Seed bacterial microbiota in post-submergence tolerant and sensitive barley genotypes

Plant materials

Accessions of barley (Hordeium vulgare L. ssp. vulgare) WB-433, WB-281, WB-050 and WB-352 were sourced from the Whealbi project (https://www.whealbi.eu/). Seeds were derived from lines obtained from two rounds of single seed descent (Bustos-Korts et al. 2019), and subsequently reproduced during the 2020 growing season at Fiorenzuola d’Arda in Northern Italy (44°56′N 9°54′E, 80 m.a.s.l.). The soils in Fiorenzuola d’Arda have the following composition: silt 50%, clay 36%, sand 14%; the soil is slightly basic with pH 7.85 and an organic mass of 3.0% in the upper 30 cm of topsoil. They are classified as fine silty, mixed and mesic Udic Ustochrepts (Rizza et al. 2018).

Accessions were selected in terms of previously identified contrasting phenotypes in relation to the Submergence Tolerance Index (STI, Gómez-Álvarez et al. 2023), i.e. whether they have a capacity to germinate after 2 days of submergence followed by 5 days of recovery. This index is high in WB-281 (Tajikistan) and WB-433 (Montenegro) and low in WB-050 (Greece) and WB-352 (Syria).

Seeds were stored at 4°C in a LAB-MIDI refrigerator (Desmon Scientific, Nusco, Italy). Prior to the submergence treatment, hulled seeds were surface sterilised with 75% ethanol for 2 min, 3.5% sodium hypochlorite for 2 min, and 75% ethanol for 1 min (Bertani et al. 2016). Seeds were washed with sterile water at each step, and at the end of the process.

Sampling and submergence experiment

Sterile seeds were placed in Magenta GA-7 (Merck, 77 mm × 77 mm × 97 mm) boxes. They were germinated in a growing incubator (Percival Scientific, Perry, IA, USA) at 20°C and 50% relative humidity (RH) in the dark. In air, seeds were placed on wet sterile filter paper. In submergence, they were transferred to Magenta GA-7 boxes filled with sterile water. The experiments included four replicas for each barley accession under air and submergence, with 15 seeds for each replicate. After 48 h, 1 g of barley seeds for each replicate was stored at −80°C. DNA extraction, library preparation and sequencing were performed at IGATech Technology Services (Udine, Italy). Possible external contamination of submergence water was checked by incubation on LB (Luria–Bertani) plates at 30°C for 48 h under dark aerobic conditions.

DNA extraction, library preparation and sequencing

DNA was extracted from barley seeds using the kit Dneasy Mericon Food (Qiagen, Venlo, NL) IGATech standard protocol. The hypervariable regions V3 and V4 of the 16S rRNA gene were amplified with the 16S-341F 5′-CCTACGGGNBGCASCAG-3′ and 16S-805R 5′-GACTACNVGGGTATCTAATCC-3′ primer set. An initial peptide nucleic acid (PNA) clamping was applied during the first polymerase chain reaction (PCR) step to block amplification of the host mitochondrial and chloroplast 16S sequences (Lundberg et al. 2013), following the PNA Bio protocol (Newbury Park, CA, USA). The amplification included 25 PCR cycles using locus-specific primers, and a subsequent amplification that integrated the relevant flow-cell binding domains and unique indices (NexteraXT Index Kit, Illumina, San Diego, CA, USA). The sequencing of libraries was performed with the 300 bp paired-end mode using MiSeq (Illumina).

Bioinformatic analysis

The quality of the sequencing process was verified with FastQC (Andrews 2010). The removal of low-quality reads was performed with ERNE-FILTERING (Del Fabbro et al. 2013). After denoising using the QIIME2 ver. 1.9.0 DADA2 plugin (Caporaso et al. 2010), high quality sequences with a minimum length of 275 bp were converted into amplicon sequencing variants (ASVs). Chimeras were removed using the QIIME2 UCHIME plugin. ASVs sequences were aligned to the SILVA (Ribosomal RNA Gene Database Project Data1) database (Quast et al. 2013) (ver. 132), using a 99% sequence identity. The taxonomy matrix and ASVs table were used for subsequent studies.

Statistical analysis

R (R Core Team 2021) was used to perform statistical analyses. Briefly, a phyloseq object was obtained with R/Phyloseq (McMurdie and Holmes 2013). After the removal of reads of chloroplasts, mitochondria and unassigned features, the rarefraction of the ASVs table was performed to the minimum library size. The ‘estimate_richness’ function of R/Phyloseq was used to calculate Chao1 and Shannon indexes related to alpha-diversity. After the Shapiro–Wilk test, Student’s t-test was used to calculate differences between treatments and phenotype.

Cumulative sum scaling (CSS) was used to normalise data with cumNorm() from the R/MetagenomeSeq v.3.8 (Paulson et al. 2013). For the calculation of the beta-diversity, the canonical analysis of principal coordinates (CAPs) feature of the ordinate() function of the R/Phyloseq was utilised, with constraining of the phenotype and treatment variables. Weighted Unifrac, Bray–Curtis and Jaccard distances were used to analyse differences, which were confirmed using PERMANOVA analysis for the three distance matrixes. Venn diagrams were created using the ps_venn() function from the MicEco package (Russel 2023), taking into account the number of ASVs belonging to each group. Pheatmap() function from Pheatmap() package (Kolde and Kolde 2015)was utilised to perform the heatmaps, where the data was transformed to log (1 + x).

Functional prediction based on 16S rRNA gene sequence analysis

Functional prediction based on 16S rRNA gene sequence was performed by using the MicFunPred resource (Mongad et al. 2021).The ASVs table and the representative reads were utilised, maintaining borderline chimeras. The analysis was performed using Pfam and KEGG Orthology (KO) databases (Li et al. 2014; Kanehisa et al. 2016a, 2016b; Keller-Costa et al. 2021; Mistry et al. 2021). The analysis of results was conducted using R (R Core Team 2021). ANOVA analysis (P-value < 0.05) corrected by the False Discovery Rate (FDR) was applied to detect any significant differences among KO groups and Pfam of identified bacteria. The visualisation of MicFunPred results was obtained with the Pheatmap() function, using the viridis colour palette from R/Viridis (Garnier et al. 2021).

Seed microbiota differs in tolerant and sensitive barley accessions

Two tolerant (WB-281, WB-433) and two sensitive (WB-050, WB-352) barley accessions were previously found to be able or not to germinate after 2 days of submergence followed by 5 days of recovery (Gómez-Álvarez et al. 2023). Seeds belonging to these accessions were collected after 48 h under air and complete submergence before the recovery period (Fig. 1a). In a previous work, we found that a differential lignification during seed development is responsible for the subsequent phenotype observed during germination (Gómez-Álvarez et al. 2023) (Fig. 1a). Our goal was to study whether seed characteristics could be linked to differences in seed microbial communities, which in turn may play a role in the dissimilar responses observed during the recovery period.

Fig. 1.

(a) The experimental set up, showing the sampling point and differential phenotypes of tolerant and sensitive barley accessions after the recovery period. (b) Alpha-diversity analysis performed by using Shannon and Chao1 indexes (Student t-test) on samples collected before the recovery period. (c) Beta-diversity analysis performed using the canonical analysis of principal coordinates (CAP) constrained for variables on the Bray–Curtis distance (PERMANOVA test). n.s., not significant; sub, submergence.

One sampling point was analysed: 48 h in air (control, both tolerant and sensitive barley accessions subsequently germinated in this condition) and 48 h under submergence (treatment, only tolerant varieties subsequently germinated during the recovery period) (Fig. 1a). Our hypothesis is that the differences observed during the recovery period between sensitive and tolerant accessions might be associated with the presence of dissimilar microbial communities after submergence.

The 16S rRNA gene analysis generated 5 016 648 paired-end reads. After in silico removal of chimeras and unassigned sequences and querying the SILVA database (99% sequence similarity), 2010 ASVs were assigned. Alpha-diversity was estimated using the Shannon and Chao1 indexes, comparing the different treatments (air and submergence), the different phenotypes (sensitive and tolerant) and the interaction between them.

The Shannon index of diversity revealed a high range of variability among samples in both air and submergence, whose difference was not significant (Fig. 1b). Although the Shannon index distribution among samples was large in both sensitive and tolerant barley accessions, the difference between phenotypes was significant (Student’s t-test, P < 0.05). This revealed that tolerant accessions have a higher diversity than sensitive accessions regardless of the growth conditions (Fig. 1b). The Chao1 index of richness also showed a large range of variation and the difference between air and submergence conditions was not significant (Fig. 1b). On the contrary, the Chao1 index was significantly different between sensitive and tolerant accessions, although there was a large variation in the tolerant samples (Fig. 1b). For both the indexes, the interaction between treatments and phenotypes was not significant.

Beta-diversity analysis of community composition was performed. Visualisation of the data distribution was computed using the canonical analysis of principal components (CAP) with a constraint for the treatment condition and phenotype (Fig. 1c). The Bray–Curtis CAP was followed by the permutation analysis of variance (PERMANOVA), which found an effect for phenotypes (P < 0.05) but not for treatments and phenotype-by-treatment interactions. The use of weighted Unifrac and Jaccard distances showed the same pattern, albeit not significant (Supplementary Table S1).

These results support the hypothesis of a putative endophytic bacterial microbiota difference between sensitive and tolerant barley accessions, irrespective of the air/submergence conditions. This conclusion differs with the result previously found for rice under submergence (Ahumada et al. 2022), where the tolerant rice accessions under submergence were part of a separate cluster, suggesting a specific bacterial community. In fact, rice and barley differ in their response to submergence in that rice is able to germinate under water whereas barley is not. The Arborio rice tolerant variety is able to elongate much of the coleoptile toward the water surface where it acts as a snorkel in order to allow O₂ diffusion to underwater organs (Narsai et al. 2015; Shiono et al. 2022). This particular characteristic may be associated with a different bacterial community than the one in the variety that elongates the coleoptile less. However, this does not happen with barley, where none of the accessions germinate under water.

Sensitive and tolerant barley accessions show specific bacteria phyla and families

A Venn diagram was used to highlight the distribution of the ASVs in barley accessions characterised by different phenotypes, showing significant differences in alpha and beta-diversity indexes. The results show that sensitive and tolerant accessions harbour unique communities with only 117 ASVs in common. Tolerant barley accessions had double the amount of ASVs relative to sensitive accessions (Fig. 2a), reflecting the results obtained with the Chao1 index (Fig. 1b).

Fig. 2.

(a) Venn diagram for the variable phenotype representing the total number of ASVs identified as described in the ‘Bioinformatic Analysis’ section. (b–e) Heatmap of the relative abundance for the 10 most abundant unique ASVs belonging to sensitive and tolerant varieties grouped by phylum (b, c) and family (d, e). Unique ASVs for sensitive and tolerant accessions are in bold. (f) Boxplots of the significant genera depending on the phenotype variable (FDR < 0.05) and showing differences between air and submergence samples. Relative abundance is calculated as −log(1 + X) of number of counts for given ASVs. m., members.

The heatmap of the ten most abundant phyla (representing 96% of the community) and the most abundant families (representing 46% of the community) for tolerant and sensitive accessions (Fig. 2b–e) showed the predominance of the phyla Proteobacteria and Actinobacteria (alt. Actinomicetota) in both sensitive and tolerant accessions. Both phyla were previously found to be abundant in the rhizosphere of H. vulgare ssp. vulgare elite cultivars (Escudero-Martinez et al. 2022). Proteobacteria and Actinobacteria were also found to be predominant in the rhizosphere and root of the H. vulgare ssp. vulgare landrace, in modern cultivars and in H. vulgare ssp. spontaneum (Bulgarelli et al. 2015) and in the barley seed microbiome of different genotypes (Bziuk et al. 2021). The phylum of Proteobacteria remained high under submergence only in sensitive accessions.

The phylum Cyanobacteria was abundant only in tolerant barley accessions (Fig. 2c). Cyanobacteria have been found in association with terrestrial plants, and their capacity to fix CO₂ and atmospheric nitrogen can have a large impact on the host plant, with a growing interest when available in the rice field (Álvarez et al. 2020). On the other hand, the phylum Deferribacterota (alt. Deferribacteres) was abundant only in the list of the sensitive accessions (Fig. 2b). Deferribacterota are halotolerant and frequently found in marine niches (Alauzet and Jumas-Bilak 2014). Interestingly, they are preferentially anaerobes and were found mostly in sensitive seeds, which are likely hypoxic (Gómez-Álvarez et al. 2023).

Among the families, Paenibacillaceae, Oxalobacteraceae, Bacillaceae, Erwiniaceae and Beijerinckiaceae were common to both tolerant and sensitive barley accessions (Fig. 2d, e). Paenibacillaceae and Oxalobacteraceae were among the most abundant in barley seeds also in previous works (Yang et al. 2017; Bziuk et al. 2021), whereas the other families seem to be specific to the accessions that we analysed. In the sensitive barley accessions, the Paenibacillaceae family was the most abundant in air and was reduced under submergence (Fig. 2d).

The unique ASVs of sensitive barley accessions belong to the Deferribacteraceae, Micrococcaceae, Azospirillaceae, Sphingomonadaceae and Verrucomicrobiaceae, which are absent from the list of predominant families in tolerant accessions (Fig. 2d–e). At the family level, in tolerant accessions there is an enrichment of ASVs from Microbacteriaceae, Comamonadaceae, Rhizobiaceae, Flavobacteriaceae and Sphingobacteriaceae, not observed in the 10 most represented groups of sensitive barley accessions (Fig. 2d–e). Of these, Microbacteriaceae, Comamonadaceae and Flavobacteriaceae have been found to be abundant in barley root (Bulgarelli et al. 2015). Sphingobacteriaceae are known to be endophytes of maize (Kämpfer et al. 2016). Microbacteriaceae and Oxalobacteriaceae are reduced under submergence, whereas there were more Bacillaceae (Fig. 2e).

Air and submergence samples show a similar amount of unique ASVs (Fig. S1), reflecting the result of the alpha and beta-diversity analyses. The ASVs that were only present in the submergence samples belong to Chloroflexi, whereas those present only in air are Patescibacteria, which have been found to dominate in the rhizosphere and roots of spring wheat microbiota respectively (Francioli et al. 2021).

At the family level, Sphingobacteriaceae, Xanthobacteriaceae, Bacteroidaceae Vicinamibacteraceae and Pseudomonadaceae are specifically present in submergence (Fig. S1). These families are different from what has been found in spring wheat microbiota under flooding conditions at different growing stages, suggesting a specific response of these plant species (Francioli et al. 2021, 2022).

The most significant genus difference (FDR, for all the single comparisons P-value < 0.05) between sensitive and tolerant barley accessions were analysed (Fig. 2f). Bacillus, Flavobacterium, RB41, and Rubrobacter were found only in tolerant samples both in air and in submergence conditions. Bacillus is one of the twenty most abundant genera in seed and rhizosphere microbiome in barley (Bziuk et al. 2021). Bacillus species with an endophytic lifestyle can provide plants with protection from phytopathogens, increased resistance to abiotic stresses, and plant growth promotion (Lopes et al. 2018). Flavobacterium has been described in Norwegian barley germinated seeds (3–4 days) (Jiao et al. 2022). RB41 has also been previously observed in barley (Lewin et al. 2021).

In our study, the genera Gemmata, Paenibacillus, Reyranella and Saccharimonadales members were found only in sensitive varieties. Paenibacillus has been reported as one of the most abundant in barley seeds (Yang et al. 2017; Bziuk et al. 2021) and, when isolated, it is able to produce protease, β-1-3-glucanase, cellulase and indole-3-acetic acid (IAA) (Bziuk et al. 2021). It has also been reported in ungerminated Norwegian barley seeds (Østlie et al. 2021).

Bacteria identified in tolerant barley accessions are functionally enriched

Significantly different functions between bacteria identified in sensitive and tolerant barley accessions were predicted, using the KEGG ORTHOLOGY (KO) database (Fig. 3). Tolerant barley accessions showed a bacterial functional enrichment compared to sensitive accessions in all the categories analysed. Interestingly, these include several categories involved in sugar-related mechanisms, such as the alpha-glucoside transport system, pentose phosphate pathway, maltogenic alpha-amylase, starch and sucrose metabolism, multiple sugar transport system and D-allose transport system (Fig. 3). Such activities might be involved in the use of sugar when the plant germinates after submergence during the recovery period and thus help the plant to establish the seedlings. On the other hand, they may suggest that the availability of sugars from seeds in germination enriches the bacterial community.

Fig. 3.

Functional analysis with MicFunPred using the KEGG ORTHOLOGY (KO) database depending on the phenotype variable (FDR, P < 0.05). Relative abundance is calculated as −log(1 + X) of number of counts for given ASVs. PTS, phosphotransferase system.

Significantly different functions between air and submergence were also predicted using the KO database (Fig. S2). Interestingly, samples from submerged seeds were enriched in bacteria with putative functions involved in starch and sucrose metabolism, which is crucial in overcoming submergence stress. On the contrary, nitric oxide synthase function is reduced under submergence. Nitric oxide is crucial for seeds to germinate, and submergence stress has been found to induce secondary dormancy in sensitive barley accessions (Gómez-Álvarez et al. 2023), suggesting a potential link with plant phenotype.