Exploring the Algarrobo Decline in the Pómac Forest: Unraveling the Relationship between the Endomicrobiome of Neltuma pallida and Enallodiplosis discordis

TL;DR

This study investigates the bacterial communities in algarrobo trees and their pest, Enallodiplosis discordis, to understand the factors contributing to the decline of Neltuma pallida in the Pómac Forest.

Contribution

The study provides new insights into the bacterial endomicrobiome of Neltuma pallida and its pest, revealing limited bacterial overlap and low diversity in tree tissues.

Findings

Neltuma pallida branches showed low bacterial diversity, dominated by host sequences rather than true bacterial endophytes.

E. discordis larvae had higher bacterial diversity, with limited overlap with the host's microbiota.

The decline symptoms may not be directly caused by a specific high-abundance bacterial component in the tree stems.

Abstract

A progressive decline in populations of Neltuma pallida (Humb. & Bonpl. ex Willd.) C.E. Hughes & G.P. Lewis (algarrobo) has been documented in Peru. While this phenomenon has been primarily attributed to herbivorous insect attacks, it has been proposed that disease severity and symptoms are exacerbated by infection with opportunistic microorganisms. Enallodiplosis discordis has been identified as the main pest of N. pallida; with larvae producing chlorotic halos on leaflets that have been suggested to involve microbial activity. In this study, we explored the relationship between the bacterial endomicrobiome of N. pallida branches and the microbiota of E. discordis larvae in the Pómac Forest Historical Sanctuary, a protected, Neltuma forests. We sampled branches from asymptomatic, incipiently diseased, late-stage symptomatic trees, and E. discordis larvae. Bacterial DNA was extracted, and 16 S rRNA gene libraries were prepared and sequenced using ONT long-read sequencing. Our results revealed that N. pallida branch-associated samples, regardless of health status, exhibited low bacterial diversity and were dominated by host sequences, indicating organellar carryover and insufficient detection of true bacterial endophytes in branch tissues. To assess potential bacterial overlap between the host and the pest, we characterized the microbiota of E. discordis larvae, which exhibited greater bacterial diversity, with limited overlap restricted to low- to moderate-abundance bacterial families such as Acidithiobacillaceae and Erwiniaceae. These exploratory analyses suggest that the observed decline symptoms may not be directly mediated by a specific high-abundance component of the bacterial community in the stems and contribute to our understanding of the interplay between biotic factors influencing algarrobo decline.

The online version contains supplementary material available at 10.1007/s00248-026-02711-8.

Figures

Visual summary of the workflow, including sampling, surface sterilization, DNA extraction, 16 S rRNA library preparation, sequencing, and downstream analyses

Bacterial microbiota composition of endophytes from branches of asymptomatic and symptomatic *N. pallida* trees at the family level (a) and genus level (b). AA: Asymptomatic samples, AS: Late-stage symptomatic samples

Bacterial microbiota composition of the endomicrobiome of *E. discordis* and of the endophytes from branches of incipient diseased *N. pallida* trees infested with *E. discordis* at the family (a) and genus (b) levels

Assessment of alpha diversity in the different groups analyzed using the Shannon (a), Simpson (b), and Chao1 (c) indices

PCoA analysis of the different samples from *N. pallida*: asymptomatic tree, incipient diseased tree and late-stage diseased tree, and of the larvae from incipient diseased trees

Heatmap showing the percentage of bacterial genera in *E. discordis* samples and in *N. pallida* trees under different health conditions

Full text

Introduction

Trees of the genus Neltuma previously known as Prosopis [1], commonly known as algarrobo in Peru, are distributed across the arid and desert zones of the seasonally dry tropical forest ecosystem in the country’s northwest, with Neltuma pallida being the dominant species [2, 3]. Algarrobo is considered one of the most ecologically and economically valuable tree species in this ecosystem [2–4]. However, its populations are increasingly threatened by land-use change, illegal logging, and the rising incidence of pests and diseases [3], Currently, it has been stated that the North Peruvian dry forest ecosystem could disappear within the next few years [5], a phenomenon termed as “algarrobo decline”. As the etiology of the non-anthropogenic factors driving this decline remains poorly understood, identifying its potential causes is critical [5].

Algarrobo tree mortality has been mainly attributed to biotic factors, with insects from the families Tortricidae, Cecidomyiidae, and Noctuidae identified as the principal pests [6]. In 2016, eight insect species were recorded as associated with algarrobo damage [2]; however, it was not until 2019 that the dipteran Enallodiplosis discordis was identified as the most significant pest affecting this species in Peru [2]. E. discordis is present year-round on infested trees and is considered a prolific pest due to its life cycle. In its larval stage, it attaches to the leaflets of seedlings and mature trees, causing discoloration, wilting, and defoliation [2, 6]. While insect damage is often the most conspicuous symptom, insects frequently interact with opportunistic microorganisms such as bacteria and fungi [6]. Accordingly, in N. pallida, larvae of E. discordis have been observed to produce chlorotic rings at their attachment sites on foliar tissue, potentially linked to filamentous fungi or bacteria [6]. Some diseases are caused by the presence of endomicroorganisms capable of inducing dysbiosis in host tissues [7]. This observation has led to the hypothesis that certain bacterial species may play an important role in the wilting mechanism observed in N. pallida [6, 7].

Several studies have shown that, beyond genetic factors, plant resistance and tolerance to disease are strongly influenced by the presence of endophytic microorganisms that naturally inhabit plant tissues [8]. These microorganisms can modulate host–pathogen interactions through the production of antimicrobial compounds or by competing with pathogens, thereby enhancing host tolerance [7–9]. The gut microbiota of phytophagous insects plays essential roles in counteracting plant defensive metabolites and reducing their toxicity [10–12].

Recent advances in molecular technologies have greatly enhanced our ability to assess the true breadth of microbial biodiversity [12, 13]. Among these, metagenomics has enabled broad access to prokaryotic information from environmental samples, regardless of culturability [12, 13]. Metabarcoding allows for the reconstruction of microbial community taxonomy from environmental or host-associated samples—such as endophytes and endomicrobiomes—through the analysis of specific marker genes like 16 S rRNA for bacterial identification [12]. Here, we used full-length 16 S rRNA sequencing to profile N. pallida branch endophytes and the larval microbiota of E. discordis, comparing diversity, composition, and host–pest microbiome overlap as a baseline for microbial surveillance of algarrobo decline.

Methods

Our workflow encompassed field sampling, surface sterilization and DNA extraction, 16 S rRNA library preparation and sequencing, and downstream taxonomic and diversity analyses; Fig. 1 provides a visual summary, with full procedures in the following subsections.

Sample Collection

Sampling was conducted in the Pómac Forest Historical Sanctuary (SHBP) in Lambayeque, Peru. Three categories of N. pallida trees were selected: visually healthy (asymptomatic) trees (AA), characterized by dense foliage, green leaflets, and absence of E. discordis larvae; incipiently diseased trees (AI), exhibiting yellowing leaves and visible E. discordis larvae on the leaflets; and late-stage symptomatic trees (AS), nearly defoliated, with foliage restricted to a few apical branches due to long-standing progressive wilting (Fig. 1). E. discordis larvae (ED) from the incipiently diseased trees (AI) were also collected.

Fig. 1. Visual summary of the workflow, including sampling, surface sterilization, DNA extraction, 16 S rRNA library preparation, sequencing, and downstream analyses

Apical branches, consisting of the proximal stem to the leaf lamina (excluding leaflets), were collected from three asymptomatic (AA), incipiently diseased (AI), and late-stage symptomatic (AS) trees (three branches per tree; a total of nine branch samples per health category; n = 3 trees [biological replicates] per category). Each branch was cut, rinsed with sterile water, and surface-sterilized in the field with 70% ethanol. Larvae (ED) were obtained from the same AI trees by gently detaching them from infested leaflets borne on the same branches, with one larval subsample per branch, yielding three larval subsamples per tree (nine in total). Samples (branches and larvae) were collected from three different branches within each tree canopy to maximize representativeness, coded, and transported on ice to the Genomics and Bioinformatics Laboratory for Biodiversity at the Universidad Nacional Mayor de San Marcos (UNMSM).

Surface Sterilization and DNA Extraction

To specifically assess the endophytic microbiota, all branch samples (AA, AI, AS) were re-sterilized in the laboratory (1% NaOCl, followed by three rinses in sterile water; 60 s each) to remove epiphytic bacteria. Internal tissues were then dissected following the protocol described in [14]. The three branch samples belonging to the same tree were pooled and ground together in liquid nitrogen, and approximately 100 mg of homogenized tissue was used for total DNA extraction using the DNeasy Plant Pro Kit (Qiagen, Germany), following the manufacturer’s protocol.

For larval samples, approximately 15 mg of Enallodiplosis discordis larvae were pooled in a 1.5 mL tube. To minimize external DNA contamination, we followed an optimized larval pre-treatment protocol for insect endophytic/endosymbiont analyses: larvae underwent three 1 min washes in 70% ethanol, followed by three 1 min rinses in sterile distilled water [15]. This cycle was repeated twice. DNA was extracted using the ZymoBIOMICS DNA Miniprep Kit (Zymo Research, USA) according to the manufacturer’s protocol.

DNA integrity was assessed by electrophoresis on a 1% agarose gel and visualized under UV light. DNA concentration and purity were determined spectrophotometrically using a NanoDrop Lite (Thermo Scientific), with the 260/280 absorbance ratio used as an indicator of DNA purity.

16 S rRNA Library Preparation and Sequencing

Two libraries were prepared using the Oxford Nanopore 16 S Barcoding Kit 1–24 (SQK-16S024), which includes barcoded primers (27 F: AGAGTTTGATCCTGGCTCAG; 1492R: TACGGYTACCTTGTTACGACTT) for 16 S rRNA gene amplification for up to 24 samples to enable subsequent demultiplexing. PCR reactions were performed under the following conditions to increase sequencing yield: reactions were carried out in a final volume of 40 µL per sample, consisting of 25 µL of LongAmp Hot Start Taq 2X Master Mix (New England Biolabs), 10 ng of template DNA, and 11 µL of nuclease-free water. Thermocycling conditions were as follows: initial denaturation at 95 °C for 1 min; 35 cycles of 95 °C for 20 s, 55 °C for 30 s, and 65 °C for 2 min; and a final extension at 65 °C for 5 min.

PCR products were purified using AMPure XP magnetic beads (Beckman Coulter) according to the manufacturer’s protocol and resuspended in 10 µL of elution buffer (10 mM Tris-HCl, pH 8.0; 50 mM NaCl). Purified amplicons were normalized to equimolar concentrations using Tris-HCl/NaCl buffer, and 10 µL aliquots from each sample were pooled. Subsequently, 1 µL of Rapid Adapter (RAP) was added to the pooled library and incubated for 5 min prior to loading onto a MinION Mk1C sequencer (Oxford Nanopore Technologies).

Sequencing was performed using R9.4.1 flongle flow cells (FLO-MIN106). The flow cell was primed, and the sequencing mixture—comprising the sequencing buffer, loading beads, and the pooled DNA library—was loaded onto the flongle. This mixture stabilizes DNA strands and facilitates their translocation through the nanopores. Sequencing of the 16 S rRNA gene was carried out for approximately 12–18 h under default operating conditions. No mock community control was sequenced.

Bioinformatic Analysis

Sequencing data were first converted to POD5 format using POD5 v0.3.12 (https://github.com/nanoporetech/pod5-file-format) and subsequently basecalled and demultiplexed with Dorado v0.7.2 (https://github.com/nanoporetech/dorado). Adapter sequences (Rapid Adapters, RAP) were trimmed using Porechop v0.2.4 (https://github.com/rrwick/Porechop), and reads were then filtered by length and quality with Chopper v0.8.0 (https://github.com/wdecoster/chopper). Potential plant-derived contaminants were identified by aligning reads against the N. pallida genome (sequenced in-house) [16]. For larval datasets (ED), no explicit host-read subtraction was applied, as no reference genome or closely related assembly is currently available for E. discordis. File handling was performed using Samtools v1.13 [17] and Bedtools v2.30.0 [18]. Quality-control metrics were calculated with SeqKit v2.9.0 [19] before and after filtering.

Taxonomic classification of filtered reads was performed using the NanoCLUST pipeline [17]. NanoCLUST groups similar reads using dimensionality reduction and unsupervised clustering, generates a polished consensus sequence for each cluster, and classifies these sequences using BLAST against the updated NCBI 16 S rRNA database [20]. Descriptive and comparative analyses, as well as data visualization, were conducted in Python v3.10.12 via Google Colab, using libraries including pandas, numpy, matplotlib, seaborn, and scikit-learn. Outputs included bacterial composition profiles, alpha diversity metrics (Shannon, Simpson, and Chao1 indices), principal coordinates analysis (PCoA), and heatmaps based on the relative abundance of taxa at the genus and family levels. Although the aim of this study was to reveal trends at the descriptive level, formal statistical testing between sample groups was performed using R and included non-parametric methods PERMANOVA based on Bray-Curtis dissimilarities with 999 permutations for comparing bacterial compositions, and Kruskal-Wallis for evaluating statistically significant differences in alpha diversity metrics.

Results

Sequencing Success and Read Classification

A total of 28,420 reads were obtained from bacterial endophyte samples derived from both symptomatic and asymptomatic trees. In addition, 71,794 reads were generated from the endomicrobiome of E. discordis larvae and from endophytes of incipiently diseased (AI) trees associated with this insect. All raw sequences were deposited in the Sequence Read Archive (SRA) under BioProject PRJNA1314707. After NanoCLUST classification, read depth per group was 13,496 reads for AA, 31,570 for AI, 14,924 for AS, and 40,224 for (ED) (Table 1). Taxonomic assignments were successfully obtained for all samples.

Table 1. Number of sequences obtained after nanoclust classification, number of reads for each sample group, sample and sample replicate; and number of clusters per replicate for samples from asymptomatic trees, incipient diseased trees and late-stage symptomatic trees, and E. discordis larvaeSample groupSampleSample replicateClassification with NanoCLUST N°. of reads

Reads per sample

N°. of clusters Asymptomatic trees(AA) AA-01

AA-01.1 108445922 AA-01.2 20443 AA-01.3 14643 AA-02

AA-02.1 118037284 AA-02.2 8483 AA-02.3 17003 AA-03

AA-03.1 223251764 AA-03.2 15324 AA-03.3 14122Late-stage symptomatic trees(AS) AS-01

AS-01.1 70438203 AS-01.2 13883 AS-01.3 17283 AS-02

AS-02.1 133654203 AS-02.2 21883 AS-02.3 18963 AS-03

AS-03.1 182456843 AS-03.2 20923 AS-03.3 17683Incipient diseased trees(AI) AI-01

AI-01.1 185011,0665 AI-01.2 8843 AI-01.3 83323 AI-02

AI-02.1 650693983 AI-02.2 3543 AI-02.3 25384 AI-03

AI-03.1 971411,1063 AI-03.2 3383 AI-03.3 10544E. discordis larvae(ED) ED-01

ED-01.1 524012,6568 ED-01.2 34203 ED-01.3 399632 ED-02

ED-02.1 404412,77217 ED-02.2 43322 ED-02.3 439610 ED-03

ED-03.1 428414,79633 ED-03.2 58202 ED-03.3 46925

Composition of Bacterial Endophytes

At the family level (Fig. 2A), asymptomatic trees (AA) were dominated by Symphyonemataceae (75.8–80.3%), with Stellaceae as the secondary family (≈ 10.7–24.2%). However, as detailed below and discussed later, these assignments predominantly reflect plant organellar (chloroplast) DNA rather than true bacterial endophytes. Additional low-abundance families ( ≤ ~ 10%) were detected sporadically, including Acidithiobacillaceae, Erwiniaceae, and Caedimonadaceae. In late-stage symptomatic trees (AS), Symphyonemataceae remained the most abundant family (60.2–80.9%), while Stellaceae increased in two samples (≈ 38–40%). These patterns indicate that the endophytic microbiota of N. pallida stems is relatively simple and remains broadly similar between asymptomatic (AA) and late-stage symptomatic (AS) trees, suggesting that tree health status does not markedly affect the dominant bacterial families detected in woody tissues. Although this study was designed to describe compositional patterns, and the small sample size and number of taxa limits statistical, the non-parametric PERMANOVA test based on Bray-Curtis dissimilarities with 999 permutations showed no significant differences between any sample group (AA, AS, AI, ED) at the family level (pseudo-F: 0.47, p = 1.00; 999 permutations).

Fig. 2. Bacterial microbiota composition of endophytes from branches of asymptomatic and symptomatic N. pallida trees at the family level (a) and genus level (b). AA: Asymptomatic samples, AS: Late-stage symptomatic samples

At the genus level (Fig. 2B), Loriellopsis predominated in asymptomatic trees (AA; 75.8–80.3%), with Stella as the secondary genus (≈ 10–24%), both of which were predominantly organellar. Low-abundance genera (< 8%), such as Acidithiobacillus, Caedimonas, and Erwinia, were detected in specific samples. Late-stage symptomatic trees (AS) exhibited a similar pattern, with Loriellopsis remaining dominant (60.2–80.9%) and Stella occurring at relatively higher proportions in two trees (≈ 38–40%), while other genera were present at lower abundances. Similar to what was observed for the differences at the family level, bacterial composition was not significantly different among the four sample groups at the genus level (pseudo-F: 0.68, p = 1.00).

Comparison between the Endomicrobiome of E. discordis and Endophytes of N. pallida

At the family level (Fig. 3A), incipiently diseased (AI) branches were dominated by Symphyonemataceae (78.9–88.3%), with Stellaceae as the secondary family (≈ 10.7–20.2%), whereas all other families occurred at lower proportions. In contrast, E. discordis larvae (ED) exhibited higher family-level diversity, with dominant families varying among samples. Across ED samples, Bacillaceae, Caryophanaceae, Acidithiobacillaceae, Aurantimonadaceae, and Symphyonemataceae each reached approximately 14–44% in at least one sample (Fig. 3A), along with several additional low-abundance families.

Fig. 3. Bacterial microbiota composition of the endomicrobiome of E. discordis and of the endophytes from branches of incipient diseased N. pallida trees infested with E. discordis at the family (a) and genus (b) levels

At the genus level (Fig. 3B), incipiently diseased (AI) branches were characterized by a strong dominance of Loriellopsis (78.9–88.4%), with Stella as the secondary genus (≈ 10.7–20.2%); all other genera were present at minor proportions (each ≤ 2%). In contrast, E. discordis larvae (ED) displayed heterogeneous microbial profiles, in which Bacillus, Planococcus, Planomicrobium, Aureimonas, Acidithiobacillus, and Loriellopsis each reached peaks of approximately 12–19% in at least one sample, with additional low-abundance genera illustrated in Fig. 3B.

These findings indicate that, whereas the microbiota of incipiently diseased N. pallida branches is overwhelmingly dominated by Loriellopsis with limited variation, E. discordis larvae harbor a more complex and diverse microbial community comprising a broader array of bacterial genera. Our results suggest that larvae acquire part of the tree-associated microbiota but also harbor additional bacterial groups that are likely linked to their physiology, digestive processes, or microhabitat within plant tissues.

Diversity Analysis

Alpha diversity was consistently higher in E. discordis larvae than in tree-associated samples across the dataset. Shannon diversity values were highest in the ED group, whereas AA, AI, and AS clustered at lower values (Fig. 4A). Similarly, Simpson indices indicated greater evenness in ED samples (Fig. 4B), and Chao1 estimates likewise suggested higher richness potential in ED compared with any tree-associated group (Fig. 4C). Although these trends were clear at the descriptive level, statistical testing revealed significant differences only for the Simpson index (Kruskal-Wallis test; H = 8.27, p = 0.0407), while Shannon (H = 7.45, p = 0.0587) and Chao1 (H = 4.90, p = 0.1790) diversity indices did not differ significantly among sample groups, likely reflecting the high intra-group variability observed in ED samples. Collectively, these patterns indicate a broader and potentially more functionally flexible gut microbial community in larvae, whereas branch endophytes reflect host-filtered, lower-diversity assemblages in this dataset. Motivated by these within-sample differences, we next evaluated between-sample structure using principal coordinates analysis (PCoA).

Fig. 4. Assessment of alpha diversity in the different groups analyzed using the Shannon (a), Simpson (b), and Chao1 (c) indices

PCoA Analysis

PCoA revealed a clear separation between larval (ED) and branch samples (AA, AI, AS) (Fig. 5), suggesting that the endophytic microbiota of branches remains relatively stable despite differences in plant health status. Nonetheless, some variation was observed among AS-01 and AS-02 samples, attributable to differences in the relative abundance of specific genera, such as Loriellopsis and Stella. In contrast, E. discordis larvae (ED) exhibited a completely distinct and highly variable microbiota in our dataset (Fig. 5), indicating a community adapted to the larval lifestyle and potentially changing according to developmental stage. This marked difference in microbiome diversity between branches and larvae is consistent with the contrasting nature of the microhabitats represented by woody tissues and insect larvae. To identify the taxa underlying this contrasting pattern, we next examined genus-level heatmaps.

Fig. 5PCoA analysis of the different samples from N. pallida: asymptomatic tree, incipient diseased tree and late-stage diseased tree, and of the larvae from incipient diseased trees

Heatmaps of Microbial Taxa Correlations

To explore patterns among bacterial taxa within each group, heatmaps were generated to display the relative abundance (percentage) of bacterial genera across individual samples (Fig. 6). Higher relative abundance values are represented by lighter colors (yellow and green), whereas low or absent values are shown in darker tones.

The genus Loriellopsis exhibited high relative abundance across all sample groups (AA, AI, AS, and ED), exceeding 85% in certain samples. This pattern suggests that Loriellopsis is a common endophyte of N. pallida and that E. discordis larvae may acquire it through feeding on leaflets. In contrast, the genus Stella was detected exclusively in tree-associated samples (AA, AI, and AS), with a stronger presence in branch tissues. This finding indicates that Stella is a branch-associated bacterial endophyte of N. pallida and is absent from leaflets and from E. discordis larvae.

With respect to the microbiota associated with E. discordis, genera such as Bacillus, Planococcus, and Aureimonas were detected in the ED group, suggesting potential symbiotic associations with the insect host. In addition, although the genus Acidithiobacillus was detected in some samples, its intermediate relative abundance (below 20%) indicates that it is not a dominant component of the microbiota of either N. pallida or E. discordis and may represent a transient presence (Fig. 6). Similar patterns were observed when analyses were performed at the family level (Fig. S1). Taken together, these results indicate that, although Loriellopsis represents a shared component across hosts and tissues, the microbiota of E. discordis displays greater breadth and compositional differentiation, most likely driven by insect-associated factors rather than by the tree environment alone.

Fig. 6. Heatmap showing the percentage of bacterial genera in E. discordis samples and in N. pallida trees under different health conditions

Discussion

In branches from late-stage symptomatic trees (AS), the apparent bacterial signal was dominated by sequences matching plant organelles (chloroplasts and mitochondria), and genuine endophytic bacterial taxa were scarce in our dataset. By contrast, low-abundance bacterial families were detected in some asymptomatic (AA) and incipiently diseased (AI) branches, including Acidithiobacillaceae, Caedimonadaceae, and Erwiniaceae. Acidithiobacillus may be associated with residual mining activity in the Pómac Historical Sanctuary, as members of this genus are known to oxidize sulfur compounds and mobilize metals such as gold, copper, and uranium [21, 22]. The presence of Caedimonas, described as an endosymbiont of protozoa [23], could reflect environmental contamination or an as-yet undescribed ecological function. Erwinia, reported as a pathogen in guava and eucalyptus [24, 25], may act as an opportunistic pathogen under specific host physiological conditions.

Families Symphyonemataceae and Stellaceae accounted for much of the apparent signal across AA, AI, and AS branches. However, further alignment diagnostics suggested that these reads matched plant chloroplast and, to a lesser extent, mitochondrial genomes more closely than bona fide bacterial references. We therefore interpret these assignments as organellar carryover rather than genuine endophytic cyanobacteria in our dataset. This interpretation is also consistent with the limited evidence for endophytic occurrences of these families; except, Stellaceae has been reported mainly from agricultural soils, with proposed roles in phytoremediation and cellulose degradation [26]. Given the homology between chloroplast loci and bacterial 16 S rRNA targets, aggressive host-read subtraction can introduce false negatives; accordingly, we retained these reads for completeness, interpreted them as organellar background, and analyzed genuine low-abundance bacterial families separately. The methodological limitations in identifying endophytes was reported before using 16 S primers, including the pair used in this study [27], requiring more specific primers or alternative methods [28, 29]. In our dataset, larval communities displayed greater richness and evenness than branch endophytes (Fig. 4) and occupied a broader ordination space (Fig. 5). Rather than being dominated by a single lineage, ED samples harbored a multi-lineage assemblage encompassing families commonly reported from insect guts, including Acidithiobacillaceae, Aurantimonadaceae, Bacillaceae, Moraxellaceae, Pseudomonadaceae, Sphingomonadaceae, and Methylobacteriaceae [30–39]. Between-sample heterogeneity may also reflect differences in larval developmental stages [40]. This elevated diversity was accompanied by substantial intra-group variability among larval samples, which likely constrained the statistical detection of consistent differences across all alpha diversity metrics. The detection of Symphyonemataceae in E. discordis samples is most likely attributable to ingestion of N. pallida foliage; as noted above, sequences assigned to this family show greater similarity to chloroplast genomes than to genuine bacterial references. Thus, their presence in the insect gut likely reflects dietary intake rather than an endosymbiotic association.

At the genus level, the larval microbiome was more consistent with insect-associated metabolism than with tree endophytes. Bacillus—recurrently reported in agricultural pests such as Hermetia illucens and Rhynchophorus ferrugineus—supports cellulose degradation and contributes to immunity and detoxification processes [33, 34, 39, 41–43]. Pseudomonas, documented in Nilaparvata lugens and Tuta absoluta and also detected in Gynaikothrips uzeli and adult Liometopum apiculatum, is consistent with facultative symbiosis and the processing of complex polysaccharides and aromatic compounds [44–47]. Complementing these functions, Sphingomonas participates in the turnover of recalcitrant substrates and has been linked to nitrogen cycling in L. apiculatum [44, 47]. Acinetobacter aligns with insecticide degradation (e.g., in T. absoluta) [44, 48], whereas Aureimonas has been associated with gut responses to insecticides in Bombyx mori and Piezodorus guildinii [28, 29]. Finally, Methylobacterium has been implicated in amino acid metabolism and digestive enzyme synthesis in insects [15, 39]. Taken together, these genera suggest an intestinal community configured for cellulose depolymerization, xenobiotic detoxification, and nutrient acquisition, consistent with the broad and heterogeneous profiles observed in E. discordis larvae. Comparisons of relative bacterial abundance between N. pallida endophytes and the gut microbiota of E. discordis should be interpreted with caution, as these communities occur in distinct biological kingdoms and ecological contexts. Taxa that are abundant or potentially beneficial in the tree may therefore occur at low relative abundance in the insect host without implying limited functional relevance. Accordingly, host–insect microbiome comparisons in this study are presented descriptively and do not allow inference of microbial transfer based on relative abundance alone.

Overall, branch endophytes represent a restricted community dominated by Symphyonemataceae/Loriellopsis, which we interpret as organellar carryover rather than true bacterial endophytes. In contrast, E. discordis larvae harbors a broader and more heterogeneous assemblage of taxa that partially overlaps with branch-associated sequences but also includes bacterial lineages enriched in insect-associated functions. In this context, no pattern consistent with systemic bacterial dysbiosis was observed in the branches of N. pallida. Consequently, the hypothesis that E. discordis acts as a vector of systemic phytopathogenic bacteria is not supported by our results. Instead, it is more plausible that the observed syndrome is localized in foliar tissues or involves other microorganisms (e.g., fungi or viruses) acting under larval feeding stress. This interpretation is consistent with acoustic tomography data taken previously (C. Arbizu, PhD, Director de la Subdirección de Investigación y de Estudios Especiales, Instituto Nacional de Innovación Agraria-INIA, email, April 29, 2022) confirming internal trunk integrity in symptomatic trees prior to the onset of this study (Fig. S2). Nevertheless, a larger sample size and/or greater sequencing depth would be valuable to further test this hypothesis. In addition, classical microbiological isolation of bacteria from necrotic or chlorotic leaf tissues, followed by 16 S rRNA gene sequencing of individual colonies, could help assess microbial viability and clarify any potential involvement in defoliation. Finally, the implementation of Nanopore Adaptive Sampling (NAS) or the use of more selective 16 S rRNA primer sets designed to minimize chloroplast amplification would likely enhance the detection of low-abundance bacterial taxa that may otherwise remain undetected.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1: Fig. S1 Heatmap showing the percentage presence of bacterial families in E. discordis samples and in N. pallida trees at different health conditions

Supplementary Material 2: Fig. S2 Forest tomographies were performed on nine healthy and diseased trees with diameters at breast height (DBH) ranging from 92 to 210 cm, including individuals analyzed in this study for their microbiome. An ArborSonic 3D tomograph equipped with SD-2 sensors (Fakopp) was used. All trees were georeferenced using a GPS device, and morphological as well as dasometric characteristics were recorded. Three cross-sectional layers per tree (at 40, 100, and 150 cm above ground level) were analyzed. Based on the results obtained from each layer, a multilayer reconstruction of the trunk was generated. No structural differences were detected between symptomatic (a) and asymptomatic (b) trees with respect to trunk integrity, suggesting that leaf loss is unlikely to be caused by decay transmitted through the vascular bundles. (C. Arbizu, PhD, Director de la Subdirección de Investigación y de Estudios Especiales, Instituto Nacional de Innovación Agraria-INIA, email, April 29, 2022)