measured in a tropical mountain rainforest (Tischer et al. 2014) and as reported along an altitudinal gradient in the Andes, albeit lower than on lowest altitudes reported (Nottingham et al. 2016). BG activity, liberating glucose (available C) as last step in the breakdown of cellulose, was low overall, yet with an observable increase during the drier season synchronized with litterfall. NAG and AP activity, used as a proxy for N and P demand, also showed increases during the drier season and synchronization with litterfall. Notably, NAG activity, breaking down chitin and possibly indicative of microbial turnover, increased just before the drier season (June); in contrast, AP activity showed a relative decrease in the same month.

The measured enzyme activities may shed light on microbial nutrient demand or allocation to resource acquisition. Our results suggest a relatively higher investment of AP compared to BG and NAG; which according to our third hypothesis would indicate that central-eastern Amazon forest soils are limited by P availability. Using EE activity vector analysis, as conceptualized by Moorhead (2016), we identified temporal changes in microbial nutrient demand that are likely related to the phenology of soil microbial biomass and activity. We found an increased N-demand at the end of the wet season (lower vector angle), which decreased during the drier season, in favor of the P-demand (larger vector angle). This was mainly driven by the different pattern of NAG activity compared to the other enzymes. NAG catalyzes the breakdown of chitin present in for example fungal cell walls, and a relatively higher activity could also indicate higher microbial turnover (Zeglin et al. 2013). This has been observed by Mori (2020; see also Mori et al. 2021) as well, who challenged the idea that low enzymatic C:N activity ratios (proportional activities in our case) reflect microbial N limitation if the dominant substrate is not cellulose. An alternative to the microbial limitation-hypothesis could be changes in substrate, such as a switch from more plant derived substrate to higher turnover of the microbial biomass, driving NAG-dynamics. Indeed, inputs of C-rich plant material to subtropical soils can shift the fungal community to N-limitation, while the bacterial community shifts towards a co-limitation by C and N (Rosinger et al. 2019).

During the drier season, P demand was relatively more pronounced (larger vector angle). Once leaf litter reaches the soil, there are different pathways for the incorporation of organic matter (SOM) into the soil, where the labile components are released first, and particulate recalcitrant matter is incorporated in later stages (Cotrufo et al. 2013, 2015). This time lag is a possible explanation for the trend indicated by the vectors towards more P-acquisition towards the end of the drier season; P-loss from litter does not occur immediately (Martins et al. 2021), which might cause a delay in enzymatic response (Schaap et al. 2021). However, we found no evidence of a significant lag effect between litter inputs and AP dynamics in our data (no significant autocorrelation of model residuals).

Vectors of proportional enzyme activities showed a relative increase in microbial P demand towards the end of the drier season and seem to indicate an increase of relative N-demand at the end of the rainy season—before litterfall increases. Changes in microbial biomass size a month prior to litter inputs have been reported for tropical forests, which were attributed to plant mediated shifts in belowground C and nutrient inputs and decreased nutrient uptake related to seasonal leaf senescence (Ruan et al. 2004). An equal mechanism could be at play here. A possible link that might explain our findings is that the litterfall is driven by the seasonality of precipitation, and consequently, N—required for enzyme production—is mobilized (into an accessible form, measured here as eN) rather quickly from substrate through increased activity of NAG, allowing for increased EE production in the drier season for all enzymes. This might also explain the comparably low AP activity before the drier season (June), and the relative dip in the vector angles—and thus a stronger N-demand. This suggests enzyme dynamics are mainly controlled by microbial access to available N, which is also in line with the observed relations between eN and EE activities.

In summary, our study shows that microbial activity is synchronized with litter seasonality, as shown by the relation between leaf litter and EE activities. Moreover, our results suggest available forms of nutrients (measured here as eoC, eN and Olsen P) in the mineral soil were taken up quickly (hours-days, e.g., Menge et al. 2009; Helfenstein et al. 2018) and did therefore not show strong relations to the corresponding enzymes in the monthly measurement intervals, yet available N facilitates enzymatic activity. Microbial activity showed a permanent high demand for P, although before the drier season, an increased N-demand was observed. We conclude that in the studied tropical forest ecosystem, soil nutrient availability is an important determinant of dynamic changes in EE mediated nutrient acquisition capacity, which is in turn related to plant phenology and climate seasonality. Our results suggest a supply of available N is paramount to EE activity to maintain microbial enzyme production, yet also demonstrate a persistently high P demand. Future research should untangle the temporal dependencies between nutrient cycles—such as the N and P cycle—to address the timing of constraints and limitations to microbial functioning under different biotic and abiotic conditions. This may increase insight into the response of nutritional cycles in tropical forests to shifts in seasonality, such as more prolonged and more severe dry seasons through climate change.