Microclimate is a strong predictor of the native and invasive plant-associated soil microbiome on San Cristóbal Island, Galápagos archipelago
Alexi A. Schoenborn and Sarah M. Yannarell are co-first authors.
Abstract
Understanding the drivers that affect soil bacterial and fungal communities is essential to understanding and mitigating the impacts of human activity on vulnerable ecosystems like those on the Galápagos Islands. The volcanic slopes of these Islands lead to steep elevation gradients that generate distinct microclimates across small spatial scales. Although much is known about the impacts of invasive plant species on the above-ground biodiversity of the Galápagos Islands, little is known about their resident soil microbial communities and the factors shaping them. Here, we investigate the bacterial and fungal soil communities associated with invasive and native plant species across three distinct microclimates on San Cristóbal Island (arid, transition zone and humid). At each site, we collected soil at three depths (rhizosphere, 5 cm and 15 cm) from multiple plants. Sampling location was the strongest driver of both bacterial and fungal communities, explaining 73% and 43% of variation in the bacterial and fungal community structure, respectively, with additional minor but significant impacts from soil depth and plant type (invasive vs. native). This study highlights the continued need to explore microbial communities across diverse environments and demonstrates how both abiotic and biotic factors impact soil microbial communities in the Galápagos archipelago.
INTRODUCTION
Biodiversity is essential for sustaining proper ecosystem function (Naeem et al., 2012). However, due to global climate change, human impact (deforestation, introduction of invasive species, mining, population growth) and other factors, biodiversity across a vast array of environments is decreasing at an alarming rate (Butchart et al., 2010). Invasive species are one of the largest human-induced sources of global environmental change (Pejchar & Mooney, 2009; Pimentel et al., 2001). Much of the habitable world suffers from the introduction of invasive species; however, there are key locations that are particularly vulnerable (Caujapé-Castells et al., 2010; Dulloo et al., 2002). The Pacific Islands, including the Galápagos Islands, are under intense threat from invasive plant species (van Kleunen et al., 2015). Roughly 1400 invasive species have become naturalized in the Galápagos Islands, the majority of which are plants or plant-associated organisms (Toral-Granda et al., 2017).
The Galápagos Islands are a chain of islands of volcanic origin located ~1000 km west of the coast of Ecuador. Due to their location along the equatorial Pacific, the islands experience distinct weather changes throughout the year, with major influences coming from the El Niño-Southern Oscillation (Atwood & Sachs, 2014) and the Intertropical Convection Zone (ITCZ) (Putnam & Broecker, 2017). The ITCZ largely dictates the seasonal weather patterns observed on the islands: January–June is the hot/wet season with large rain storms occurring across the entire islands while July–December is the cooler/dry season, with coastal zones experiencing rainfall and cloudy conditions at elevations around 700 metres above sea level (masl) (Percy et al., 2016; Schmitt et al., 2018). This seasonal shift results in unique microclimates that are strongly dependent on elevation, ranging from arid near the coast to higher elevation ‘transition zones’ to humid at the volcano summits (Schmitt et al., 2018). These unique microclimates create distinct ecosystems over small geographic scales, which have led to speciation and the endemism of much of the Islands' native flora and fauna. The high levels of endemism make the Galápagos Islands particularly susceptible to invasive species, where perturbations can quickly alter resident biodiversity and harm fragile ecosystems (Robertson & Cramer, 2009).
Environmental microbes in soil and surrounding plants have the potential to greatly alter ecosystem functions (Graham et al., 2016); however, microbes are often overlooked when discussing how to preserve biodiversity. Microbes play a pivotal role in global carbon and nitrogen cycling (Gougoulias et al., 2014), as well as in maintaining plant and soil health (Msimbira & Smith, 2020; Naik et al., 2019) through the prevention of plant pathogens and pests (Qadri et al., 2020) as well as the secretion of metabolites that enhance plant growth (Souza et al., 2015; Spaepen et al., 2007). Soil perturbations, such as those induced in agriculture and crop production, have been shown to reduce soil microbial diversity (Gomiero et al., 2011; Hendgen et al., 2018; U. Singh et al., 2020). Although there have been extensive investigations into the factors that drive microbial communities, such as pH (Fierer & Jackson, 2006; Griffiths et al., 2011; Lauber et al., 2009), salinity (George et al., 2021; Van Horn et al., 2013; Zeglin et al., 2011), soil depth (Xu et al., 2013), soil organic carbon content (Sul et al., 2013), temperature (Oliverio et al., 2017), soil moisture (Serna-Chavez et al., 2013), redox status (Pett-Ridge & Firestone, 2005) and plant communities (Barberán et al., 2015; Peay et al., 2013; Prober et al., 2015), these factors are likely environment specific; thus, continued investigation into the different drivers that apply to unique environments and ecosystems are still needed.
Unlike endemic flora, which are specialized for and adapted to specific microclimates, invasive plants are often able to tolerate a range of environmental conditions, providing them with a growth advantage that allows them to spread quickly. Beyond outcompeting endemic flora, invasive plants can negatively impact the soil nutrient levels as well as inhibit carbon and nitrogen cycling (Ehrenfeld, 2003). However, while much is known about the impacts invasive plant species have on above-ground flora and fauna, we are still limited in our understanding of the role invasive plant species have on below-ground communities such as resident soil microbiota; many plant–microbe interactions are expected to be plant-species specific (Broz et al., 2007; Farrer et al., 2021; Knapp et al., 2012; McLeod et al., 2021; Tian et al., 2021). The invasive plant Psidium guajava (common name: guava) is one of the most detrimental invasive plants on the Galápagos Islands. Unlike the native plants on San Cristóbal, which typically thrive in only a single microclimate, the invasive guava species can be found in most areas of San Cristóbal as well as other inhabited islands, except in the most arid locations on the coasts (Urquía et al., 2019).
The Galápagos Islands are uniquely suited to investigate the interplay of multiple factors on the composition of soil microbial communities. The physical features of the island landscape allow the impact of microclimate and plant species on soil microbial communities to be examined across distinct environments within a small geographic location, removing confounding variables typically present when larger physical distances must be examined to access distinct microclimates. In this study, we investigated soil microbial community structure on San Cristóbal Island in the Galápagos archipelago across arid, transition zone and humid microclimates at the end of the wet season. We characterized and compared both the bacterial and fungal communities associated with the rhizosphere (the region of soil surrounding plant roots) of invasive and native plant types from these three distinct microclimates, designing a study to examine these environmental variables as well as soil sampling depth. Our goal was to better understand the biotic and abiotic factors impacting the composition of soil microbial communities and their potential function. Such an understanding may allow us to harness the biological potential of these microbes to mitigate invasive plant species or alter the impacts of human activity on microbial communities. We determined that the strongest predictor of the bacterial and fungal communities in the unique landscape of San Cristóbal was microclimate, with soil depth and plant invasiveness both modestly but significantly influencing these communities.
EXPERIMENTAL PROCEDURES
Soil sample collection
Soil samples were collected at three sites on San Cristóbal Island, Galápagos Islands and Ecuador on consecutive days in June 2019. The three collection sites were Mirador (June 20, 2019; lat: −0.886068, long: −89.539818), Cerro Alto (June 21, 2019; lat: −0.883333, long: −89.516667) and El Junco (June 22, 2019; lat: −0.8966, long: −89.480206). At Mirador, we collected soil surrounding six guava and six scalesia plants. At Cerro Alto, we collected soil surrounding six guava and six verbena, and at El Junco, we collected soil surrounding 12 guava and 12 miconia plants. Enhanced sampling was performed at El Junco as there were sufficient numbers of plant species within our 50 m × 50 m plot. Additionally, this site was the most preserved and had limited human impact; thus, we wanted to have enough power to determine if invasive plants had unique or altered microbial communities compared with native plants. All equipment and gloves were wiped down between sampling using isopropanol wipes. From each plant we collected roughly 5 g of soil from three depths per plant (rhizosphere, 5 cm below ground and 15–20 cm below ground surface). To collect soil from the rhizosphere, we pulled roots out and shook off soil into a 5 mL conical tube. To collect 5 cm and 15–20 cm sample depths, we measured the depth from the surface and collected plugs horizontally from the soil. Collections took less than 4 h at each site and samples were stored in a cooler with ice packs for transport until returning to the lab, where samples were allotted for microbiological analysis and frozen at −20°C until further processing.
DNA isolation
After transport from the collection site on ice, soil samples for DNA isolation were aliquoted (~1 g into 1.5 mL Eppendorf tubes) and stored immediately at −20°C. DNA isolation was performed using Qiagen DNeasy PowerSoil Kit (Cat # 12888–100), according to the manufacturer's instructions with slight modifications: samples were vortexed for 30 min at maximum speed in the PowerBead tubes, and DNA was eluted in 50 μL of Solution C6. Isolated DNA yield and quality was checked using a spectrophotometer—Thermo Scientific NanoDrop 2000 (Cat # ND-2000). DNA extracts were transferred on ice to the high-throughput sequencing facility at University of North Carolina at Chapel Hill for amplicon sequencing.
Amplicon 16S and ITS rRNA gene sequencing
Members of bacterial and fungal communities were identified by sequencing the V3–V4 region of the 16S rRNA gene and the ITS region, respectively. Soil samples were sequenced according to the protocol described by Caporaso et al. (2011). The extracted DNA was amplified with barcoded primers to enable multiplexed sequencing. We used 341F 5′-CCTACGGGNGGCWGCAG-3′ and 806R 5′-GACTACHVGGGTWTCTAAT-3′ for bacteria and ITS-9F 5′-GAACGCAGCRAAIIGYGA-3′ and ITS-4R 5′-TCCTCCGCTTATTGATATGC-3′ for fungi. The 16S and ITS libraries were constructed with the same barcodes for each sample and combined into the same amplicon pool after they were purified using AMPure beads (Beckman Coulter, Illinois, USA), pooled into a library (100 ng) and quantified by qPCR. Twenty percent of denatured PhiX was added to the amplicon pool (12 pM) and sequenced on the MiSeq platform using 300 + 300 bp paired-end V3 chemistry (Illumina, San Diego, CA).
Processing of the sequence data
We processed sequences in QIIME 2 v2020.8 (Bolyen et al., 2019), using a modified pipeline workflow (Comeau et al., 2017). Primers were removed using the cutadapt (v2.10) plugin (Martin, 2011) and the reads were sorted based on the primers into either the 16S (bacterial) or ITS (fungal) group. Raw sequence reads from soil samples were denoised, filtered and clustered into amplicon sequence variants (ASVs) using the Divisive Amplicon Denoising Algorithm (DADA2, v1.10.0) plugin (Callahan et al., 2016). For the bacterial sequence data, we used the following parameters: ASVs with a frequency of <0.1% of the mean sample depth were removed to account for a bleed-through between MiSeq runs, and the remaining reads were used for subsequent analysis. For the 16S rRNA gene dataset, sequences belonging to mitochondria and chloroplasts were filtered out. After denoising and filtering, the number of sequences per sample ranged between 50 and 555,320 sequences for the 16S run and 7–108,745 sequences for the ITS run. Samples with fewer than 18,800 sequences were excluded from the analysis. These included samples M6-A, C3-C and C1-B for the ITS dataset and C1-C and C6-A for the 16S dataset. Taxonomy was assigned to the ASVs using a pre-trained Naïve-Bayes classifier compared against SILVA v138.99 reference database (Bokulich et al., 2018; Robeson 2nd et al., 2021; Yilmaz et al., 2014). For fungal ASVs, Naïve-Bayes classifier was compared against the UNITE (ver8_99_s_all_04.02.2020) ITS database files (Abarenkov et al., 2020). Alpha diversity was estimated using Shannon entropy index (Shannon, 1948). Bray–Curtis dissimilarity was used to assess compositional differences between conditions (sites or plants or soil depth). ANCOM analysis (q2-composition v.2020.8.0) (Mandal et al., 2015) within the QIIME2 was used to determine differential abundance of fungal and bacterial taxa between the native and invasive plants.
To visualize the community composition, we stacked bar plots of bacterial relative abundances using the ggplot2 package in R v4.0.3 (Wickham, 2016). The clustering of bacterial communities was visualized via NMDS plots constructed from Bray–Curtis distance matrices using ggplot2 (v3.3.3) (Wickham, 2016). We used permutational multivariate analysis of variance (PERMANOVA) tests using the adonis function from the R vegan package (Oksanen et al., 2012) to determine how individual factors (microclimate (site), sampling depth and the plant invasiveness) and their interactions influence differences in microbial communities (Anderson, 2017) as measured by Bray–Curtis dissimilarity. We used 9999 permutations. To obtain Figures 3 and S6 graphs, the analysis and graphing program GraphPad Prism 9.0 was used.
Weather stations
Weather stations were deployed in 2015 at each of the sampling sites (images of weather stations from Mirador and Cerro Alto in Figure S1). The weather stations collected data on precipitation, wind speed and wind direction, temperature and solar radiation. Data gaps were the result of instrument failure. Due to remote location and difficult terrain, repairs to instrumentation took time.
Soil analysis
Soil was collected and dried at 56°C for 72 h and then moved to room temperature and air-dried for an additional 2 weeks. Soil was then stored in plastic ziploc bags at room temperature in the dark until soil analysis was performed. We pooled ~1 g from each of the 144 by site, depth and plant type (ex. El Junco guava rhizosphere samples), resulting in 18 samples total. Soil analysis was performed by the University of Georgia Athens, Agriculture, and Environmental Service Lab. The Dumas method was used for the analysis of total carbon and nitrogen and previously described (Kirsten, 1979); briefly, for total organic carbon analysis, soil is treated with water and 8% sulphuric acid, mixed, and incubated for several hours and then placed in the soil dryer to volatilize carbonates from the sample. Soil (~0.5 g) is loaded into a steel crucible and combusted in an oxygen atmosphere at 1200°C in an Elementar Vario Max Total Combustion Analyser, where elemental C and N are converted into CO2, NOx and N2. Carbon content was determined using an infrared cell and N2 was determined using a thermal conductivity cell. Results were reported as a %. To measure anions, the USEPA Method 300 was used; briefly, a Dionex Model DX-120 ion chromatograph (Sunnyvale, CA) was used with a 25 μL sample loop injector and AS40 automated sampler employed along with a Dionex IonPac AS14A separation column (3 × 150 mm) and IonPac AG14A guard column (4 × 50 mm). A Dionex ACRS-500 (2 mm) Anion Self-Regenerating Suppressor was used for conductivity suppression before detection by a Dionex DS4-1 conductivity cell. Soil was mixed at a 1:1 ratio in deionized water for an hour and filtered before running the samples. Eluent: 8 mM Na2CO3 + 1 mM NaHCO3 (1 mL min−1) was added at a flow rate of 0.48 mL per minute. Data were visualized using the graphing program GraphPad Prism 9.0. Initial significance was determined by conducting Kruskal–Wallis ANOVAs followed by Mann–Whitney t-tests for pairwise comparisons where significance was observed.
RESULTS
Sampling sites on San Cristóbal Island
To determine the effects of distinct microclimates and invasive plant species on resident soil microbiota in unique geographic locations, we conducted this study on the Island of San Cristóbal in the Galápagos Archipelago. We selected three sampling sites that encompassed three distinct microclimates within 4 miles of each other. The sites were Mirador (leeward side, 320 masl and arid), Cerro Alto (leeward side, 520 masl elevation and transition zone) and El Junco (windward side, 690 masl elevation and very humid) (Figures 1A and S1A). These sites were chosen not only because they embody distinct environmental microclimates that are geographically close to one another, but also because they house weather stations that have been continuously collecting environmental variables since 2015, including temperature (°C), % relative humidity and average rainfall (Figure 1B, Table S1). Historically, Mirador is the warmest site in both rainy (January–June) and dry (July–December) seasons compared with the other two sites (Figure 1B). Cerro Alto experiences moderate temperatures and moderate rainfall throughout the year, while El Junco has the lowest average temperatures, highest % humidity and highest average rainfall (Figure 1B).
In addition to the three field sites having distinct microclimates, the vegetation, soil properties and plants also varied with each site (Figure 1). Prior to 2015, the land surrounding Mirador was used for agricultural purposes; it is now used as a conservation farm to grow native plants. The plot of land sampled at Mirador had several citrus trees and shrub grasses present with larger bushes, guava and scalesia spaced out across the plot (Figure S1A). The Cerro Alto sampling site was densely packed with a range of plants including tall grasses, guava and herbs; this land has not previously been used for agriculture, and, although it is elevated above cow pastures, it is fenced off and unlikely to have received any manure runoff (Figure S1A). The land surrounding El Junco is protected by the Galápagos National Park; we sampled from a plot containing guava, miconia, ferns and small grasses (Figure S1A).
Soil sampling scheme
We performed soil collections at the end of the wet season, in June of 2019. At the Mirador and Cerro Alto sites, we obtained soil samples from six native and six non-native plants, while at El Junco we sampled from 12 of each plant type. The non-native invasive plant, P. guajava (guava), was present at all sites. However, there was not a single native plant species sufficiently abundant to consistently sample at all three sites. Therefore, for each sampling site we selected a distinct native plant species that was abundant at that site, and within the same 50 m × 50 m sampling plot as the guava: Scalesia gordilloi at Mirador, Verbena spp. at Cerro Alto and Miconia robinsoniana at El Junco (Figure S1A). From each plant, we collected soil at three depths: (a) rhizosphere, the soil particles adhered to the plant root and shaken off, (b) 5 cm below surface from where the plant emerged from the soil, and (c) 15–20 cm below the plant (hereafter simply 15 cm) (Figures 1D and S1B). The soil collected at each site differed visually in colour, water content and texture (Figure S1C); we did not observe visual differences between different soil depths within a single sample site. We therefore collected 144 soil samples (Mirador and Cerro Alto each having six native and six non-native plants sampled at three depths and El Junco having 12 native and 12 non-native plants sampled at three depths, where sufficient numbers of plants were available to support more extensive sampling).
Soil carbon and nitrogen are influenced by plant type and soil depth
To better quantify the physiochemical differences between the soils at each site, we conducted soil analyses on pooled samples. From each site we pooled soils from the same plant type and sample depth (i.e., all guava from 5 cm; all miconia from rhizosphere, etc.), which led to 18 samples for soil analysis. When the data were analysed by averaging all of the results from samples from the same site, we did not observe significant differences for %TOC (total organic carbon), %C, %N or Cl− mg/kg, but we did observe significant differences in SO42− mg/kg and pH (Figure 2A, Table S2). El Junco had significantly (P < 0.05) elevated levels of sulphate and lower soil pH than the other sites (Figure 2A). When we then compared samples collected from guava versus the native plants across all sites, we observed significantly (P < 0.05) elevated %TOC, %C and %N in samples from guava compared with the native plant samples (Figure 2B). Additionally, we observed significantly lower %TOC, %N and %C for the 15-cm samples compared with either the rhizosphere or 5-cm-depth samples (Figure 2C). These data indicate that the invasive plant guava may either directly or indirectly impact soil carbon levels compared with the native flora. Additionally, soil closest to the plant, regardless of plant type, had higher levels of %TOC, N and C compared with the 15-cm samples.
The dominant bacterial and fungal taxa show distinct distribution patterns across sites
We identified the soil microbial communities present using 16S rDNA and ITS (internal transcribed spacer) amplicon sequencing from the collected soil samples. Rarefaction analysis indicated sufficient read depth was achieved during sequencing (Figure S2). We assigned bacterial taxonomy to amplicon sequence variants (ASVs) using a Naïve-Bayes classifier compared against a SILVA reference database that was trained on the V3–V4 region of the 16S rRNA gene (Bokulich et al., 2018). When we assessed the relative abundances of bacteria and fungi at the phylum, order and genus levels, we saw distinct distribution patterns of community composition at each site (Figure S3). Furthermore, the overall patterns of both bacterial and fungal relative abundances were similar across all samples from a particular site, regardless of plant type or soil depth from which they were collected, indicating that each site has distinct microbial communities.
Across all three sites, the three dominant bacterial phyla were Proteobacteria (26%), Acidobacteriota (formerly Acidobacteria; 25%) and Actinobacteriota (formerly Actinobacteria; 20%) (Figure 2A) (Oren & Garrity, 2021). The most abundant phyla showed distinct distribution patterns across sites. The relative abundances of the Actinobacteriota, Methylomirabilota, Myxococcota and Firmicutes were all highest at Mirador, intermediate at Cerro Alto and then lowest at El Junco (Figure 3A), while the Acidobacteriota and Proteobacteria exhibited the opposite pattern (Figure 3A). In contrast, the relative abundances of the Verrucomicrobiota and Bacteroidota were significantly higher at Cerro Alto than at Mirador or El Junco, while the relative abundances of the Plantcomycetota and Chloroflexi were lowest at Cerro Alto compared with both Mirador and El Junco (Figure 3A).
We similarly assessed the relative abundance of the dominant fungal groups across the three sampling sites. The Ascomycota (56%), Basidomycota (20%) and Chytridiomycota (11%) were the dominant fungal phyla at all three sites. The relative abundance of the Ascomycota was significantly higher at Mirador (78%) compared with Cerro Alto (54%) and El Junco (46%) (Figure 3B), while the Basidiomycota and Mortierellomycota exhibited the opposite pattern (Figure 3B). Finally, the relative abundance of the Chytridiomycota was highest in Cerro Alto compared with the other two sites (Figure 3B).
Although we were not targeting archaea, we did detect two phyla using 16S rRNA sequences, the Thermoplasmatota and Crenarchaeota at El Junco, with undetectable levels at either Mirador or Cerro Alto (Figure 3C).
The microbial communities at El Junco are the least diverse of the three sites
We next investigated the impact sampling site and plant type had on alpha diversity, or mean species richness. The Shannon entropy was similar at Mirador and Cerro Alto for both bacterial and fungal communities (Kruskal–Wallis with Benjamini–Hochberg adjusted P-value (q): q = 0.25 and q = 0.71, respectively) but it was significantly lower for the El Junco samples (q < 0.001 for both bacterial and fungal communities) as compared with either Mirador or Cerro Alto (Figure 4A,B), indicating that El Junco has significantly lower microbial alpha diversity than either Mirador or Cerro Alto.
We then compared the alpha diversity of samples grouped by plant type (invasive or native) to investigate if the invasive guava had lower or higher diversity compared with the native plants at each site. At Mirador and Cerro Alto, there were no significant differences in bacterial alpha diversity between samples from each plant type at any sampled depth (Figure 4C). However, at El Junco, guava-associated soil had significantly higher Shannon entropy (or bacterial diversity) than the native miconia-associated soil (for both the rhizosphere and 5 cm, q = 0.01) (Figure 4C, right). In contrast, the Shannon entropy for the fungal communities at all sites was not significantly different for the native-associated compared with guava-associated soil samples except at Mirador (Figure 4D). This indicates that, although microbial alpha diversity is not uniformly influenced by plant type, the bacterial communities at El Junco have distinct Shannon entropy values depending on the microbial association with the particular plant species.
Geographic location predicts soil microbial community composition
To determine the major drivers of the microbial communities, we explored the variation in community composition across 142 samples (two Cerro Alto samples were excluded due to quality concerns) (Figure 5A,B). We ran PERMANOVA (adonis function in vegan package) with site, plant invasiveness, soil depth and all their interactions. Geographic location (site) explained most of the observed variation between soil communities (PERMANOVA, R2 = 0.73 for bacterial and R2 = 0.41 for fungal communities, P < 0.001 for both), followed by the soil depth (R2 = 0.03 for bacterial and R2 = 0.02 for fungal communities, P < 0.001 for both). Note that all PERMANOVA outputs for these and all following analyses are reported in Table S3. Plant invasiveness explained 1% (P = 0.02) of the bacterial and 3.3% (P < 0.001) of the fungal community variation (Figure S4). The model as a whole explains 82.5% of the observed bacterial community variation and 54.9% of the observed fungal community variation.
The microbial communities of El Junco soils were most distinct from those at Mirador and Cerro Alto, which had some minimal cluster overlap (Figure 5A,B). We observed significantly different bacterial and fungal beta diversity at each site (PERMANOVA, P < 0.001; Figure 5C) and increasing taxonomic divergence between Mirador and the other two sites, Cerro Alto and El Junco (PERMANOVA, P < 0.001) (Figure 5C), with the most dissimilar community from Mirador occurring at El Junco. Altogether, these data indicate that the composition of both the bacterial and fungal communities strongly correlate with geographic location.
Interaction between the geographic location and plant invasiveness and the geographic location and soil sampling depth also contributed to the observed fungal and bacterial variation, with El Junco having the greatest plant invasiveness-induced variance for both bacterial and fungal communities (Figure S4). Overall, these data indicate that while the bacterial and fungal communities are primarily influenced by geographic location, soil depth and plant species also affect their composition, with the extent of that effect dependent on sampling site.
Dominant phyla are distinctly distributed between native and invasive plants
We used analysis of the composition of microbiomes (ANCOM) to compare the differential abundance of taxa associated with plant invasiveness (Table S4). ANCOM analysis detected two bacterial taxa abundances to be significantly different (q < 0.05), one 99.07% identical to a sequence obtained from an uncultured forest soil microorganism invaded by moso bamboo (GenBank# JN851513). The other taxon was 99.75% identical to an uncultured bacterium clone NC3F4h9 (GenBank# JQ387433.2), recovered from ecosystems exposed to 10 years of elevated atmospheric carbon dioxide (Dunbar et al., 2012). Both of these were more prevalent and abundant in invasive plants. Two fungal taxa were statistically more abundant in invasive plants, and these included an uncultured Neopestalotiopsis fungal clone (GenBank# MW163322) and uncultured Archaeospora clone MG-2-06-07 (GenBank# MF590013) recovered as part of the arbuscular mycorrhizal fungal community composition from weathered oil ponds in the Ecuadorian Amazon (Garcés-Ruiz et al., 2017). Neither bacterial nor fungal taxa were previously associated with the invasive guava plant.
Having observed the strongest plant-induced variance at El Junco, we next wanted to compare the most abundant microbial phyla between invasive (guava) and native (miconia) samples at El Junco. We examined the 10 most abundant bacterial and six most abundant fungal phyla in the rhizosphere and 5-cm soil samples (since these depths are most impacted by plant type, Figure S5). The majority of the bacterial (8 out of 10) and fungal phyla (5 out of 6) relative abundances were significantly (P < 0.05) different between guava and miconia as determined by Mann–Whitney t-tests (Figure S6). In addition, based on guava-associated soil having a high percentage of total nitrogen (Figure 2), we looked specifically at the relative abundances of genera suggested to be free-living nitrogen fixers, including Rhizobiums, Rhodomicrobiums, Roseiarcus, MND1 and Frankia. All of the genera investigated, with the exception of Rhodomicrobium, had significantly elevated relative abundances in guava-associated soil compared with miconia-associated soil at El Junco (P < 0.05) (Figure S7).
DISCUSSION
Elevation gradients are often exploited to investigate the environmental impacts on biodiversity. Previous work characterizing microbial community diversity has demonstrated conflicting impacts of elevation. Some studies found higher microbial diversity at higher elevations (Li et al., 2016; Peay et al., 2017), while others found microbial diversity did not change with elevation (Fierer et al., 2011; D. Singh et al., 2014). These discrepancies in how elevation impacts soil microbiota may reflect the different environmental factors found at the specific sites and elevations being investigated (Looby & Martin, 2020). Some of the most analogous studies to ours have been conducted in Hawaii. Both the Galápagos and Hawaiian Islands are volcanic in origin, forming as a result of hotspots beneath their respective tectonic plates. Although both of the above-sea-level islands of these archipelagos are largely similar in age (Candra et al., 2021; Clague & Dalrymple, 2021), the soils at our sites on San Cristóbal are ~800,000–1,000,000 years in age (Candra et al., 2021), which is in contrast to the very young soils (~20,000 years old) examined in a Hawaiian study examining temperature gradients (Selmants et al., 2016) or the young soils (~150,000 years in age) examined in an extensive Hawaiian elevation-precipitation gradient study (Peay et al., 2017). In addition, the Galápagos tends to be cooler than expected for its equatorial location, and the climate in the Galápagos is drier than in Hawaii, with fewer dust depositions from Asia (Candra et al., 2021). Notably, our elevation resolution is sparse (since we only sampled from three elevations) compared with the larger number of samples taken in these Hawaiian studies (Peay et al., 2017; Selmants et al., 2016). Nevertheless, a comparison of our results is informative. Selmants et al. (2016) examined sites where the mean annual temperature (as well as the carbon flux, but not the total carbon storage) varied, but where vegetation, soil type, moisture and pH were constant. No differences in bacteria richness was observed at these sites, indicating that other factors (i.e., pH, plants, precipitation) appear more important than temperature in driving bacterial community composition (Selmants et al., 2016). In Paey et al.'s study (2017) of Hawaiian elevation-precipitation gradients, they observed that fungal richness increased with increasing precipitation, while bacterial richness appeared unimodal. This is in contrast to our results, where both bacterial and fungal diversity metrics were lowest at the highest elevation (El Junco) compared with lower elevations. The relationship between pH and microbial diversity was also different between our work and these Hawaiian studies: in our study, higher pH correlated with higher bacterial and fungal diversity, while in Hawaii, fungal diversity was decreased at higher pH (Peay et al., 2017). Overall, as the review by Looby and Martin (2020) points out, differences in biodiversity across topographic gradients are likely due to variations in a range of abiotic factors such as precipitation, soil moisture, landscape and temperature.
Abiotic factors previously shown to influence microbial community diversity and composition include pH (Borneman & Triplett, 1997; Fierer & Jackson, 2006; Hartman et al., 2008), temperature, soil moisture (Fierer & Jackson, 2006) and organic content (Fierer & Jackson, 2006; Shen et al., 2013; Shen et al., 2019; Wang et al., 2015). Here we determined that, overall, the site from which a sample is obtained (Mirador, Cerro Alto, or El Junco) accounts for ~70% of the bacterial and ~40% of the fungal variance observed. El Junco having lower overall diversity is consistent with prior studies indicating that acidic soils lead to lower bacterial richness (Fierer & Jackson, 2006). In addition, El Junco is significantly more humid than the other two sites, and samples from this site exhibited the most distinct communities compared with Cerro Alto (transition zone) and Mirador (arid). Previous work exploring the influence of precipitation on microbial communities has demonstrated inconsistent patterns. In some cases, more humid environments have increased biodiversity compared with more arid sites (Crits-Christoph et al., 2013; Neilson et al., 2017), while other studies have seen no changes in biodiversity metrics across precipitation gradients (Angel et al., 2010; Bachar et al., 2010). Cerro Alto, the site with the highest bacterial and fungal diversity, receives a moderate and fluctuating amount of precipitation. Previous work exploring wet–dry cycles and transition zones has indicated that differing microbial communities are present depending on whether the location is in the wet or dry part of the cycle (Balasooriya et al., 2008; Chowdhury et al., 2014; Lee et al., 2018); this could explain the enhanced biodiversity observed at Cerro Alto. One limitation of environmental DNA sequencing is that sampling captures all DNA present in the sample regardless of whether the cells are dead, dormant or alive. Thus, future work specifically investigating which microbial members are metabolically active during different seasons and environmental conditions (for instance, with BONCAT or stable isotope probing; Couradeau et al., 2019; Haichar et al., 2016) along with manipulative experiments (such as altering water availability in soil cores) could provide additional insights into the influence of moisture and precipitation on microbial communities at these sites.
Consistent with other soil microbiota studies (Fierer et al., 2012; Janssen, 2006), the dominant bacterial phyla we observed were Acidobacteria, Actinobacteria, Bacteroidetes, Proteobacteria and Verrucomicrobia, and the abundance of these taxa tracked with different environmental variables present at each site. Actinobacteria, which form spores and thus are abundantly found in deserts (Fierer et al., 2012; Harris, 1981), were most abundant at Mirador, the most arid site. At El Junco, the most humid site, Proteobacteria and Acidobacteria were abundant, consistent with data showing they are less abundant in arid environments (Fierer et al., 2012). The acidic soil of El Junco (pH ~4.4) is consistent with the high prevalence of the acid-loving Acidobacteria found there (Fierer et al., 2012; Lauber et al., 2009; Naidoo et al., 2021; Tripathi et al., 2017). In terms of fungi, the Ascomycota and Basidiomycota, frequently dominant phyla in soils (Cloutier et al., 2020; Gargouri et al., 2021; Maestre et al., 2015; Sun et al., 2017), were also abundant in our samples. Ascomycota are more abundant in more arid climates (Gargouri et al., 2021; Smith et al., 2007), and were found at high levels at Mirador, while Basidiomycota are more abundant in high precipitation areas (Peay et al., 2017) and were most abundant at El Junco. Glomeromycota, which include the arbuscular mycorrhizal fungi (AMF) (Lanfranco et al., 2016; Taylor et al., 2015), represented ~3% of the overall fungal abundance observed per site. This finding was intriguing, since previous work has shown that invasive plants have a higher abundance of these fungi compared with native flora (Schmidt & Scow, 1986) (Bever et al., 2001); however, our current analyses are insufficient to accurately probe this clade based on known issues identifying Glomeromycota using amplicon sequencing (Delavaux et al., 2022). Our data indicate that site location is a primary driver of microbial community composition at these sites in the Galápagos Islands, consistent with environmental factors such as soil moisture and pH being important determinants of microbial community composition.
In addition to expanding our understanding of how abiotic elements tied to elevation gradients and site environment impact soil microbial communities, our sampling scheme was unusual in its ability to additionally probe the role of plant invasiveness in impacting bacterial and fungal community composition. Plant roots excrete chemicals, nutrients and carbon sources into the soil, which can alter microbial communities within the rhizosphere (Hu et al., 2018). Plant–microbe interactions are a significant contributor to ecosystem functions and processes (Van Der Heijden et al., 2008), and the microbes involved in these interactions have the potential to enhance plant health (Backer et al., 2018; de Bruijn, 2015; Schirawski & Perlin, 2018). However, how plant–microbe interactions shape the microbial community composition remains unclear, with some studies demonstrating that plant type influences community (Barberán et al., 2015; Hough et al., 2020; Peay et al., 2013; Prober et al., 2015), while others do not observe a plant influence (Lekberg & Waller, 2016; Nunan et al., 2005; Peay et al., 2015; Schlatter et al., 2015); one study showed that fungal endophytes varied over an elevation gradient, even from within a single plant host (Zimmerman & Vitousek, 2012). Our work identified ‘plant invasiveness’ as a minor but significant driver influencing microbial community variance between samples. Although we had only a single invasive plant (guava), the three native plants we examined were phylogenetically diverse, ranging from grasses to shrubs or small tress. We found that the bacterial community composition varied between native and invasive plants in samples from Mirador (arid) and El Junco (humid), while the fungal community was influenced by plant invasiveness at all three sites. In our study, except for the bacterial communities at Cerro Alto (where plant invasiveness does not significantly contribute to the observed community variance), we saw that approximately 9% of the bacterial and 12% of the fungal variance is explained by ‘plant invasiveness’, and that these values increase when selectively analysing the samples closest to the plant root (rhizosphere and 5 cm). One caveat is that—because we examined only one invasive plant species (guava) but grouped the microbiota of three distinct native plant species—we have more deeply probed the taxa associated with guava than those of the native plants. Thus, we may be identifying a guava-specific microbiome rather than for a microbiome specifically associated with invasiveness.
Overall, this study provides insight into the major predictors of soil microbial communities on the Island of San Cristóbal, Galápagos. We identified non-uniform correlations between elevation and environmental variables and microbial patterns at different geological and geographic locations. Because these volcanic islands and their soils are relatively young compared with continental soils, our findings suggest that microbial populations may consistently form under similar environmental conditions regardless of geographic location or relative age of landmass. The snapshot in time described here demonstrates that the bacterial and fungal soil communities at these sites are largely influenced by the microclimate and soil characteristics of each site. We also identified ‘plant invasiveness’ as a smaller, albeit still significant, driver of microbial community composition. Therefore, although invasive plants are dramatically altering the macroscopic ecosystem of the Galápagos Islands, their impact on soil microbial communities does not appear to override the stronger environmental drivers imposed by site location. One critical caveat, however, is that abundance does not necessarily correlate with environmentally relevant functional differences: the key functional actors may not be the most abundant microbes (Jousset et al., 2017). Thus, although the overall microbial community structure may be largely driven by the environmental factors tied to site location, those microbes most relevant to soil health or function may be those taxa more strongly impacted by plant invasiveness. Future metagenomic research for these sites is justified to better understand which microbes are metabolically active (via SIP-metagenomic or meta-transcriptomic studies) and what the plasticity of these communities is over time and in response to climatic changes (via temporal amplicon and metagenomic studies as well as microbial transplantation studies). We expect that such future work in this area would permit direct evaluation of microbial activity in these soils and thus a more nuanced understanding of the differences and similarities between soil sites and the impact of plant invasiveness. Finally, the Galápagos Islands are expected to experience the impacts of climate change early relative to other areas, and thus are hotspots to investigate this global environmental perturbation (Paltán et al., 2021). Investigating how these soil communities shift over time and in response to changes in the seasons is needed to address and potentially mitigate the effects of human-induced global changes such as climate change.
AUTHOR CONTRIBUTIONS
Alexi A Schoenborn: Conceptualization (equal); data curation (equal); formal analysis (equal); investigation (equal); methodology (equal); visualization (equal); writing – original draft (equal); writing – review and editing (equal). Sarah M Yannarell: Conceptualization (equal); data curation (equal); formal analysis (equal); investigation (equal); methodology (equal); writing – review and editing (equal). Caroline T MacVicar: Data curation (equal); formal analysis (supporting); visualization (supporting); writing – original draft (supporting); writing – review and editing (supporting). Noelia N Barriga-Medina: Investigation (equal); writing – review and editing (supporting). Kevin S Bonham: Formal analysis (supporting); investigation (supporting). Antonio Leon-Reyes: Investigation (equal); resources (equal); writing – review and editing (supporting). Diego Riveros-Iregui: Investigation (supporting); resources (equal); writing – review and editing (supporting). Vanja Klepac-Ceraj: Data curation (equal); formal analysis (equal); methodology (equal); supervision (equal); visualization (equal); writing – review and editing (equal). Elizabeth A Shank: Conceptualization (equal); data curation (supporting); funding acquisition (equal); project administration (equal); supervision (equal); visualization (supporting); writing – review and editing (equal).
ACKNOWLEDGEMENTS
This research was authorized under the agreement between the GSC-USFQ, Galápagos National Park and the Ministry of Environment of Ecuador with reference number MAE-DNB-CM-2016-0041. We would like to thank the UNC Center for Galápagos Studies (CGS) for funding provided to Elizabeth A. Shank and Diego Riveros-Iregui through their Galápagos Seed Grant Program in 2018 and 2019 that allowed us to initiate and carry out these studies. We would also like to thank the administrators, staff and scientists at the Galápagos Science Center for providing space, assistance and equipment to carry out this work. We would like to thank Cassandra Pattanayak at Quantitative Analysis Institute at Wellesley College for her advice on statistical analyses and the Wellesley College Undergraduate Student Summer Research Program for providing funding to support Caroline T. MacVicar work on this research.
CONFLICT OF INTEREST STATEMENT
The authors declare no conflicts of interest.
Open Research
DATA AVAILABILITY STATEMENT
QIIME 2 pipeline and R code used to analyse the data and generate the figures can be accessed at https://github.com/Klepac-Ceraj-Lab/Galapagos_soil and https://doi.org/10.5281/zenodo.7553018. Raw sequence files of bacterial 16S rRNA gene and fungal ITS amplicons were deposited to the NCBI Sequence Read Archive under the BioProject accession number: PRJNA780894.
