Microbes participate in ecological communities, much like multicellular organisms. However, microbial communities lack the centuries of observation and theory describing and predicting ecological processes available for multicellular organisms. Here, we examine early bacterial community assembly in the water-filled internodes of Amazonian bamboos from the genus Guadua. Bamboo stands form distinct habitat patches within the lowland Amazonian rainforest and provide habitat for a suite of vertebrate and invertebrate species. Guadua bamboos develop sealed, water-filled internodes as they grow. Internodes are presumed sterile or near-sterile (ie, containing small loads of endosphere-associated microbes) while closed, but most are eventually opened to the environment by animals, after which they are colonized by microbes.  We find that microbial community diversity increases sharply over the first few days of environmental exposure, and taxonomic identity of the microbes changes through this time period as is predicted for early community assembly in macroscopic communities. Microbial community taxonomic turnover is consistent at the bacteria phylum level, but at the level of Operational Taxonomic Units (OTUs), internode communities become increasingly differentiated through time. We argue that these tropical bamboos form an ideal study system for microbial community ecology due to their near-sterile condition prior to opening, relatively consistent environment after opening, and functionally limitless possibilities for replicates. Given the potential importance of opened internode habitats as locations of transmission for both pathogenic and beneficial microbes among animals, understanding the microbial dynamics of the internode is a key conservation concern for the insect and amphibian species that use this microhabitat.

Sample Collection

We collected our samples at Los Amigos Biological Station in southern Peru starting on 30 November 2016. Los Amigos is located in lowland tropical rain forest adjacent to the Madre de Dios river, an Amazon river tributary. We collected water samples from eight bamboo stalks of the species Guadua weberbaueri that were distributed across two plots in upland terra firme forest (Figure 1A). We sampled four stalks from each plot. We determined the following eligibility for internode sampling: the internode must be 1 to 2 meters above the ground, contain water, and be closed. We opened the walls of bamboo internodes using a bleach-sterilized blade to cut a 20 – 40 mm notch horizontally near the top of the node, above the water level. Then we pulled back a section of the internode’s external skin using the edge of the blade (Figure 1B). We chose this method because it mimics the actions of foraging brown capuchin monkeys (Davidson et al. 2006; Jacobs and von May 2012).

At the time of opening, we collected 0.5mL of water from each bamboo internode using sterilized transfer pipettes (Figure 1B). We preserved these samples in 1.5mL of CTAB buffer. This buffer has been used in similar studies of microbial communities in other phytotelmata (e.g., pitcher plants; (Bittleston et al. 2016)). Our subsequent sampling occurred 24 hours after opening, 48 hours after opening, and 168 hours after opening. 

Nucleic Acid Extraction:

We extracted the total metagenomic DNA from the samples by incubating 200 microliters of CTAB and water mixture with 400 microliters of buffer ATL from Qiagen’s DNEasy Blood and Tissue kit and 10 microliters of proteinase K. We incubated the samples at 56 ̊C for 24 hours. We then performed a phenol-chloroform-isoamyl alcohol extraction following Barker 1998. We chose the phenol-chloroform approach to minimize contamination by plant secondary metabolites that could inhibit PCR or sequencing reactions. We included a negative control in the extraction and processed and sequenced the control along with our samples. Samples were amplified using primers specifically developed to target the V3-V4 hypervariable region of the bacterial 16S rRNA gene (Kozich et al. 2013). We ran 25 PCR cycles with an annealing temperature of 55 C. Sequencing libraries were created with the NEBNext DNA Ultra prep kit 2 (cat# E7645L) with barcoded dual-index primers.

Purified and eluted DNA amplicons were stored at 4 ̊C before being sent for sequencing. We standardized all PCR products using a Qubit measurement and pooled equimolar quantities of each separate PCR well for the final run. Each sample, taken from a single bamboo on a sampling day, was individually barcoded prior to DNA sequencing. Libraries were sequenced using the Illumina MiSeq platform. Data from this project can be found on NCBI’s Short Read Archive under BioProject PRJNA766623.

Bioinformatic pipeline:

We used the program mothur v.1.48.0 to identify OTUs (Schloss et al. 2009). Following the mothur miseq SOP, we made contigs from our sequences, removed fragments that were not within 8 base pairs of our 292 base pair target length, and removed sequences with ambiguous base calls or homopolymers over 8 bp long. We removed sequences that did not align within an 80% similarity cutoff to a 16S bacteria sequence in the SILVA v.138 database (Quast et al. 2012). We identified the unique sequences in the dataset and removed chimeras. We matched each of our sequences to the RPD Classifier database using the mothur classify.seqs and removed any sequence that aligned most closely with a non-bacteria reference. We then clustered sequences at a 0.03 distance threshold. Finally, we found the consensus classification for our OTUs using the RPD Classifier database v18. The RPD Classifier lists some chloroplast sequences under their “Bacteria” kingdom designation, using the phylum-level labels “Cyanobacteria/Chloroplast.” Since we were likely to have bamboo DNA in our extractions, we removed any plant contamination from our sequences, resulting in the removal of two OTUs.

We used a custom script in R 4.1.1 to parse the output ‘mcc.shared’ sample by OTU matrix and ‘mcc.0.03.cons.taxonomy’ file into an OTU-by-sample matrix in which each cell of the matrix represented the number of reads of an OTU in a specific bamboo-day sample (R Core Team 2021). Both input files and the code used for processing are available Dryad repository. To account for possible contamination, we removed any OTU represented by more than 10 sequences in our negative control. Finally, we set all OTUs with a read depth of 10 or less to zero to account for index hopping during sequencing. To account for possible amplification failure, we removed all days of any bamboo stems that had zero OTUs during any day besides the first. After these steps, we retained seven stems, 19-5, 19-7, 19-8, 19-10, 24-3, 24-6, and 24-9.