Inﬂuence of Two Important Leguminous Trees on Their Soil Microbiomes and Nitrogen Cycle Activities in a Primary and Recovering Secondary Forest in the Northern Zone of Costa Rica

: Inga edulis and Pentaclethra macroloba are dominant N-ﬁxing forest trees in Costa Rica, likely important for recovery of soil N and C after deforestation, yet little is known of their soil microbiomes nor how land use impacts them. Soils from both trees in a primary and secondary forest were assessed for N-cycle metrics and DNA sequence-based composition of total bacterial, potential N-ﬁxing bacterial, and potential ammonium oxidizing bacterial genera. The compositions of the functional groups of bacteria, but not their total relative abundance of DNA, were di ﬀ erent across the soils. The P. macroloba soils had greater NO 3 − levels and richness of both functional groups, while I. edulis soils had greater NH 4 + levels, consistent with its NH 4 + preference for root nodule development. The bacterial communities were di ﬀ erent by habitat, as secondary forest I. edulis microbiomes were less rich, more dominant, possibly more a ﬀ ected by the disturbance, or reached equilibrium status quicker than the richer, less dominant P. macroloba microbiomes, which may be developing slower along with secondary forest succession, or were less a ﬀ ected by the disturbance. Functional redundancy and switching of 10 N-cycle bacterial genera was evident between the primary and secondary forest soils, likely to maintain stable levels of N-cycle activity following disturbance. In summary, the two tree soil microbiomes are di ﬀ erent, land use di ﬀ erentially a ﬀ ects them, and, thus, both tree species should be used during forest regeneration strategies in this region.


Introduction
Deforestation for agricultural uses has long been occurring in tropical regions such as the Northern Zone of Costa [1,2], resulting in diminished rates and levels of activity of soil nitrogen (N) and carbon (C) cycle dynamics and related soil ecosystem conditions, and the colonization of scrub growth or invasive species rather than development healthy secondary forests [1]. The N-fixing soil microbe activities represent the principal pathway by which deforested tropical areas recuperate the soil N and C [3][4][5][6] that are depleted during deforestation and subsequent agricultural uses [7][8][9][10], and are Soil Syst. 2020, 4, 0065 2 of 17 also in high demand for the rapidly growing vegetation during reforestation [11]. To remediate these disturbed soils and forested areas, secondary forest regeneration following different types of land use damage is becoming common in these regions [2,[12][13][14]. It has also been recognized that leguminous trees and their soil microbiomes provide the principal pathways for soil N and C recuperation, and belowground biomass development during tropical forest restoration [3][4][5][6][15][16][17][18][19][20]. This recognition has led to the use of N-fixing trees as an important component in tropical forest regeneration and restoration strategies [3,16,18,19,21].
Currently, little is known of the ecological relationships between these leguminous trees, their soil ecosystems, or their soil N-cycle-associated microbiome communities. It is also unclear how land management or other disturbances and restoration activities impact the overall soil ecosystem conditions, the composition or recovery of the soil N-cycle microbial communities, or even how these impacts might influence rates of regrowth of tropical forests. This information could provide insights useful in maximizing the efficacy of tropical forest and soil restoration and recovery plans [22][23][24][25][26][27]. For example, some suggest that N-fixing trees appear to interfere with tropical tree regrowth trajectories, while others suggest they appear to enhance tropical tree regrowth trajectories [7,[28][29][30]. Certainly, these observed differences in forest regrowth trajectories could be linked to inhibition by N-fixing trees, or covarying components like stand age and N-fixing tree species composition [4,11,31,32], or past land-use histories [1,15,31,[33][34][35][36]. However, it is also possible that specific leguminous tree species are differentially influencing the soil microbiome community compositions such that they result in altered soil C and N cycle dynamics, sequestration and distribution, and rates of soil biomass development, all of which could certainly affect both the soil recovery and forest regrowth patterns [10,37,38]. For example, it is known that bacterial N-fixation is important for production of ammonium (NH 4 + ) and stimulation of the ammonium oxidizing bacteria (AMO) that produce nitrate (NO 3 − ), and that both NH 4 + and NO 3 − are critical for healthy forest and soil recovery. However, if different leguminous tree species have dissimilar soil microbiome community compositions, their N-cycle activities could be differentially regulated, resulting in variability in N cycle dynamics that could ultimately impact forest regrowth trajectories. For example, it is well known that N-fixing activities in different bacterial species are facultative and feedback-regulated, which results in variations of N-fixation rates not necessarily proportional to the abundance or species of either N-fixing tree or N-fixing bacteria within a reforested area [39,40]. Clearly, such differential regulation and activities of the N-cycle associated with the soil microbiome can alter the soil ecosystem and, thus, have an impact on tree regrowth patterns. However, as there is currently insufficient knowledge about the ecological relationships between different leguminous tree species soil N cycle-linked microbiome communities and associated N-cycle components, either between tree species or along restoration gradients following land disturbance or management practices, a study was conducted on the N-fixing and AMO bacterial communities and N-cycle components in the soils of two species of leguminous trees in a primary forest and a regenerating secondary forest in the Northern Zone of Costa Rica. Pentaclethra macroloba (Willd.) Kuntze (Fabaceae) and Inga edulis (Sw.) Willd. (Fabaceae) are two common leguminous tree species that are early colonizers and dominant later successional stage trees, thought to be important for N resource inputs in the Northern Zone forests of Costa Rica, [19,21,[41][42][43] and as such have been used in tropical reforestation practices, although I. edulis has been more commonly used [19,21,42]. The current study was the first to characterize the compositions of the total bacterial assemblage and N-cycle-associated bacterial microbiomes and the N components within soils of these two tree species and to identify whether any differences in these metrics could be found between the different species or between the species under different land use histories. Thus, we examined the differences in the composition of the general bacterial, N-fixing, and AMO bacterial soil communities and N components from soils of I. edulis and P. macroloba trees within a recovering 23-year old secondary forest compared to a primary forest within the Northern Zone of Costa Rica. The goals were to determine: (a) if differences existed between the two plant species soil microbiomes; (b) if these microbiomes showed evidence of being differentially influenced by the land management practice used; and (c) if there was evidence that one species might better serve for reforestation purposes. To address the goals, two questions were asked:

1.
Are there differences in composition of the communities of total bacterial genera and bacterial genera associated with potential N-Fixing or AMO activities between soils of P.  [41]. The I. edulis and P. macroloba trees used were from Upland Primary Forest sites not previously harvested, and Upland Secondary Forest sites that were part of the same primary Uplands forests, but cleared in 1982, grazed for 10 years, then abandoned and allowed to naturally regenerate since 1992. The different forest sites used were 100 m to 1 km apart, with the same oxisol soil types, no significant differences in pH, soil texture structure, % water holding capacity, density, or topography, and with slopes of 0% to 10% [26,27].
Six trees of each species were identified from the Primary Forest sites and six of each species from the Secondary Forest sites that were at least 50 m apart, with no overlapping canopy cover. All trees chosen were considered to be actively growing adult trees, with diameter at breast height (DBH) values between about 35% to 50% of the maximum for the species in the region (for P. macroloba = 24-47 cm, for I. edulis = 10-19 cm), which ensured active root nodule functionality based on previous studies [44][45][46][47]. Tree size (i.e., DBH) is commonly used as a proxy for tree age in this region of Costa Rica, as annual tree rings are inconsistently and minimally evident, at best, due to there being no strong dry season-wet season moisture variations [48,49]. The use of tree size correlated with the previous literature on the two species allowed us to approximate a relatively equal tree age for the tree soils.
The Tree Protection Zone or TPZ method [26], recently used in these same forests [26], was used to identify soil sample locations around tree bases as it better represents the critical root area associated with tree health and vigor, while the drip-line method often underestimates this important area [50,51]. Four soil cores (7.5 cm wide × 15 cm deep × 1.25 cm thick) per tree were aseptically collected (70% ethanol disinfection of gloves and tools between trees) at approximately 15 cm depth, at the 4 cardinal points around the tree bases, and at 10% of the calculated TPZ, after which the 4 cores were pooled into a single tree soil sample, in June 2015. All soil samples were passed through sterilized 4-mm sieve at field moist conditions prior to any analysis.

Soil Nitrogen Properties
Standing pools of soil N-cycle metrics were chosen as potential indicators of microbial group activity. Standing pools of nutrients are thought to represent more long-term effects on the soil biota due to land management compared to fluxes, which tend to represent rapid nutrient turnover rates of nutrients that are more associated with short-term changes in the soil biota and do not as effectively reflect land management-induced changes [7,50,52,53] levels were used as indicators of soil NH 4 + production, such as from N-fixation or organic N compound decomposition.

DNA Extraction, Sequencing, and Bioinformatics
Total soil community environmental DNA (eDNA) was extracted from three 0.33g replicate sub-samples for a total of 1g for each soil sample using the MoBio PowerSoil DNA Isolation Kit (MO BIO Laboratories Inc., Carlsbad, CA, USA). The concentration and purity (A 260 /A 280 ratio) of soil eDNA were then determined using a NanoDrop 1000 spectrophotometer (ThermoFisher Scientific, Waltham, MA, USA). All the amplification and sequencing methods have been previously described in detail [26,55]. Briefly, the different soil sample eDNA extracts were used for a 2-step PCR amplification, targeting the v3 and v4 of 16S ribosomal RNA gene region for bacteria and archaea [56]. All generated soil amplicons were sequenced in several Illumina MiSeq runs using a V3 MiSeq sequencing kit (FC-131-1002 and MS-102-3003). The subsequent 16S Illumina-generated sequences were processed using semiautomated pipelines, producing operational taxonomic units (OTUs). These OTUs were processed and taxonomically assigned from Phylum to Genus using the Ribosomal Database Project (RDP) classifier v2.12 [57] for bacteria using a standard QIIME protocol [56] with in-house Python data management scripts. Input reads were processed through QIIME's open reference clustering with standard parameters and 97% similarity to the GreenGenes database for bacterial analysis and assigned to known taxonomy where possible using the UCLUST algorithm. The identified soil bacterial DNA-based OTUs were organized taxonomically down to the genus rank.
This process resulted in 6,796,607 bacterial DNA sequences that were successfully identified at the phylum (24 phyla), class (36 classes), order (78 orders) and family (147 families) levels, with 1,123,988 of these sequences clearly identified to the level of 296 genera (hereafter called total bacterial genera). For this project, we chose to analyze the differences in the composition of the various soil bacterial communities at the genus level in the soil microbiome by Tree Species and by Land Management. The variable library sizes for each soil sample per habitat were normalized by converting each specific sequence (or OTU) read number to the mean proportion of the sequences (MPS) per sample [58], as the number of sequences from each within each soil sample, divided by the total number of sequences per that sample.
In microbial ecology studies, one can determine the relative amount of a functional gene in an eDNA sample through qPCR studies based on amplification of that functional gene sequence, but this does not concretely connect that specific sequence of the functional gene and its abundance to a specific microbial taxon, as identification of the gene is based on the amplification of that gene DNA sequence while identification of the taxon is generally based on amplification of variable regions of microbial rRNA. However, reference-based associations between certain microbial groups and functional genes have been used to connect functional genes with specific microbial genera or species [59,60]. Such efforts have resulted in valuable databases that, while not exhaustive nor said to be 100% accurate, provide a means of linking specific microbial taxa with potential functional activities. We took an analogous approach [61], using these same databases (www.genome.jp/kegg/; http://rdp.cme.msu.edu/hierarchy/; www.zoology.ubc.ca/louca/FAPROTAX/lib/php/index.php) and additional literature sources (references presented in Table S1) to suggest that specific bacterial genera have a strong possibility of having the potential for the functional group activities of nitrogen fixation or ammonium oxidation. These groups were called N-Fix pot and AMO pot , respectively; there were 41 potential N-Fixers bacterial genera (N-Fix pot ), and 16 potential AMO bacterial genera (AMO pot ) identified (Table S1). We only included genera that were clearly considered to commonly perform these functions. This method is only meant as a preliminary, exploratory tool to suggest the possibility for the representation of the different functional groups in the soils. Although this is not meant as a definitive characterization of the functions of the different genera, we believe this approach can provide a useful preliminary determination of the effect that various treatments might have on microbial community Soil Syst. 2020, 4, 0065 5 of 17 composition associated with different functions. The MPS values of the N-Fix pot and AMO pot groups were also determined.

Data Analysis
All tree soil data were analyzed by two groupings as predictor variables: Tree Species (P. macroloba vs. I. edulis), and Land Management (Primary or Secondary Forest). To address Question 1, several approaches were used. Initially, the total MPS values for the N-Fix pot and AMO pot bacterial communities were summed and compared between the different Tree Species and between Land Management practices (Primary vs. Secondary Forest) using the Mann-Whitney test for significance. In addition, the MPS values for the total bacterial, the N-Fix pot and AMO pot bacterial communities were 4th root transformed to account for dominant and rare tax [62]. The MPS transformed DNA data was then converted into Bray-Curtis dissimilarity matrices using PRIMER-E v6 and its' add-on PERMANOVA [62,63]. A series of 2-factorial main and pairwise permutational multivariate analysis of variance (PERMANOVA) tests were conducted on the Bray-Curtis matrices from the three different bacterial groups of genera using PRIMER-E v. 6 and PERMANOVA+, with Tree Species alone (P. macroloba vs. I. edulis), and Land Management (Primary Forest and Secondary Forest) as the factors to determine if either Tree Species or Land Management were drivers of any differences found in the bacterial community compositions. As suggested [62,63], all main and pairwise PERMANOVA tests were based on 9999 unrestricted permutations. A discriminant Canonical Analysis of the Principal Coordinates (CAP; [64]) was also performed on the Bray-Curtis matrices that showed a significant difference between them by PERMANOVA, following PERMANOVA+ guidelines, to provide a rigorous assessment of the strength of the differences of the microbial community compositions between the soils. Strong differences in microbiome compositions between soil comparisons are indicated by CAP axis squared canonical correlations are represented by R 2 values > 0.7, moderate differences by R 2 values ≥ 0.5 to 0.69, and weak differences by R 2 values < 0.5 [64].
The Bray-Curtis matrices in Primer-E v6 were also used to calculate Pielou's evenness (J = H'/ln[S]), and the Margalef's richness (d = (S − 1)/Log(N)) of the total bacterial genera, the N-Fix pot and the AMO pot bacterial genera communities. The Margalef's method compensates for the effects of sample size, while other measurements of richness do not, such as Shannon's index or simple OTU count, both of which can increase simply with sample size. The strength of the differences in the richness and evenness across the different tree soil comparisons were determined using a one-way ANOVA followed by Tukey's HSD or Dunnett's T3 post-hoc analyses or independent samples t-test in SPSS (v.25, Armonk, NY, USA). Prior to ANOVA, the Levene's test was performed in SPSS to determine homogeneity of the variances of the data, and the Shapiro-Wilk's test was performed in SPSS to determine normality of all the data, and whether Tukey's HSD or Dunnett's T3 post-hoc test should be used. Lastly, Mann-Whitney analyses of the MPS values were conducted in SPSS on the N-Fix pot and AMO pot genera with a MPS values of >0.5 in at least one of the soil types analyzed to identify differences in specific genera between the different soil groups.
To address Question 2, all N-cycle data were examined for mean differences between the three tree soil comparisons by one-way ANOVA followed by Tukey's HSD or Dunnett's T3 post-hoc tests, as appropriate using SPSS (v.25, Armonk, NY, USA). In addition, a Distance-Based Linear Model (DistLM) permutation test in PERMANOVA+ was used to determine if there were soil N-cycle metrics or ratios that were significant predictors of the differences in multivariate patterns observed between the different tree soil bacterial community comparisons. We used the previously mentioned MPS data Bray-Curtis matrices as response variables, and log (x + 1) transformed N-cycle data as predictor variables for the DistLM analyses, along with a step-wise selection process, along with an AICc (Akaike's Information Criterion Corrected) selection criterion and 9999 permutations. The AICc criterion is the appropriate method to use when the number of samples/the number of predictor variables is <40 [62], as was the case in this study.

Differences in Bacterial Community MPS, Richness, Evenness, and Specific Genera
Although there were no differences observed in the total MPS values of the N-Fix pot or AMO pot genera between the soils of the two tree species (Figure 1a), differences in richness and evenness of the bacterial communities were observed ( Table 2). The richness of the total bacterial, N-Fix pot , and AMO pot communities was greater in the P. macroloba than I. edulis soils (T-Test, F (2,22) range = 2.18 to 3.62, p range ≤ 0.01 to 0.03), and the evenness of the total bacterial and N-Fix pot (but not the AMO pot ) community was less in the P. macroloba than I. edulis soils (T-Test, F (2,22) range = 6.03 and 7.12, p value ≤ 0.01 and 0.05, respectively).
3.62, p range ≤ 0.01 to 0.03), and the evenness of the total bacterial and N-Fixpot (but not the AMOpot) community was less in the P. macroloba than I. edulis soils (T-Test, F(2,22) range = 6.03 and 7.12, p value ≤ 0.01 and 0.05, respectively). There were differences in the overall MPS values of both the N-Fixpot and AMOpot communities in the P. macroloba soils between the Primary and Secondary forests (i.e., intraspecies differences) but not in the I. edulis soils (Figure 1b). Specifically, MPS values of the N-Fixpot community were greater in the Secondary (34.3% ± 4.5%) than in the Primary (26.9% ± 2.6%) P. macroloba soils (p = 0.008) and the MPS of the AMOpot communities were greater in the Primary (13.0% ± 1.8%) than Secondary (10.4% ± 2.3%) P. macroloba soils (p = 0.05). The only interspecies differences in the overall MPS of the two bacterial groups observed was that the MPS of the N-Fixpot communities were greater in the Secondary P. macroloba (34.3% ± 4.5%) than Secondary I. edulis (26.4% ± 4.5%) soils (p = 0.015). The Primary and Secondary Forest soils also had differences in the richness and evenness of the total bacterial and N-Fixpot bacterial (but not the AMOpot) communities (Table 1, ANOVA, F(2, 22) range = 4.164 to 11.795, p range ≤ 0.01 to 0.02). Specifically, intertree species differences in both habitats were evident as the richness of the total bacterial and N-Fixpot communities were greater and the evenness less in both the Primary Forest and Secondary Forest P. macroloba compared to the Primary and Secondary Forest I. edulis soils (ANOVA. p range ≤ 0.01 to 0.05). There were differences in the overall MPS values of both the N-Fix pot and AMO pot communities in the P. macroloba soils between the Primary and Secondary forests (i.e., intraspecies differences) but Soil Syst. 2020, 4, 0065 7 of 17 not in the I. edulis soils (Figure 1b). Specifically, MPS values of the N-Fix pot community were greater in the Secondary (34.3% ± 4.5%) than in the Primary (26.9% ± 2.6%) P. macroloba soils (p = 0.008) and the MPS of the AMO pot communities were greater in the Primary (13.0% ± 1.8%) than Secondary (10.4% ± 2.3%) P. macroloba soils (p = 0.05). The only interspecies differences in the overall MPS of the two bacterial groups observed was that the MPS of the N-Fix pot communities were greater in the Secondary P. macroloba (34.3% ± 4.5%) than Secondary I. edulis (26.4% ± 4.5%) soils (p = 0.015). The Primary and Secondary Forest soils also had differences in the richness and evenness of the total bacterial and N-Fix pot bacterial (but not the AMO pot ) communities (Table 1, ANOVA, F(2, 22) range = 4.164 to 11.795, p range ≤ 0.01 to 0.02). Specifically, intertree species differences in both habitats were evident as the richness of the total bacterial and N-Fixpot communities were greater and the evenness less in both the Primary Forest and Secondary Forest P. macroloba compared to the Primary and Secondary Forest I. edulis soils (ANOVA. p range ≤ 0.01 to 0.05).    There were 10 genera with MPS values > 0.5% that were significantly different between the soil groups studied ( Table 2). The Mann-Whitney analyses by Tree Species showed the MPS values for the N-Fix pot and AMO pot genera Burkholderia and Bacillus were greater in the P. macroloba soils (p ≤ 0.01 and 0.05, respectively), while the MPS of the AMO pot genera Nitrospira were greater in the I. edulis soils (p = 0.03). Intertree species differences by habitat were observed as the MPS values of Burkholderia were greater in the Primary Forest P. macroloba than I. edulis soils (p ≤ 0.01); the MPS of Burkholderia, Bacillus, and the N-Fix pot genera Rhodoplanes were greater in the Secondary Forest P. macroloba than I. edulis soils (p ≤ 0.01, 0.05, and 0.04, respectively). Intratree species differences between habitats were observed as the MPS values of Bacillus and the N-Fix pot genera Clostridium and Geobacter were greater in the Secondary than Primary Forest P. macroloba soils (p = 0.01, <0.01, and 0.05, respectively); those of N-Fix pot genera Rhizobium, Beijerinckia, and the AMO pot genera Nitrospira, Sphingomonas, and Comamonas, were greater in the Primary than Secondary P. macroloba soils (p ≤ 0.01, 0.02, 0.05, <0.01, respectively). Table 2. Mann-Whitney analyses potential Nitrogen-Fixation and potential Ammonium Oxidizing bacterial genera with MPS values > 0.5% in soils from Pentaclethra macroloba and Inga edulis by Tree Species (Pmacro vs. Inga) and by Land Management (PP = Primary Forest P. macroloba, SP = Secondary Forest P. macroloba, PI = Primary Forest I. edulis, SI = Secondary Forest I. edulis.

Multivariate Analyses of the Bacterial Community Compositions
Significant differences were found in the composition of the total bacterial, N-Fix pot , and AMO pot communities between the different Tree Species soils (Table 3. PERMANOVA, Pseudo-F (2,22) range = 3.716 to 10.853, p range = all < 0.01, respectively). There were strong differences found in the compositions of the total bacterial communities (CAP R 2 = 0.809, p < 0.001) and moderate differences in the N-Fix pot and AMO pot bacterial communities (CAP R 2 = 0.591, and 0.514, p < 0.001, and 0.004, respectively) between the two Tree Species soils (Table 4). At the Land Management level of analysis, there were differences observed in the bacterial community compositions between the Primary and Secondary Forest soils (Table 3. PERMANOVA, Pseudo-F (2,22) RANGE = 2.579 to 6.982, p range = all < 0.01). There were intratree species differences found in the bacterial communities between the P. macroloba (but not the I. edulis) Primary and Secondary Forest soils ( Table 4). The P. macroloba Primary and Secondary Forest soils had strong and significant differences in the compositions of the total bacterial and AMO pot communities (CAP R 2 = 0.917 and 0.850, p = 0.015 and < 0.001, respectively; and PERMANOVA, Pseudo-F (2,22) = 5.295 and 4.533, p both < 0.01, respectively) and moderate differences in the N-Fix pot communities (CAP R 2 = 0.529, p = 0.002; PERMANOVA, Pseudo-F (2,22) = 2.205, p < 0.01). There were intertree species differences observed within the two different habitats (Table 3) as the total bacterial, N-Fix pot , and AMO pot community compositions were significantly different, although weakly, between the Primary Forest P. macroloba and I. edulis tree soils (CAP R 2 range = 0.312 to 0.468, p range ≤ 0.001 to 0.002; PERMANOVA, Pseudo-F (2,22) range = 1.847 to 2.347, p range = 0.01 to 0.05), and moderately to strongly different between the Secondary Forest P. macroloba and I. edulis tree soils (CAP R 2 range = 0.505 to 0.867, p range = 0.001 to 0.052; PERMANOVA, Pseudo-F (2,22) range = 2.108 to 4.335, p range ≤ 0.01 to 0.03).

Differences in N-Metrics
The patterns of the N-metrics levels were different between the Tree Species and the Land Management (Table 4). The P. macroloba tree soils had greater levels of and NO 3 − /TN were greater in the P. macroloba than I. edulis secondary Forest soils (p = 0.01, 0.04, and <0.01, respectively). There were no differences in the NH 4 + and NH 4 + /TN levels between habitats, although these were generally greater, but not statistically, in the different I. edulis-related samples.

Distance Based Linear Modeling
The DistLM analyses (Table 5) by Tree Species showed that NO 3 − levels best predicted the differences in composition of the total bacterial and N-Fix pot communities, explaining 20.85% and 18.09% of the differences, respectively, and NO 3 − , NO 3 − /NH 4 + best predicted the differences in the composition of the AMO pot community, explaining 14.72% of the differences in group composition. The DistLM analyses by Land Management showed that NO 3 − and NO 3 − /TN levels best predicted the compositional differences in total bacterial and N-Fix pot communities, explaining 22.75% and 18.43% of the differences, respectively, and that NO 3 − and NO 3 − /NH 4 + best predicted the differences in AMO poy community composition, explaining 21.4% of the differences in the composition of this group.

Discussion
In this study, we observed significant differences in the soil microbiome communities between P. macroloba and I. edulis and also found inter-and intratree species differences in soil microbiomes when they were compared by the effect of the local land use practice of conversion of forest-to-pasture-to-secondary forest. Our conclusions are that: (a) P. macroloba and I. edulis stimulate the development of different soil microbiome assemblages; (b) differences in these microbiome are also driven by the different the Land Management practice of Secondary Forest development used in this region; (c) I. edulis soils had a less rich but more evenly distributed total bacterial soil microbiome that may have been more impacted by the land use practice or may be developing more slowly than that of P. macroloba soils; (d) the P. macroloba soil microbiome had a richer and less evenly distributed total bacterial microbiome, than I. edulis, suggesting it may have been less impacted by the land use practice or may be developing more quickly with the vegetation changes during secondary forest succession; (e) there was greater generation of the indicators of ammonium oxidation along with an increased richness of the Nit-Fix pot and AMO pot communities associated with the P.macroloba soil microbiomes compared to those of I. edulis, suggesting P. macroloba soil microbiomes may be more important for ammonium oxidizing activity; (f) there was greater generation of the indicators of NH 4 + production in the I. edulis soil microbiomes compared to those of P. macroloba, suggesting that the I.edulis soil microbiomes may be more important for ammonium production activity in these forests; (g) the land management-driven changes in the soil NO 3 − , NH 4 + , NO 3 − /NH 4 + , NO 3 − /TN were closely associated with changes in the soil microbiomes; moreover, the levels of NO 3 − , NO 3 − /NH 4 + , and NO 3 − /TN were found to be predictive of these changes in the soil microbiomes; and (h) the impacts of the land management may be stimulating functional redundancy and taxonomic switching of functional group genera associated with both tree species. These findings indicate that P. macroloba should at minimum be included with I. edulis in future restoration activities in the region, as it promotes a different microbiome and is either less impacted or recovers quicker than the I. edulis soil microbiomes.

Differences between Tree Species
The Tree Species-driven differences observed in the composition, richness, and evenness of the total bacterial, AMO pot , and N-Fix pot bacterial microbiome communities were best predicted by the differences in levels of NO 3 − , NO 3 − /NH 4 + , or NO 3 − /TN in the different soils using the DistLM modeling assessment. Moreover, the differences found in the in bacterial communities between the tree soil microbiomes, the greater levels of NO 3 − -related metrics, and the reduced levels of NH 4 + -related metrics found in the P. macroloba soils, and the opposite found in the I. edulis soils, suggests P. macroloba soil microbiomes are more critical for producing NO 3 − , and the I. edulis microbiomes more critical for producing NH 4 + into these forest soil ecosystems. These findings build on previous studies that have shown I. edulis trees enhance the quantity and quality of soil inorganic N, organic C and N, stimulate C-sequestration and long-term N accumulation, and above ground biomass development in tropical areas more than other non-Inga species [19,42,65,66] and that I. edulis nitrogen-fixing root nodule development and activity is stimulated by the presence of NH 4 + and more inhibited by NO 3 − [67][68][69].
These previous and current findings suggest I. edulis may be stimulating development of a bacterial community that is stable, less rich well-fitted for its niche, and more efficiently producing NH 4 + in comparison to that of P. macroloba, and, thus, may be enhancing the success of nodulation and subsequent growth for I. edulis in these forests. At this point, whether these relationships between plant and soil microbiome are occurring in a "cause and effect" manner is not known, but these observations suggest a logical hypothesis that should be further studied. It would also be of interest to determine if P. macroloba root nodules are stimulated by NO 3 − , in contrast to those of I. edulis.
Much less is known about the influence of P. macroloba on soil ecosystems, although it has been suggested that the activity of the root nodular N-fixing microbes associated with the tree represent an important mechanism for N availability in soils in this region of Costa Rica [65,70]. In these same forests, McGee et al. [26] showed that P. macroloba soil microbiomes were different and associated with greater levels of production of both NH 4 + and C-biomass activities than those of Dipteryx panamensis, which is another member of the Fabaceae family and a dominant canopy tree in the region and also suggested that P. macroloba would be a critically important tree for remediation of damaged soils in the region.

Differences in Land Use Legacy Effects
Legacy effects of land management have been shown to influence composition and succession of microbial communities in damaged habitats [71,72]. Our study suggests legacy effects from the forest-to-pasture-to-secondary forest land management practice common to this region of Costa Rica are influencing the differences in composition and succession of the two tree species soil microbiomes between the Primary and Secondary Forests. There were differences in the composition, along with greater richness and reduced evenness of the total bacterial and N-Fix pot communities, and the greater richness of the AMO pot community, within the both Primary and Secondary Forest soils of P. macroloba, compared to I. edulis tree soils. This could suggest that P. macroloba soil microbiomes are either less affected by the specific land management practice used, or recovering quicker during recovery, along the successional vegetation changes occurring in the developing Secondary Forests, when compared to the I. edulis soil microbiomes. Conversely, the legacy effects of this land management practice could also be resulting in I. edulis soil microbiome communities that are either more impacted by the damage or are recovering more slowly than those of P. macroloba, resulting in a less rich, more evenly distributed group of bacteria that is changing more slowly with the vegetation successional gradient, which could be providing more longer-term soil ecosystem changes. These would be the expected patterns for bacterial communities differentially impacted by habitat damage or recovering from the damage at different rates [71][72][73] during the processes of competitive exclusion [74].
Thus, these ecological patterns suggest that either P. macroloba soil microbial communities are further along the competitive exclusion processes and have stabilized more than in I. edulis soils or that P. macroloba soil microbiomes are less affected by the land management practice implemented. Although these ideas may be speculative interpretations based on the data, they provide the first evidence that this land management practice commonly used in the Northern Zone of Costa Rica may influence the soil microbiomes of P. macroloba differently than I. edulis. Regardless of the reasons or causes of the response, these findings support the idea that P. macroloba should be included with I. edulis in future restoration practices the P. macroloba microbiome may be less affected by the land management disturbance, or the P. macroloba microbiome recovers more quickly from the damage. However, our data also show that the I. edulis soil microbiomes appear to be very important for the production of NH 4 + that may actually enhance growth of that tree. These findings support the use of both trees in remediation of damage.

Evidence of Functional Redundancy and Taxonomic Switching
Functional redundancy is considered as the number of taxa within community guilds performing the same functions [69,75], which could be represented by multiple microbial genera in communities within the same niches, with high degrees of flexibility and resilience, and with similar metabolic capabilities that are phylogenetically conserved [75][76][77][78][79][80]. The switching of the taxa within a niche might serve as predictors of the influences that environmental changes have on microbial communities [77,79,81]. Our study suggests functional redundancy and taxonomic switching may be occurring across the different tree soil treatments, with the changes in functional group genera occurring in response to land management. There were differences in MPS values of certain N-Fix pot and AMO pot genera across the tree soil comparisons, even though there were no differences in total MPS values of either functional group across the comparisons. This suggests these genera may be part of a larger group of functionally redundant bacteria critical for N-cycle activities such that some genera may be replaced by others following land disturbances or successional changes in the vegetation community, but the overall abundance and activity of the entire functional group remains at a steady state. This may be ecologically important for maintaining steady state metabolic activities in these soils during these environmental disturbances/vegetational changes. Future studies are planned to measure the actual functional group metabolic activities, simultaneous with the assessment of the MPS values and diversity of the 7 N-Fix pot genera and 5 AMO pot genera found to "switch" in these tree soil comparisons.

Conclusions
This project showed that P. macroloba and I. edulis tree soil microbiomes and N-cycle metrics were different, and that each of the tree soil systems responded differently to the previous forest-to-pasture-to-secondary forest conversion. The I. edulis soil microbiome communities may be more affected by the land disturbances or may be reaching a steady state equilibrium in composition faster than the P. macrobloba soil microbiomes, which may be more resilient to the land management used or are responding slower to the disturbance and in concert with changes in vegetation that occur during forest succession. However, both of the tree species microbiomes appear to be beneficial to soil ecosystem recovery. Functional redundancy and taxonomic switching are also occurring in the soils, differentially by tree species, in response to the land disturbance. It appears that the I. edulis soil microbiomes may be developing an environment that facilitates the growth of I. edulis and development of its root nodules through the production of greater amounts of NH 4 + . Thus, the soil microbiomes of these two N-fixing tree species are different and are differentially influenced by land management. Both trees should be planted together during reforestation in this region as they would likely provide a more rapid initial microbiome response, followed by a slower developing microbiome response that allows successional vegetation changes occurring during secondary forest regeneration.

Nucleotide Sequence Accession Numbers
The DNA sequence data were submitted to the NCBI Sequence Read Archive (SRA) SUB6799852, the NCBI BioProject PRJNA599579.
Supplementary Materials: The following are available online at http://www.mdpi.com/2571-8789/4/4/0065/s1, Table S1: Genera identified as potentially within one or more of bacterial functional groups Nitrogen-Fixing or Ammonium Oxidizing, using the scientific articles listed, in addition to the databases mentioned in the text.