Introduction

Salt marshes are among the most productive ecosystems in terms of primary production and are a habitat to several species with commercial value (Alongi 1998). The vegetation in this type of ecosystem is usually subjected to harsh environmental conditions including periodic tidal flooding, high salinity levels, and contaminants such as metals from urban and industrial effluents. As a consequence, the typical salt marsh vegetation presents low species diversity and high specialization (Woerner and Hackney 1997).

Ria de Aveiro, a costal lagoon on the northwest of Portugal, is one of the most mercury-contaminated coastal systems in Portugal. For over four decades (1950–1994), industrial effluents containing mercury (Hg), arsenic (As), zinc (Zn), lead (Pb), and other contaminants were discharged into natural water streams of this lagoon which led to contamination of water and sediments of adjacent marshes (Costa and Jesus-Rydin 2001). Although discharges have substantially decreased over the last two decades, sediments in the Laranjo Bay salt marshes are still considerably contaminated, namely with Hg. This highly toxic nonessential metal is a hazardous contaminant with potential to be released to the water column and assimilated by the biota (Pereira et al. 2009).

The perennial halophyte Halimione portulacoides (L.) Aellen is one of the most productive and abundant species in European salt marshes (Bouchard et al. 1998), including those along the Ria de Aveiro lagoon. This Hg-tolerant halophyte allocates this toxic metal preferably to belowground tissues and was proposed as a bioindicator and biomonitor for Hg pollution (Válega et al. 2008a). Additionally, this halophyte has been proposed for phytoremediation of Hg (Anjum et al. 2011), tributyltin (Carvalho et al. 2010), and petroleum hydrocarbons (Couto et al. 2011).

Endophytic bacteria colonize internal plant tissues without causing symptoms of disease (Lodewyckx et al. 2002). These bacteria are highly diverse taxonomically, ubiquitous from herbaceous plants to trees, and contribute to plant health and productivity (Berg et al. 2014). Endophytes present a great potential for applications in agricultural, medical, biotechnological, and industrial fields. Industrially important enzymes such as amylases, cellulases, lipases, pectinases, proteases, and xylanases, among others, have previously been found in endophytic bacteria (e.g., Lodewyckx et al. 2002).

Plant-associated bacteria, including endophytic bacteria, are known to be very relevant for plant growth promotion (PGP; Hardoim et al. 2008), including in cases of biotic or abiotic stresses, such as contaminated soils (reviewed in Rajkumar et al. 2012). This PGP action may be conducted through direct or indirect mechanisms. Direct PGP mechanisms include enhancement of plant nutrient acquisition, by aiding in nitrogen fixation and phosphate solubilization; plant growth by production of the auxin indol-3-acetic acid (IAA); and lowering of ethylene-related plant stress by producing the enzyme 1-aminocyclopropane-1-carboxylate (ACC) deaminase. Indirect PGP is achieved by preventing phytopathogen prejudicial effects on plant development by producing metabolites involved in antimicrobial activities and iron chelation (reviewed in Bulgarelli et al. 2013). It is also known that plant-associated bacteria are able to aid in phytoremediation processes by phytoextraction (reviewed in Sessitsch et al. 2013). In fact, Oliveira et al. (2014a) have shown that endophytic bacteria isolated from H. portulacoides have potential to be used in detoxification of petroleum hydrocarbons.

Notwithstanding the acknowledged importance of endophytic bacteria, comprehensive studies regarding the basal culturable communities of these bacteria in the specific environment of salt marshes are scarce, as studies are usually directed toward studying contaminant-degrading bacteria. H. portulacoides is a highly relevant and ecologically important halophyte in salt marshes; however, the structure of endophytic bacterial communities of this plant has not yet been fully explored, only its fraction that is able to degrade hydrocarbons (Oliveira et al. 2014a).

This study aimed to characterize the culturable endophytic bacterial communities of this halophyte from a highly contaminated site, an intermediately contaminated site, and a noncontaminated site in Ria de Aveiro, regarding their phylogenetic diversity and traits associated with direct and indirect PGP effects.

Methods

Study area and sampling

The study was conducted in the estuary Ria de Aveiro, in Aveiro, Portugal. A confined area in this lagoon was subjected to Hg-rich industrial discharges for four decades (1950s–1990s, Pereira et al. 1998). Three locations in the salt marsh of Ria de Aveiro were assessed: one highly contaminated by industrial activity (site B, 450 m downstream from the contamination source in the estuary, 40° 43′ 47.46″ N, 8° 36′ 45.40″ W), one with intermediate levels of contamination (site E, 2500 m downstream from the contamination source, 40° 43′ 15.90″ N, 8° 38′ 15.76″ W) (Válega et al. 2008b), and a noncontaminated location (site C, 40° 38′ 05.36″ N, 8° 39′ 38.87″ W; 11.3 km from the contamination source).

Healthy specimens of the halophyte H. portulacoides were sampled in monospecific stands, at least in triplicate in each site. Adjacent sediment was sampled in triplicate with a 3 cm in diameter and 13-cm depth corer. Sampling was performed during low tide in November 2012 (average temperature 16 °C). Samples were transported to the laboratory under refrigerated conditions and promptly processed.

Metal(loid) quantification

Composite samples of aboveground (AG) tissues (leaves, stems, and flowers; 1 g fresh weight), belowground (BG) tissues (roots; 1 g FW), and sediment (2 g FW) from each site were used for metal(loid) quantification, in triplicate. Chromium (Cr), nickel (Ni), copper (Cu), zinc (Zn), As (as total arsenic), and mercury (Hg) were quantified. Samples were subjected to acid digestion with concentrated nitric acid overnight at 115 °C (Figueira and Freitas 2013) and analyzed by inductively coupled plasma mass spectrometry (ICP-MS) according to the International Standard ISO 17294 by a certified laboratory at the University of Aveiro.

One-way ANOVA (function aov) combined with post hoc Tukey’s honest significant differences method (function TukeyHSD) were used to assess differences in metal(loid) concentrations in sediment and plant tissue samples on all sites. Welch two-sample t test (function t.test) was used to estimate differences in metal(loid) concentrations (i) among plant tissues, for each isolation site and (ii) between sediment and plant samples, for each isolation site. These analyses were performed using the stats package in the free software R (R Core Team 2014) with abovementioned functions aov, TukeyHSD, and t.test, considering a 95 % confidence interval.

Isolation of endophytic bacteria, growth media, and conditions

Eleven plants were processed at the laboratory: five from site C, three from site B, and three from site E. First, all plants were thoroughly washed with tap water. For each plant, 5 g (FW) of AG tissues and 2 g (FW) of BG tissues were used for endophytic bacteria isolation. Surface sterilization was optimized and performed as follows: samples were sequentially immersed in 50 mL of phosphate-buffered saline (PBS; 10 min), 96 % ethanol (1 min), 5 % NaOCl (30 min), 96 % ethanol (1 min), and rinsing in distilled sterile water three times. Surface sterilization was confirmed by inoculating water from the last wash in Tryptic Soy Agar (TSA, Merck, Germany), Marine Agar (MA, Difco Laboratories, France), and R2A (Merck, Germany) culture media. The plates were examined for bacterial growth after incubation at 28 °C for 72 h.

Tissues were then ground with a pestle in a mortar containing 10 mL of PBS. Serial dilutions were prepared in PBS and spread, in duplicate, on TSA, MA, and R2A media. The plates were incubated up to 72 h at 28 °C and were observed every 24 h for colony-forming unit (CFU) count. Morphologically distinct endophytic bacterial colonies were selected and purified by streaking, resulting in a collection of 665 endophytic bacterial isolates, and stored at −80 °C in 20 % glycerol. Function aov combined with TukeyHSD was used to assess differences in average CFU per gram of tissue (fresh weight) in different isolation sites and culture media, and t.test was used to assess tissue-based differences.

DNA extraction, PCR-based fingerprinting, and selection of representative isolates

DNA from all 665 isolates was extracted using the Genomic DNA Purification kit #0513 (Thermo Scientific, USA) according to the manufacturer’s instructions. BOX-PCR with BOXA1 primer (5′-CTACGGCAAGGCGACGCTGAC-3′; Versalovic et al. 1994) was used for molecular typing of all isolates. PCR reactions were performed in a volume of 25 μL, using 5× Green GoTaq buffer, GoTaq DNA polymerase (5U μL−1), MgCl2 (25 mM), and a dNTP mixture (2 mM) (Promega, USA). Conditions for amplification were as follows: one cycle at 95 °C (7 min), 30 cycles at 94 °C (1 min), 53 °C (1 min), 65 °C (8 min), and a final cycle at 65 °C (16 min). One hundred thirty-three isolates did not yield a clear profile with this methodology and were typed using ERIC-PCR with ERIC1R (5′-ATGTAAGCTCCTGGGGATTCAC-3′; Versalovic et al. 1991) and ERIC2 (5′-AAGTAAGTGACTGGGGTGAGCG-3′; Versalovic et al. 1991) primers. Conditions for PCR amplification were the same as for BOX-PCR, except the annealing temperature, which was 52 °C.

Analysis of the genetic fingerprinting patterns was performed with GelCompar II software (Applied Maths, Belgium). The Pearson correlation coefficient was applied, and cluster analysis was performed using the unweighted pair group method with arithmetic mean (UPGMA) algorithm. The resulting dendrograms were analyzed in order to obtain groups of isolates with at least 85 % similarity. This cutoff was determined so that patterns that were known to be equal (molecular-weight size markers) would be considered to be in the same group. Representative isolates of each group were randomly selected. A total of 467 representative isolates were obtained: 419 from molecular typing and 48 isolates that did not yield clear profiles by either BOX- or ERIC-PCR.

16S rRNA gene amplification, phylogenetic affiliation, and multivariate analyses

Four hundred sixty-seven representative isolates were subjected to PCR amplification of 16S rRNA gene using the universal primers 27f (5′-AGAGTTTGATCCTGGCTCAG-3′; Lane 1991) and 1492r (5′-GGTTACCTTGTTACGACTT-3′; Lane 1991) and the NZYTaq 2× Green Master Mix (NZYTech, Portugal) in 25-μL reactions: one cycle at 94 °C (5 min), 30 cycles at 94 °C (1 min), 55 °C (1 min), 72 °C (1.5 min), and a final cycle at 72 °C (10 min). Five isolates did not yield quality sequences after several attempts and were excluded from the analysis. The remaining 462 amplification products were purified with the DNA Clean & Concentrator™-5 (Zymoresearch, USA) according to manufacturer’s instructions. Amplicons were partially sequenced with primer 27f using GATC Biotech services (Germany) according to company specifications. The partial 16S rRNA gene sequences obtained in this study were submitted to GenBank and are available under the accession numbers KT324749 to KT325209.

Quality sequences with an overall average of 893 bp were examined and edited with FinchTV V1.4.0 software (Geospiza, USA). The Basic Local Alignment Search Tool (BLAST; Altschul et al. 1997) and the EzTaxon server (Kim et al. 2012) were used to obtain taxonomic classification up to the genus level. Information from the representative isolates was used to associate taxonomic affiliation to the respective 658 isolates from the collection. Sequences were then subjected to the Ribosomal Database Project (RDP) pyrosequencing pipeline (Cole et al. 2014), where they were aligned and clustered into operational taxonomic units (OTUs) considering a distance limit of 0.03 (97 % similarity). Consequently, an OTU table was built with abundances for 21 samples: five for CAG—community from AG tissues from site C; five for CBG—BG tissues from site C; two for EAG—AG tissues from site E; three for EBG—BG tissues from site E; three for BAG—AG tissues from site B; and three for BBG—BG tissues from site B. Three AG samples were collected from site E; however, only two yielded culturable bacteria. As such, only those were considered for subsequent analyses.

Using R version 3.1.1 (R Core Team 2014), several analyses were performed to characterize the bacterial communities regarding their OTU abundance. Diversity indices Shannon’s H′ diversity and Pielou’s evenness were assessed with functions diversity and with diversity/log(specnumber), respectively, from the vegan package (Oksanen et al. 2015). The multipatt function from package indicspecies (De Caceres and Legendre 2009) was used to find indicator species (here, OTUs) in samples from different sites, tissues and communities, using a 10 % significance level for selecting indicators. The species-site group association parameter “IndVal.g” was used, since it corrects for unequal group sizes.

The variation in OTU composition among the samples was analyzed by assessing differences in a dissimilarity matrix, built with Bray-Curtis dissimilarity measure (function vegdist from package vegan in R) after transforming the OTU abundance table with square root (function sqrt in R). PERMANOVA was used to analyze variances in the Bray-Curtis dissimilarity matrix using the adonis function (from package vegan in R) on the interaction between isolation site and tissue, and to both variables independently. For this analysis, the number of permutations was 999 in all tests.

Information in the Bray-Curtis dissimilarity matrix was then graphically assessed with a principal coordinate analysis (PCoA) using cmdscale function (from package vegan in R) to create the PCoA ordination. The significance of the observed patterns in the PCoA was analyzed by assessing differences in sample scores across PCoA axis 1 (PCoA1) and PCoA axis 2 (PCoA2). The normality test shapiro.test (stats package in R) revealed nonnormality in PCoA1 (Shapiro-Wilk: W = 0.8848, p value < 0.05) and normality in PCoA2 (Shapiro-Wilk: W = 9352, p value > 0.1). As a consequence, patterns in PCoA1 were assessed with Kruskal-Wallis rank sum test (kruskal.test from stats package in R) combined with posthoc.kruskal.nemenyi.test (from PMCMR package in R; Pohlert 2005) for samples from different sites and communities, and with Wilcoxon rank sum test (wilcox.test from stats package in R) for samples from different tissues. PCoA2 patterns were analyzed with aov for samples from different sites and communities, and with t.test for samples from different tissues.

Screening of endophytic bacteria for enzymatic activity

The 467 representative isolates were individually screened, in triplicate, for different enzymatic activities: amylolytic, cellulolytic, lipolytic, pectinolytic (at pH 5.0 and 7.0), proteolytic, and xylanolytic. The ability of the isolates to degrade a substrate was evaluated using 0.2 % starch, 0.5 % carboxymethyl cellulose, 1 % Tween 20, 0.5 % pectin, 1 % skim milk, or 0.5 % xylan, respectively, on TSA or MA culture medium. The basal culture medium (TSA or MA) was selected for each isolate, according to where best bacterial growth was obtained. Tween 20 and skim milk solutions were sterilized separately and only added to the culture media after cooling down. The agar medium plates were inoculated with 5 μL of a bacterial suspension with optical density (OD) at 600 nm = 0.8 U and incubated at 30 °C until sufficient growth was obtained, or up to 120 h.

The enzymatic activities were assessed according to the presence (positive result) or absence (negative result) of a halo surrounding the bacterial growth. Amylolytic activity was detected when a clear halo was observed upon addition of Lugol solution to the agar plate. Cellulolytic and xylanolytic activities were detected when a clear halo was observed upon incubation with a 1 % Congo red solution (with 10 % ethanol) and subsequent wash with 1 M NaCl. Lipolytic activity was observed by the formation of a white precipitate around the bacterial growth. Pectinolytic activity was detected by formation of a clear halo after addition of 1 % hexadecyltrimethylammonium bromide (CTAB) solution. Proteolytic activity was revealed by visualization of a clear halo against an opaque background. The results were interpreted according to the visibility of the halo obtained: negative, no halo observed; positive, faint to intense halo observed.

Screening of endophytic bacteria for plant growth promotion traits

Representative isolates were individually screened for different PGP traits: ACC deaminase activity, IAA production, phosphate solubilization, and siderophore production. Isolates were incubated in adequate growth media at 30 °C until sufficient growth was obtained. Unless otherwise stated, strains for negative and positive controls were studied in Pereira et al. (2013).

To determine ACC deaminase activity, isolates were grown in DF salt minimal medium (Dworkin and Foster 1958) or variations of this medium. Basal DF was prepared as follows (per liter): 4 g KH2PO4, 6 g Na2HPO4, 0.2 g MgSO4·7H2O, 2 g glucose, 1 mg FeSO4·7H2O, 10 μg H3BO3, 10 μg MnSO4, 70 μg ZnSO4, 50 μg CuSO4, 10 μg MoO3, 12 g agar, pH 7.2. This medium was either supplemented with 3 mM ACC (DF + ACC; ACC was sterilized by filtration and added to cooled down medium) or 2 g L−1 (NH4)2SO4 (DF + ammonium sulfate) or no nitrogen source (DF). The ability to use ACC as a sole nitrogen source is a consequence of the activity of the enzyme ACC deaminase (Penrose and Glick 2003). Pseudomonas putida EAPC8 and Arthrobacter rombi EC32A were used as positive controls for ACC deaminase activity.

To test for IAA production, the method of Gordon and Weber (1951) was used: for each isolate, 500 μL of supernatant obtained from cultures grown in MB or TSB + l-tryptophan (1 %) was mixed with 1 mL of Salper solution. After incubating in the dark for 30 min, OD was read at 535 nm. Using a calibration curve of pure IAA as a standard, IAA concentrations (μg mL−1) were obtained for each isolate. Pseudomonas putida EAPC8 and Pseudomonas fluorescens S3X were used as positive controls for IAA production.

The ability of isolates to solubilize phosphate was screened on National Botanical Research Institute’s Phosphate (NBRIP) solid growth medium (Nautiyal 1999). After inoculation and appropriate incubation, phosphate solubilization was observed as a clear halo surrounding the bacterial growth. Pseudomonas reactans EDP28, P. putida EAPC8, A. rombi EC32A, and P. fluorescens S3X were used as negative controls for phosphate solubilization, and Arthrobacter nicotinovorans EAPAA and Rhodococcus sp. EC35 were used as positive controls.

In order to test for siderophore production, all glass materials used were deferrated by treating with 6 % nitric acid and extensively washing with sterile double-distilled water. Culture media deferration was conducted on TSA and MA medium as described by Cox (1994). After inoculation and incubation on either medium, the plates with bacterial growth were overlaid with CAS (O-CAS; Pérez‐Miranda et al. 2007) and incubated with this overlay for 2 h. Siderophore production was detected as a change of color of the overlay from blue to orange or purple. This assay was only performed with TSA medium, since the O-CAS reacted with MA medium preventing a clear reading of the result. Pseudomonas reactans EDP28, P. putida EAPC8, A. rombi EC32A, P. fluorescens S3X, A. nicotinovorans EAPAA, and Rhodococcus sp. EC35 were used as positive controls for siderophore production.

The representative isolates were also assessed for their nitrogen fixing abilities by PCR-screening for the presence of nifH gene: the primers IGK3 and DVV (Ando et al. 2005), described as the best pair of primers to use in nifH amplification (Gaby and Buckley 2012) were used with the NZYTaq 2× Green Master Mix (NZYTech, Portugal) in 25 μL reactions. Conditions for PCR amplification were as follows: initial denaturing at 95 °C for 10 min, followed by 35 cycles of denaturing at 95 °C for 45 s, annealing at 56 °C for 30 s, and extending at 72 °C for 1 min. A final extension was performed at 72 °C for 10 min. Rahnella aquatilis M72troncoD (HQ538817; Proença et al. 2010) was used as a positive control, and Escherichia coli ATCC25922 was used as a negative control (Gaby and Buckley 2012). The resulting PCR products were observed in a 2 % agarose gel electrophoresis.

Results

Metal(loid) quantification of sediments and plant tissue

Five metals (Cr, Ni, Cu, Zn, and Hg) and the metalloid As were quantified from samples collected at sites C, E, and B, in sediment samples from monospecific stands of H. portulacoides, in AG and BG tissues. The results are presented in Table 1. For some replicates of Hg concentration determination in sites C and E, the obtained values were below detection limits (<2 μg L−1). In these cases, half of the lowest detectable concentration was used in statistical analyses.

Table 1 Metal(loid) concentrations in sediment and plant tissue samples
Full size table

Sediment samples collected from sites E and B presented concentrations at least 3-fold higher than samples from site C for all metal(loid)s tested (p value < 0.05). For As and Hg, a contamination gradient could be observed, where site B (closest to the effluent discharge) had the most contaminated sediment samples and site E presented intermediate contamination (p value < 0.05). Remarkably, sediment Hg concentration was, on average, 31 (site E) to 135 (site B) times higher than in samples from site C. All sediment samples were more contaminated than plant tissue samples, in all sites, for all tested metal(loid)s (p value < 0.05), except Hg in site C, where the differences in the metal concentration did not present statistical significance.

Considering metal(loid) contents in each site separately, we observed that in site B, metal(loid)s concentration in AG tissues were significantly higher than that in BG tissues. The same was observed for Ni in site E (p value < 0.05). Remarkably, Hg contamination of BG tissues in site B was 33-fold higher than in site C.

Considering metal(loid) contents in each type of tissue separately, we saw that BG tissues were significantly more contaminated with Cr, Ni, and Hg in site B than in site C. In the specific cases of As in AG tissues and Hg in both tissues, a significant gradient was observed: C < E < B (p value < 0.05).

Phylogenetic diversity of endophytic bacteria, community structure, and multivariate analyses

Endophytic bacteria from sites C, E, and B were isolated from surface sterilized AG and BG tissues from eleven plants, in TSA, MA, and R2A culture media. Extensive results for CFU g−1 (FW) for all sampled plants are stated in Supplementary Table S1. AG tissues from site C presented significantly more CFU g−1 than BG tissues (p value < 0.05). The opposite was observed in sites E (p value < 0.05) and B (only for TSA culture medium; p value < 0.05).

The selection of morphologically distinct colonies from all culture media, plant tissues, and sampling sites resulted in a collection of 665 endophytic isolates. Molecular typing of the collection yielded 467 representative isolates (150 from CAG community, 74 from CBG, 18 from EAG, 94 from EBG, 33 from BAG, and 97 from BBG), for which partial sequences of 16S rRNA gene were obtained. Five isolates did not yield quality sequences after several attempts and were excluded from the analysis, resulting in 462 amplification products. Taxonomic classification yielded matches with similarity ranging from 92.79 to 100 % similarity. Twenty-nine of 462 isolates (6.3 %) exhibited an identity percentage lower than 97 with the closest match, suggesting the potential for novel species in the collection. Taxonomic information was then affiliated to the respective 658 isolates.

The 658 isolates were distributed amongst class-level identification and grouped according to isolation site and tissue in Fig. 1. The phylum Proteobacteria (mainly classes alpha- and gamma-Proteobacteria) clearly dominated the collection of endophytic bacteria (64.29 % of 658 isolates; Fig. 1 and Supplementary Table S2), followed by Actinobacteria (22.8%) and Firmicutes (10.33%). The phylum Bacteroidetes was also present, but less represented (2.58 %). Figure 1 also shows that alpha-Proteobacteria are highly represented in all six communities, and that gamma-Proteobacteria, although present in all communities, are most abundant in site C. Isolates from classes Bacilli and Actinobacteria are present in all communities, except Actinobacteria in the community with the lower number of isolates (EAG, 22 isolates).

Fig. 1
figure 1

Distribution of 658 endophytic isolates by class amongst isolation sites (B, E and C) and plant tissues (aboveground and belowground). a-, b- and g-Proteobacteria refers to alpha-, beta- and gamma- Proteobacteria, respectively

Full size image

The most common families of bacteria in the collection were Vibrionaceae (51 of 658 isolates, 7.75 %) Microbacteriaceae (85/658 isolates, 12.92 %), Erythrobacteraceae (87/658 isolates, 13.22 %), and Halomonadaceae (98/658 isolates, 14.89 %). The complete distribution of the 658 isolates across 79 genera in all six communities is presented in Supplementary Table S2. Overall, the most abundant genera were Altererythrobacter, Marinilactibacillus, Microbacterium, Salinicola, and Vibrio, with 49, 42, 67, 58, and 51 isolates, respectively. Altererythrobacter was the only genus present in all six communities. Isolates belonging to the genera Altererythrobacter, Bacillus, Citromicrobium, Demequina, Labrenzia, Marinilactibacillus, Martelella, Microbacterium, Micrococcus, Pseudomonas, Sphingorabdus, Stakelama, Vibrio, and Zunongwangia were present in all sites. Isolates belonging to 29 of the 79 genera were present in both BG and AG tissues.

Notably, all 58 isolates (8.8 % of the collection) associated with the genus Salinicola were exclusively isolated from site C. Hoeflea, Labrenzia, and Microbacterium were exclusive and considerably dominant in BG tissues, with 17 (2.6 % of the collection), 22 (3.3 %), and 67 (10.2 %) isolates, respectively.

An OTU table was built with abundances across all samples, using the RDP pyrosequencing pipeline. The OTUs were defined considering a distance limit of 0.03 (97 % similarity) and their distribution across communities is stated in Table 2. Analysis of indicator OTUs showed that the OTUs, here annotated with the corresponding genera, that contributed most for the characterization of each sampling site were as follows: OTU_005 (Salinicola spp., 58 isolates, p value < 0.05) for site C, and OTU_077 (Bacillus spp., 11 isolates, p value < 0.05) and OTU_028 (Rhizobium spp., 6 isolates, p value < 0.05) for site B. This analysis also showed that indicator OTUs for the isolation tissue were as follows: OTU_104 (Zunongwangia spp., 5 isolates, p value < 0.05) for AG tissues, and OTU_033 (Labrenzia spp., 22 isolates, p value < 0.05) and OTU_219 (Microbacterium spp., 44 isolates, p value < 0.05) for BG tissues.

Table 2 Analysis of 16S rRNA gene partial sequences with RDP pipeline
Full size table

A Bray-Curtis dissimilarity matrix was computed on a square root-transformed OTU table which allowed comparison of bacterial communities of all samples. The resulting matrix was tested for significant differences with PERMANOVA (adonis function in R). The interaction between isolation site and tissue was significant (adonis: F 2,15 = 1.38, R 2 = 0.12, p value < 0.05), as was the test for each independent variable: isolation site (adonis: F 2,18 = 1.65, R 2 = 0.14, p value < 0.01) and tissue (adonis: F 1,19 = 2.08, R 2 = 0.09, p value < 0.01). However, only 23 % of the variation in the bacterial composition was explained by both variables and 12 % by their interaction.

A PCoA plot was generated based on the Bray-Curtis dissimilarity matrix to compare all bacterial communities (Fig. 2). The PCoA plot shows a clear gradient across axis PCoA1 which coincides with the presence of contamination, as samples with most contamination (BG tissues from sites B and E, see Table 1) are grouped in the left side of the plot. Analysis of sample scores across PCoA1 revealed that there are significant differences in OTU composition in isolation sites (Kruskal-Wallis chi-squared = 7.66, df = 2, p value < 0.05) and in isolation tissues (Wilcoxon rank sum test: W = 93, p value < 0.01). This was not observed in PCoA2 for either isolation sites (ANOVA: F 2,18 = 2.53, p value > 0.1) nor tissues (t test: t = −1.8, df = 14.97, p value > 0.05).

Fig. 2
figure 2

Principal coordinates analysis (PCoA) plot based on 16S rRNA gene partial sequences from twenty one endophytic bacterial samples from all isolation sites and tissues. Distances between samples were calculated with Bray-Curtis dissimilarity measure on a square root-transformed OTU table. PCoA axes PCoA1 and PCoA2 are plotted with the corresponding percentages of explained variation. CAG aboveground (AG) tissues from site C, CBG belowground (BG) tissues from site C, EAG AG tissues from site E, EBG BG tissues from site E, BAG AG tissues from site B, BBG BG tissues from site B

Full size image

Enzymatic activity and plant growth promotion traits in endophytic bacteria

Assays for enzymatic activities and PGP traits were successful for 89–100 % of the representative isolates. Test success for siderophore production was 85.2 %; the remaining isolates (n = 69) were not able to grow in TSA medium, only MA, where a reaction occurred with the CAS-overlay and interfered with the interpretation of the result. Screening results are stated in Table 3 and the extensive list of results for all isolates is in Supplementary Table S3.

Table 3 Screening of bacterial endophytes for enzymatic activity and plant growth promotion traits
Full size table

Only 5.6 to 22.1 % of isolates tested positive for pectinolytic activity. A great percentage of isolates tested positive for all other assays (up to 73.2 %), especially IAA production and ACC deaminase activity (98.9 and 51.2 %, respectively). IAA production was detected in 464 isolates, with concentrations ranging from 0.48 to 206 μg mL−1. Seventeen IAA producers presented high levels of IAA concentration, over 100 μg mL−1, 15 of which belonged to Proteobacteria phylum (5 Pseudomonas sp.).

Isolates belonging to six genera (one Chromohalobacter, one Halomonas, one Marinomonas, one Marinospirillum, one Micrococcus, and seven Vibrio) tested positive for amylolytic, cellulolytic, lipolytic, pectinolytic (at pH 7), proteolytic, and xylanolytic activities. From the 97 isolates that tested positive for nifH amplification, 74 belonged to the Proteobacteria phylum, including 19 Vibrio sp. isolates. The most noteworthy genera were Altererythrobacter (from the 31 isolates of this genus, 29 were lipolytic and 24 xylanolytic, 22 presented both of these activities), Marinilactibacillus (24 cellulolytic, 18 proteolytic and 39 xylanolytic isolates, out of 41 isolates from this genus), Microbacterium (32 amylolytic, 30 cellulolytic and 29 xylanolytic isolates, from 39 Microbacterium sp. isolates), and Vibrio (14 to 32 isolates producing every enzyme tested, from a total of 33 Vibrio sp. isolates). Isolates associated to five genera (one Altererythrobacter, one Ensifer, one Oceanibulbus, one Stakelama, and two Vibrio) displayed positive results for all tested PGP traits assays (Supplementary Table S3).

The community-based and class-level distribution of representative isolates that tested positive for the assayed enzymatic activities and PGP traits are represented in Fig. 3 and Supplementary Fig. S1, respectively. The class-level distribution of isolates that tested positive in all assayed enzymatic activities and PGPs was a reflection of the class-level and community-based distribution of all isolates of our collection (Fig. 1 and Table 2, respectively), i.e., the classes with higher representability in the endophytic collection were also those that presented more isolates that tested positive for the assayed enzymes and traits. This suggests that the probability of the presence of traits in a given class depends on the number of isolates in that class, instead of depending on the site or tissue of isolation.

Fig. 3
figure 3

Distribution of isolates that tested positive for enzymatic activity assays and plant growth promotion traits from 467 representative isolates. Isolates are grouped according to isolation community (site and tissue). n represents the number of isolates that tested positive for each trait. Amylolytic, cellulolytic, lipolytic, pectinolytic (at pH 5.0 and 7.0), proteolytic, and xylanolytic activities are represented; ACC deaminase (as growth in DF + ACC medium), IAA production, presence of nifH gene, ability to solubilize phosphate, and siderophore production are also represented. CAG aboveground (AG) tissues from site C, CBG belowground (BG) tissues from site C, EAG AG tissues from site E, EBG BG tissues from site E, BAG AG tissues from site B, BBG BG tissues from site B

Full size image

Discussion

The main aim of this study was to elucidate the structure, diversity, and function of culturable endophytic bacterial communities from the salt marsh plant H. portulacoides. To our knowledge, this is the first study regarding the isolation, identification, and functional characterization of endophytic bacteria in healthy specimens of the halophyte H. portulacoides. Bacteria were isolated from AG and BG tissues from the halophyte, collected from two sites contaminated with industrial effluents and one noncontaminated control site.

Several authors have verified that a gradient of contamination exists, from Laranjo Bay to nearshore, especially regarding sediment Hg contamination (reviewed in Pereira et al. 2009). In the present study, sediment samples presented higher metal(loid) contents in sites closest to the industrial effluent discharges (sites B and E). In fact, for As and Hg, a significant concentration gradient could be observed, where content in B > E > C. These metal(loid)s were among the known main contaminants in the industrial effluents in Laranjo Bay, as is stated in the report by Costa and Jesus-Rydin (2001).

Several studies have found that H. portulacoides mainly accumulates metals in the BG tissues and that BG to AG translocation is very low (e.g., Válega et al. 2008a; Válega et al. 2008b; Anjum et al. 2011). The same pattern was observed in BG and AG tissues in this study, for all tested metals in site B (up to 6 times more metal content than in AG tissues) and for Ni in site E (average of 4.5 times more Ni in BG than in AG tissues). This allocation pattern was described by Lozano-Rodriguez et al. (1997) as a strategy for protecting the more sensitive AG tissues from the effects of metal-induced stress.

We analyzed a much higher number of isolates than what is usually observed in studies focusing endophytic bacteria. This could be attributed to the fact that we used three culture media with different nutritional compositions and that the plate incubation period for the isolation of the endophytic bacteria was long (72 h). Sample sizes were considerably smaller for EAG and BAG communities and, as a consequence of this unbalance dataset, analyzing indicator OTUs may raise concerns. However, when using the parameter “IndVal.g” in the indicspecies package in R, there is a correction for unequal group sizes in the calculations for indicator species (OTUs in our case). The quantity “A,” used in calculating the IndVal, is obtained by using the mean abundance of the species in the target site group divided by the sum of the mean abundance values over all groups. In using the sum of the mean abundance values over all groups, instead of using real abundances from all groups, the unequal sizes of the site groups are controlled for (De Caceres and Legendre 2009). Regarding the statistical analysis performed based on a dissimilarity measure (Bray-Curtis), we minimized the most relevant intrinsic bias of this measure (giving too much importance to dominant OTUs and overseeing rare ones) by applying a square root transformation to our OTU abundance table prior to obtaining the Bray-Curtis dissimilarity matrix. The selection of the index at hand was based on (i) its applicability to assess differences between communities using abundance data, (ii) its wide use in the field of ecology and microbiology for this purpose, and (iii) its suitability to answer our questions.

We found considerable culturable bacterial diversity in the internal tissues of H. portulacoides from all sampled sites. It is widely accepted that the culturable fraction of plant-associated microbes corresponds to a small fraction of the real diversity, as a consequence of growth medium inadequacy (Tanaka et al. 2014), and limitations to reproduce optimal growth conditions. Nevertheless, the culture-dependent approach allows for extensive functional characterization of the isolates in the collection, which is not possible with culture-independent approaches. Additionally, traditional methods of culture-dependent techniques are useful for the determination of the overall effect of metal contaminants in bacterial communities (Ellis et al. 2001).

As stated above, in our study, the number of isolates obtained from EAG and BAG samples was considerably lower than for other samples. Since the method for obtaining endophytic bacteria includes a harsh surface sterilization process, the viability of endophytes in the samples could be compromised. However, we do not believe this to be the main reason for such a disparity in the number of isolates in the different samples, since the same process was applied to all samples. The isolation efforts were the same for all samples, and as such, the difference in numbers of obtained endophytic isolates may be a reflection of these AG tissues being less populated, since the highest bacterial densities are usually observed in BG tissues, and decrease toward AG tissues (Lodewyckx et al. 2002). In fact, we observed a significantly lower average number of CFU per gram (FW) obtained from EAG and BAG samples, when compared with the respective BG tissues (Supplementary Table S1). This pattern was not, however, observed for AG and BG tissues from site C, where the CAG community presented a significantly higher average number of CFU per gram (FW) than CBG. This suggests that the degree of soil and plant tissue contamination could also play a role in the average number of culturable bacteria. This latter effect has been previously reported for bacterial soil communities, where the presence of metal contaminants negatively influenced the ability of readily culturable bacteria to replicate in laboratory media (Ellis et al. 2003).

Proteobacteria, Actinobacteria, Firmicutes, and Bacteroidetes were the phyla recognized in the endophytic collection from this study. This result is in accordance with studies describing the phyllosphere and endosphere assemblages in other plant hosts (reviewed in Bulgarelli et al. 2013), including halophytes (Mora-Ruiz et al. 2015), and the rhizosphere of H. portulacoides (Oliveira et al. 2014b). The genera Altererythrobacter, Marinilactibacillus, Microbacterium, Salinicola, and Vibrio were the most abundant in our collection. To our knowledge, this is the first time an isolate belonging to the genus Marinilactibacillus was obtained from the endosphere.

Isolates affiliated to the genus Altererythrobacter have been previously identified in the endosphere (e.g., Liu et al. 2014). In our collection, 73.5 % of isolated Altererythrobacter spp. were collected from contaminated sites (E and B). Some species of Altererythrobacter are described as able to resist metals (Wu et al. 2014), a trait that could be useful in the endosphere of plants colonizing contaminated locations.

Marinilactibacillus spp. have been detected in marine environments (e.g., Toffin et al. 2005) and detected with culture-independent methods in olive fermentation brine (Lucena-Padrós et al. 2015). Isolates belonging to this genus have also been isolated from many types of cheese, including as part of consortia that were able to reduce the counts of the human pathogen Listeria monocytogenes (e.g., Roth et al. 2010), indicating antimicrobial activities which could be useful in the endosphere environment.

Isolates from the genus Microbacterium, exclusively isolated from BG tissues in this study, have been thoroughly characterized and detected in many environments, including the rhizosphere of ginseng (Kim et al. 2015) and the endosphere of crops and prairie plants (Zinniel et al. 2002). Recently, three novel species from this genus were described from the endosphere of H. portulacoides (Alves et al. 2014; Alves et al. 2015). PGP traits have been previously associated with this genus (see below), suggesting a high importance in the endosphere.

Isolates from the genus Salinicola have been previously obtained from diverse environments (e.g., deep sea surface sediments, Romanenko et al. 2013) including the endosphere of a halophyte (Mora-Ruiz et al. 2015). To the best of our knowledge, PGP traits have not been explored in this genus. The prosperity of Salinicola spp. in marine environments could be related to their presence in the endosphere of the halophyte at study.

Vibrio spp. have been abundantly detected in aquatic environments and organisms (Thompson et al. 2004) and in the endosphere (e.g., seagrass, Jose et al. 2014). PGP traits have been associated with this genus (see below), suggesting that this widely known pathogen in humans and other animals may have a positive role in the endosphere.

The genera Pseudomonas and Micrococcus, less represented in this study (15 and 9 isolates, respectively), were previously detected in the endosphere of H. portulacoides in a study regarding hydrocarbon-degrading bacteria (Oliveira et al. 2014a).

From the 462 isolates retrieved in this study, 29 isolates presented less than 97 % similarity with the closest 16S rRNA gene sequence in GenBank. These isolates may potentially belong to novel species, revealing the endosphere of H. portulacoides as a source for undescribed phylogenetic groups.

As reviewed in Bulgarelli et al. (2013) for many plant hosts and discussed in Oliveira et al. (2014b) for H. portulacoides, plant-related factors contribute to the composition of plant-associated bacterial communities. In this work, 14 genera were isolated from all sites, suggesting the presence of a common structure associated to the studied halophyte. On the other hand, differences were observed, where considerably abundant genera were exclusively isolated from one site (e.g., Salinicola spp. from site C).

In the present study, Altererythrobacter spp., Marinilactibacillus spp., Microbacterium spp., and Vibrio spp. isolates were abundantly detected in all sites, suggesting an ability to resist metals, which has been previously described in these genera (Wu et al. 2014; Sheng et al. 2008; Jafari et al. 2015) except, to the best of our knowledge, for the genus Marinilactibacillus. OTUs belonging to the genera Bacillus and Rhizobium, albeit less represented in the endophyte collection (11 and 6 isolates, respectively), were considered indicator OTUs for site B, and metal resistance has previously been associated with these genera (Ma et al. 2015; Aafi et al. 2015). Isolates belonging to Salinicola spp. (n = 58) were exclusively detected in site C, and the indicator OTU for site C was also attributed to this genus. Information on whether Salinicola spp. can tolerate metal contamination is lacking in the literature; however, considering the number of isolates obtained from this genus and the extension of our collection, we suggest that it should be further investigated as to its ability as a bioindicator.

Usually, bacteria access plant tissues through natural entry points in the root tips and lateral roots emergence points, as a consequence of root growth. Once inside the plant, endophytes may colonize the plant by migration from BG to AG tissues. Additional penetration may occur through natural openings on leaves such as stomata, or through leaf wounds (reviewed in Lodewyckx et al. 2002). Here, we found that BG and AG tissues shared 20 bacterial genera, suggesting that migration from BG to AG tissues may occur in H. portulacoides. On the other hand, isolates belonging to the genera Hoeflea (17), Labrenzia (22), and Microbacterium (67) were only isolated from BG tissues, and the two latter genera were identified as indicator OTUs for these tissues by our indicator species analysis. The genera Hoeflea and Labrenzia have been isolated from root endosphere of a salt marsh halophyte (Bibi et al. 2014) and, to the best of our knowledge, have not been detected in AG tissues of plants, as was the case in our study. On the other hand, Microbacterium spp. has been detected in AG tissues of H. portulacoides (e.g., Alves et al. 2014). An OTU attributed to the genus Zunongwangia was considered indicator OTU for AG tissues in our analysis. Although this genus has been isolated from the rhizosphere of mangrove trees (Rameshkumar et al. 2014) and detected by pyrosequencing in the endosphere (Mora-Ruiz et al. 2015), it had not been previously isolated in the endosphere.

We found that the bacterial composition across our collection is significantly affected by the isolation site and tissue, although these factors only account for 23 % of the total variation observed in our data. Nevertheless, a variation in composition could be observed across the gradient of contamination in our sampling sites, suggesting that the contamination levels do, in fact, affect the composition of the culturable endosphere of H. portulacoides. Other variation parameters (e.g., soil water content, organic matter content, fine particle content, redox potential, pH, and conductivity; Válega et al. 2008b) could be explored in order to further understand the reasons for composition variation in the culturable endosphere of this halophyte.

Many isolates from the collection obtained in this study tested positive for several enzymatic activities and PGP traits, exposing the endosphere of the halophyte H. portulacoides as a novel source of bacteria and bacterial compounds with potential biotechnological applications. The ability to produce these extracellular enzymes and PGP traits is well distributed across phylogenetic groups, as the most represented classes (Fig. 1) amply exhibited all tested traits, except Bacilli for phosphate solubilization (Supplementary Fig. S1). This ability is also well represented, overall, in all isolation sites and tissues (Fig. 3). The ability to produce industrially important extracellular enzymes was abundantly detected in isolates associated to the genera Altererythrobacter, Marinilactibacillus, Microbacterium, and Vibrio. To our knowledge, the production of the tested enzymes had not been previously described in Altererythrobacter spp. and Marinilactibacillus spp.

To the best of our knowledge, the PGP traits tested in the presented study were not previously evaluated in Altererythrobacter spp. isolates. Here, we found that isolates from this genus tested positive for one to six extracellular enzymatic activities and presented one to five PGP traits. Overall, all tested enzymes and traits were produced by isolates belonging to his genus.

Traits for Marinilactibacillus spp. isolates from other studies have not been deeply characterized. Here, we found that Marinilactibacillus spp. isolates tested positive for one to five extracellular enzymatic activities, and presented one to four PGP traits. Considering all the isolates from this genus, all of the tested enzymes and traits were produced, except pectinase at pH 5.

The ability to promote plant growth has been previously detected in the genus Microbacterium with traits as ACC deaminase (Sheng et al. 2008), IAA production, phosphate solubilization (Kukla et al. 2014), and siderophore production (Pereira et al. 2013). The presence of the nifH gene was also detected (Zakhia et al. 2006). Production of protease and lipase (Pereira et al. 2013), xylanase, amylase (Park et al. 2006), cellulase (Kukla et al. 2014), and pectinase (Vinod et al. 2014) has been characterized in Microbacterium spp. Antimicrobial activity and resistance to metals have also been reported in this genus (Sheng et al. 2008; Passari et al. 2015). In the present study, one to five of the abovementioned extracellular enzymes were also detected in all Microbacterium spp. isolates, as were one to four PGP traits. Considering all the isolates from this genus, all of the tested enzymes and traits were present.

Previously studied Vibrio spp. exhibited enzymatic activities such as proteolytic, cellulolytic, amylolytic (Jose et al. 2014), xylanolytic (Kiyohara et al. 2005), and lipolytic (Ray et al. 2012), showed potential for plant growth promotion by solubilizing phosphate, fixing atmospheric nitrogen, producing IAA (Jose et al. 2014), and siderophores (Thompson et al. 2004). Vibrio spp. isolates from our collection tested positive for these activities and traits and, in addition, pectinolytic activity (at both pH = 5.0 and 7.0) and ACC deaminase activity. Other studies have also detected antimicrobial activity (Kumar and Nair 2007) and considerable Hg resistance and ability for Hg bioremediation in Vibrio spp. isolates (Jafari et al. 2015).

IAA production was detected in 460 isolates, 17 of which produced over 100 μg mL−1. The most represented IAA producing genus is Pseudomonas with five isolates, which has previously been associated with IAA producing potential in rice root exudates (Karnwal 2009). Phosphate solubilization was detected in 64 isolates, 10 of which belong to the genus Microbacterium, and 15 belong to the genus Salinicola. While Microbacterium spp. have been associated with phosphate solubilizing activity (Kukla et al. 2014), this is, to our knowledge, the first report of such activity in the Salinicola genus. Using PCR amplification, 20.8 % of representative isolates tested positive for the presence of the nifH gene. Of the 97 nifH-positive isolates, 63 originated from BG tissues and 17 belonged to the genus Vibrio, which has previously been associated with this trait (Jose et al. 2014). According to Penrose and Glick (2003), the ability to use ACC as a sole nitrogen source is a consequence of the activity of the enzyme ACC deaminase. From the 239 isolates that were able to grow in DF minimal medium supplemented with ACC, 155 (64.9 %) originated from BG tissues, and 60 % of isolates sampled in site B tested positive for this activity. Isolates associated with the genera Microbacterium (33) and Altererythrobacter (25) contributed the most for this PGP trait. Isolates from Microbacterium genus have previously been associated with ACC deaminase activity (Sheng et al. 2008); however, to our knowledge, this is the first description of such activity in the genus Altererythrobacter. Siderophore production was observed in 152 isolates (32.5 % of the collection), 19 of which belong to the genus Vibrio, which had been previously linked to this activity (Thompson et al. 2004). Nineteen Marinilactibacillus spp. isolates also tested positive, which is, to our knowledge, the first time such activity was described in this genus.

This work revealed that the endosphere of H. portulacoides represents a bacterial hot spot harboring a diverse microbiota, including putative novel species and compounds with biotechnological relevance. Endophytic bacteria were isolated from AG and BG tissues of the halophyte from contaminated and noncontaminated sites. Seventy-nine different genera were found using traditional cultivation methods. The phylogenetically diverse collection presented structural differences among sampling tissues and sites, and many of the isolates revealed their ability to produce extracellular enzymes and PGP traits. The PGP potential of these endophytes should be further characterized, aiming toward enhancing phytoextraction of metal-contaminated soils by H. portulacoides. The possibility of monitoring Salinicola spp. as a bioindicator of lack of contamination should also be further studied. Metal resistance and antimicrobial properties of these isolates should also be assessed to further characterize functional capabilities of the collection.