Desert Plant Seed Endophytes: A Reservoir of Stress-Adapted Bacillus Strains for Enhancing Wheat Salinity Tolerance

Abstract

Land desertification poses a major ecological challenge and threatens agricultural productivity. This study investigated the seed endophytic microbiomes of desert plants as a potential resource for mitigating salt stress in crops. Using high-throughput sequencing, we characterized the bacterial and fungal communities within seeds of 12 desert plant species. Dominant taxa included Firmicutes (particularly Bacillus), Bacteroidota, Proteobacteria, Ascomycota, and Basidiomycota. Culturable bacteria were subsequently isolated from Haloxylon ammodendron (C.A.Mey.) Bunge (HB) and Hedysarum scoparium Fisch. & C.A.Mey. (HSA) seeds. These isolates were screened for plant growth-promoting (PGP) traits and tolerance to salt (NaCl) and alkali (NaHCO₃). Selected strains, including the high indole-3-acetic acid (IAA)-producing Bacillus sp. HB-4, were used to inoculate wheat (Triticum aestivum L.) under 150 mM NaCl or 150 mM NaHCO₃ stress. Inoculation with strain HB-4 significantly improved wheat growth under stress. This improvement was associated with increased chlorophyll and proline content, enhanced activities of the antioxidant enzymes catalase and peroxidase, and reduced levels of malondialdehyde, a marker of oxidative damage. Our results demonstrate that desert plant seeds harbor taxonomically distinct and functionally resilient endophytes. The successful application of a desert-adapted Bacillus strain to alleviate salt stress in wheat highlights the potential of such microbiomes as a novel source of inoculants for sustainable agriculture in saline-affected regions.

1. Introduction

Desert ecosystems are characterized by severe abiotic stresses, including high salinity, drought, and nutrient-poor soils, which impose significant constraints on plant survival [1,2]. Plants have evolved a suite of adaptive strategies, among which symbiotic associations with microorganisms play a crucial role [3,4]. Endophytes, microorganisms that reside within plant tissues without causing disease, can enhance host fitness by facilitating nutrient acquisition, producing phytohormones, and inducing systemic resistance to both biotic and abiotic stresses [5,6].

Particularly relevant to stress adaptation is the role of seed-borne endophytes. Seeds can vertically transmit a select microbial community to the next generation, effectively providing seedlings with a pre-adapted “microbial starter kit” [7,8]. This vertical transmission suggests that seed endophytes are not random passengers but are likely enriched for traits that confer a fitness advantage during critical early growth stages in challenging environments [9,10]. Recent culture-independent studies have begun to characterize the taxonomic composition of seed endophytes across various plant species, frequently reporting the dominance of bacterial phyla such as Firmicutes and Proteobacteria and fungal phyla such as Ascomycota [11,12]. In particular, members of the genus Bacillus within the Firmicutes are often prevalent and are widely recognized for their PGP capabilities and stress tolerance [13,14].

However, a significant gap persists between describing these communities and mechanistically understanding their ecological function [15]. While the prevalence of Bacillus in seeds is often noted, the specific traits underlying its ecological success and its direct, causal role in mediating host plant stress tolerance remain insufficiently explored [16]. For instance, is its dominance linked to superior production of the phytohormone indole-3-acetic acid (IAA), exceptional antioxidant capacity, or the synthesis of specific osmoprotectants [17,18]? Furthermore, when beneficial effects of seed endophytes are observed in planta—such as increased proline accumulation, enhanced antioxidant enzyme activity, or reduced oxidative damage—these physiological responses are seldom explicitly linked back to the quantified functional traits of the applied microbial strain [19,20]. Moreover, the potential of harnessing endophytes from extremophyte hosts (e.g., desert plants) to confer cross-tolerance to agricultural crops remains largely untapped. Establishing clear connections between microbial taxonomy, in vitro functional attributes, and in planta physiological outcomes is essential for advancing from correlation to mechanism and for the rational development of effective microbial inoculants [21,22].

Therefore, this study was designed to bridge these gaps by employing an integrated approach. We first profiled the seed endophytic communities of 12 desert plant species using high-throughput sequencing to identify dominant and core taxa. We then focused on isolating culturable endophytes from two representative species, screening them comprehensively for multiple PGP traits (IAA production, siderophore synthesis, phosphate solubilization, nitrogen fixation) and resilience to salt and alkaline stress. Finally, selected bacterial strains were applied to wheat plants subjected to salt and alkali stress. Our objectives were to: (1) characterize the taxonomic diversity and predicted functional potential of desert plant seed endophytes; (2) isolate and functionally characterize culturable bacterial endophytes with PGP potential; and (3) evaluate the efficacy of selected strains in enhancing wheat salinity tolerance and interpret the plant’s physiological responses (growth, chlorophyll content, oxidative stress markers, osmolyte accumulation) in the context of the inoculant’s identified functional traits. By forging these links, we aim to provide a more mechanistic framework for understanding stress adaptation conferred by seed endophytes and to assess their practical potential as bio-inoculants for crops like wheat in marginal environments [23].

2. Materials and Methods

2.1. Sample Collection and Seed Surface Sterilization

Mature seeds from twelve dominant desert plant species were collected from Minqin County, Gansu Province, China (38°3′45″–39°27′37″ N, 101°49′41″–104°12′10″ E). The species were Alhagi camelorum Fisch. (AC), Salsola passerina Bunge (CA), Capparis spinosa L. (CL), Calligonum mongolicum Turcz. (CM), Caragana sinica (Buc’hoz) Rehder (CS), Haloxylon ammodendron (C.A.Mey.) Bunge (HB), Hedysarum scoparium Fisch. & C.A.Mey. (HSA), Nitraria tangutorum Bobrov (NT), Phragmites australis (Cav.) Trin. ex Steud. (PT), Sophora alopecuroides L. (SA), Tamarix chinensis Lour. (TA), and Zygophyllum xanthoxylum (Bunge) Engl. (ZX). Voucher specimens were authenticated by Professor Li Shanjia. Seeds were stored at 4 °C.

For surface sterilization, seeds were immersed in 70% (v/v) ethanol for 2 min with agitation (200× g), followed by treatment with 4% (v/v) sodium hypochlorite (NaClO) for 5 min, and then rinsed 7–8 times with sterile distilled water [24]. To verify sterilization efficacy, 100 µL of the final rinse water was plated onto Lysogeny Broth (LB) agar and Potato Dextrose Agar (PDA). The absence of microbial growth on both media after 72 h of incubation at 28 °C confirmed successful disinfection. PDA was included specifically to detect any viable fungal spores that might not grow on LB.

2.2. DNA Extraction, Amplification, and High-Throughput Sequencing

Surface-sterilized seeds were ground in liquid nitrogen. Total genomic DNA was extracted using the TGuide S96 Magnetic Soil/Stool DNA Kit (Tiangen Biotech Co., Ltd., Beijing, China) [25]. The V3–V4 hypervariable region of the bacterial 16S rRNA was amplified with primers 335F/769R [26]. The fungal internal transcribed spacer (ITS) region was amplified with primers ITS1F/ITS2 [27]. Library preparation and paired-end sequencing (2 × 250 bp) were performed on an Illumina NovaSeq 6000 platform at Beijing Bemac Biotechnology Co., Ltd, Beijing, China.

Raw sequencing data were initially processed on the BMKCloud platform, where quality filtering and DADA2-based denoising were performed to obtain high-quality non-chimeric reads and generate the Amplicon Sequence Variant (ASV) table. Subsequent bioinformatics analyses were carried out in QIIME 2 (version 2023.5). Taxonomic classification of bacterial ASVs was conducted against the SILVA database (v138.1) with a confidence threshold of ≥70%, while fungal ASVs were classified using the UNITE database (v9.0) under the same threshold. Alpha-diversity indices, including the Shannon index, Chao1 index, and Pielou’s evenness, were calculated within the BMKCloud platform after rarefaction to an even sequencing depth. To identify the core bacteriome, ASVs present in at least 80% of samples within each plant species group were retained and analyzed at the genus and family levels; Bacillus was confirmed among the core taxa.

2.3. Isolation and Identification of Culturable Endophytic Bacteria

Based on their ecological dominance and preliminary sequencing data indicating high bacterial abundance, HB and HSA were selected for intensive bacterial isolation. To maximize recovery of diverse bacteria, seven different media were initially used: LB, Reasoner’s 2A (R2A), Nutrient Agar (NA), Tryptic Soy Agar (TSA), Potato Dextrose Agar (PDA), Oatmeal Agar (OMA), and Gauze’s No.1 Agar. Surface-sterilized seeds were either placed directly on agar plates or homogenized in sterile water, with the homogenate serially diluted and plated. All plates were incubated at 28 °C for 2–7 days. Following initial isolation, subsequent purification and maintenance of all bacterial isolates were performed exclusively on LB agar. This decision was based on the observation that LB supported consistent and robust growth for all obtained bacterial morphotypes, simplifying standardized handling across subsequent high-throughput functional screens. While fungi were characterized via sequencing, this study focused on subsequent culturing and functional assays on bacterial isolates due to their predominant recovery and well-established role in abiotic stress alleviation in Gramineae crops like wheat.

Pure cultures were obtained by repeated streaking. Genomic DNA from pure cultures was extracted using the Bacterial Genomic DNA Extraction Kit (Tiangen Biotech, Beijing, China). The near-full-length 16S rRNA was amplified using universal primers 27F and 1492R. Purified PCR products were sequenced via Sanger sequencing (Genewiz, Suzhou, China). The resulting sequences were compared to the NCBI GenBank database using the BLASTn algorithm (https://blast.ncbi.nlm.nih.gov/Blast.cgi, accessed on 1 February 2026). A sequence similarity threshold of ≥97% was used for genus-level identification, with the highest similarity match to a type strain recorded (

Table 1. Identification of endophytic bacteria isolated from the seeds of HB and HSA based on 16S rRNA sequence analysis.

Strain	Similar Strain Name	Similarity (%)	Reference Numbers (NCBI)	GenBank Accession Numbers (This Study)
HB-1	Bacillus altitudinis	99.88%	NR_042337.1	PQ774797
HB-2	Bacillus pumilus	98.94%	NR_115334.1	PQ774798
HB-3	Bacillus aerius	99.88%	NR_118439.1	PQ836096
HB-4	Priestia aryabhattai	100.00%	NR_115953.1	PQ774800
HB-5	Kocuria rosea	99.89%	NR_044871.1	PQ836097
HB-6	Bacillus zhangzhouensis	99.89%	NR_148786.1	PQ836098
HB-7	Peribacillus frigoritolerans	98.25%	NR_115064.1	PQ774803
HB-8	Priestia megaterium	97.54%	NR_116873.1	PQ836099
HB-9	Priestia megaterium	98.70%	NR_112636.1	PQ774805
HB-10	Bacillus safensis	97.21%	NR_041794.1	PQ836100
HB-11	Bacillus pumilus	98.58%	NR_043242.1	PQ836101
HB-12	Bacillus licheniformis	99.78%	NR_074923.1	PQ774804
HB-13	Terribacillus aidingensis	97.90%	NR_114288.1	PQ836102
HSA-1	Bacillus zanthoxyli	98.73%	NR_043401.1	PQ835926
HSA-2	Paenibacillus rhizoplanae	98.40%	NR_156842.1	PQ835927
HSA-3	Bacillus velezensis	98.79%	NR_075005.2	PQ835928
HSA-4	Bacillus mojavensis	98.06%	NR_118290.1	PQ835929
HSA-5	Bacillus subtilis	98.29%	NR_027552.1	PQ835930
HSA-6	Bacillus zhangzhouensis	99.08%	NR_148786.1	PQ835931
HSA-7	Bacillus licheniformis	98.39%	NR_074923.1	PQ835932
HSA-8	Bacillus rugosus	97.83%	NR_181236.1	PQ835933
HSA-9	Bacillus tequilensis	97.81%	NR_104919.1	PQ835934

Note: Strain prefixes indicate the host plant species of origin: HB, Haloxylon ammodendron (C.A.Mey.) Bunge; HSA, Hedysarum scoparium Fisch. & C.A.Mey. Species identification was based on the highest similarity of the 16S rRNA gene sequences to type strains in the NCBI database. The GenBank accession numbers obtained in this study are listed.

).

2.4. In Vitro Screening for Plant Growth-Promoting Traits

All phenotypic assays were performed in triplicate.

IAA Production: Bacterial strains were grown in LB supplemented with 1 g·L⁻¹ L-tryptophan for 5 days at 28 °C with shaking (180× g). The culture supernatant was mixed with Salkowski’s reagent, and the absorbance at 530 nm was measured after 30 min in the dark [28]. IAA concentration was determined using a standard curve.

Nitrogen Fixation: Bacterial growth was assessed on Ashby’s nitrogen-free medium after 7 days of incubation at 28 °C.

Phosphate Solubilization: The formation of a clear halo around colonies was evaluated on National Botanical Research Institute’s Phosphate (NBRIP) agar plates after 2–7 days [29].

Siderophore Production: The Chrome Azurol S (CAS) agar plate assay was used. The development of an orange-yellow halo indicated siderophore production [30].

2.5. Screening for Salt and Alkali Tolerance

Salt tolerance was assessed on LB agar plates amended with NaCl at concentrations of 2%, 5%, 7%, 10%, 12%, or 15% (w/v; pH 7.0). Alkali tolerance was tested on LB agar containing 1% NaCl, with the pH adjusted to 8.0, 9.0, 10.0, 11.0, or 12.0 using NaOH. Strains were streaked onto plates and incubated at 28 °C for up to 7 days. Growth was visually assessed daily. Strains demonstrating growth at concentrations of ≥10% NaCl and/or at pH ≥ 10.0 were classified as salt-tolerant or alkali-tolerant, respectively. The most tolerant strains from this screen (i.e., those growing at the highest stress levels) were selected for subsequent pot experiments.

2.6. Plant Growth Promotion and Stress Alleviation Assay in Wheat

A pot experiment was conducted using wheat (Triticum aestivum L. cv. ‘Jiemai 22’). Seeds were surface-sterilized as described in Sample Collection and Seed Surface Sterilization and germinated on sterile vermiculite. Uniform 4-day-old seedlings were used.

The experiment followed a factorial design with two factors: Stress Treatment (Control: water; Salt stress: 150 mM NaCl; Alkali stress: 150 mM NaHCO₃) and Inoculation Treatment (Non-inoculated control, CK; and inoculation with one of six selected bacterial strains, including HB-4 and HB-9). Each of the 18 treatment combinations was replicated four times (n = 3), with each replicate consisting of one pot containing ten seedlings. Therefore, the total number of plants was 720 (18 treatments × 4 replicates × 10 plants).

For inoculation, seedlings were dipped in a bacterial suspension (OD₆₀₀ = 1.0) for 4 h prior to transplanting into pots filled with an autoclaved mixture of nutrient soil and vermiculite (1:1, w/w). The respective stress solution (water, NaCl, or NaHCO₃) was used to saturate the soil at transplanting. Plants were grown in a controlled growth chamber (25 °C, 16/8 h light/dark cycle). A bacterial suspension (or sterile water for CK) was applied as a soil drench every 7 days. Plants were harvested 30 days after transplanting.

2.7. Measurement of Growth and Physiological Parameters

Plant height, root length, and fresh weight were recorded. For physiological analyses, the youngest fully expanded leaves were sampled.

Chlorophyll content: The total chlorophyll content was determined according to the method of Wintermans and De Mots with modifications [31]. Briefly, fresh wheat leaves (0.1 g) were thoroughly washed, blotted dry, and ground into a homogeneous paste in a mortar with 0.1 g of SiO₂, 0.05 g of CaCO₃, and 2 mL of 95% (v/v) ethanol until the tissue turned white. After standing for 3 min, the homogenate was transferred to a 10 mL centrifuge tube and centrifuged at 4000× g for 10 min. The supernatant was collected, and the residue was re-extracted with 95% ethanol. The combined supernatants were diluted to a final volume of 25 mL with 95% ethanol in a brown volumetric flask. The absorbance of the extract was measured at 665 nm (A₆₆₅) and 649 nm (A₆₄₉) using a spectrophotometer. Three technical replicates were performed for each biological sample. The concentrations of chlorophyll a (Cₐ, mg L⁻¹), chlorophyll b (C_b, mg L⁻¹), and total chlorophyll (Cₜ, mg L⁻¹) were calculated using the following formulas:

Cₐ = 13.95 × A₆₆₅ − 6.88 × A₆₄₉

(1)

C_b = 24.96 × A₆₄₉ − 7.32 × A₆₆₅

(2)

C = Cₐ + C_b

(3)

The total chlorophyll content in the fresh leaf tissue (mg g⁻¹ FW) was then calculated as:

Cₜ = (C × V × D)/(1000 × m)

(4)

where V is the total volume of the extract (mL), D is the dilution factor, and m is the fresh weight of the sample (g).

Malondialdehyde (MDA) Content: The MDA content, an indicator of lipid peroxidation, was determined using the thiobarbituric acid (TBA) reaction as described by Buttar et al. with modifications [32]. Fresh wheat leaves (0.1 g) were homogenized in 1 mL of 10% (w/v) trichloroacetic acid (TCA) using a homogenizer with grinding beads (4 °C, 50 Hz, 5 min). The homogenate was then transferred to a 10 mL tube, and the volume was brought to 10 mL with 10% TCA. After centrifugation at 4000× g for 10 min, 2 mL of the supernatant was aliquoted and thoroughly mixed with 2 mL of 0.6% (w/v) TBA (prepared in 10% TCA). A control reaction contained 2 mL of distilled water mixed with 2 mL of TBA solution. The mixtures were incubated in a boiling water bath for 15 min, then rapidly cooled in an ice bath and centrifuged at 4000× g for 10 min. The absorbance of the supernatant was measured at 532 nm (A₅₃₂), 600 nm (A₆₀₀), and 450 nm (A₄₅₀). Three technical replicates were measured per biological sample. The MDA concentration (C_MDA, μmol L⁻¹) was calculated using the formula:

$${\mathrm{C}}_{\mathrm{MDA}}=6.45\times {\mathrm{A}}_{532}-0.56\times {\mathrm{A}}_{560}$$

(5)

The MDA content in the leaf tissue (nmol g⁻¹ FW) was then determined as:

MDA Content = 1000(C_MDA × V)/m

(6)

where V is the total volume of the extraction mixture (mL), and m is the fresh weight of the sample (g).

The activities of catalase (CAT) and peroxidase (POD) were determined using commercial assay kits (CAT Activity Assay Kit and POD Activity Assay Kit, Beijing Solarbio Science & Technology Co., Ltd., Beijing, China) according to the manufacturer’s instructions. For both assays, frozen leaf tissue (0.1 g) was homogenized in 1 mL of ice-cold extraction buffer specific to each kit. The homogenates were centrifuged at 12,000× g for 10–20 min at 4 °C, and the supernatants were used as crude enzyme extracts.

Antioxidant Enzyme Activities: CAT activity was assayed by monitoring the decomposition of H₂O₂ at 240 nm over 3 min [33].

POD activity was assayed by measuring the oxidation of guaiacol at 470 nm over 3 min [34]. One unit (U) of POD activity was defined as the amount of enzyme that caused an increase in absorbance of 0.01 per minute per mg of protein. The protein concentration in the crude extracts was determined using the Bradford method. Enzyme activities were expressed as U g⁻¹ protein min⁻¹. Three technical replicates were performed for each biological sample.

Proline Content: Proline was extracted from 0.1 g FW leaf tissue with 3% (w/v) sulfosalicylic acid and quantified using the acid–ninhydrin method [35]. Fresh wheat leaves (0.1 g) were homogenized in 1 mL of 3% (w/v) sulfosalicylic acid solution. The homogenate was incubated in a 90 °C water bath for 10 min and then centrifuged at 4000× g for 10 min. A 500 μL aliquot of the supernatant was mixed with 500 μL of glacial acetic acid and 500 μL of acid–ninhydrin reagent in a 5 mL tube. The mixture was incubated in a boiling water bath for 30 min. After cooling to room temperature, 1 mL of toluene was added, and the mixture was vortexed vigorously. After phase separation, the absorbance of the upper toluene layer was measured at 520 nm. A standard curve was prepared using L-proline solutions (0–50 μg mL⁻¹). The proline content in the leaf tissue (μg g⁻¹ FW) was calculated based on the standard curve (y = 0.0296x + 0.0059, R² = 0.999). Three technical replicates were performed for each biological sample.

2.8. Data Analysis

All data from the pot experiment are presented as the mean ± standard error (SE) of four independent biological replicates (n = 3), with each data point derived from three technical measurements where applicable. Prior to parametric statistical analysis, all datasets were tested for normality of distribution using the Shapiro–Wilk test and for homogeneity of variances using Levene’s test. For data meeting these assumptions, two-way analysis of variance (ANOVA) was performed to evaluate the main effects of stress treatment (Control, Salt, Alkali), bacterial inoculation, and their interaction. Where a significant interaction (p < 0.05) was found, simple main effects were analyzed. Pairwise comparisons among treatment means were conducted using Tukey’s Honestly Significant Difference (HSD) post hoc test at a significance level of α = 0.05 (two-tailed). All statistical analyses were performed using SPSS software (version 26.0, IBM Corp., Armonk, NY, USA). Data visualization was performed using Origin 2022 (OriginLab Corp., Northampton, MA, USA). The sequencing data analysis for the initial characterization of seed endophytes is described in Section 2.2.

3. Results

3.1. Diversity of Seed Endophytic Bacteria in Desert Plants

High-throughput sequencing of the 16S rRNA was carried out on seed samples from twelve desert plant species. Each species included three independent biological replicates. Analysis revealed significant interspecific differences in bacterial community composition (p < 0.05). Seeds of CM showed the highest bacterial richness at all taxonomic levels, while ZX and SA seeds contained the least diverse communities (

Table 2. Statistical table on the classification of endophytic bacteria in the seeds of 12 desert plant species.

Sample	Kindom	Phylum	Class	Order	Family	Genus	Species
AC	2	33	74	187	378	710	847
CA	2	28	70	166	316	611	722
CL	1	29	73	188	361	717	864
CM	2	36	87	220	470	994	1223
CS	2	30	64	155	279	519	589
HB	2	31	75	165	314	593	668
HSA	2	32	77	204	378	737	874
NT	2	33	73	154	289	535	624
PT	2	32	75	177	339	671	794
SA	2	31	65	152	268	481	541
TA	2	32	75	202	393	767	922
ZX	2	31	67	159	281	501	571
Total	2	43	110	316	717	1767	2525

).

A core bacterial microbiome, defined as amplicon sequence variants (ASVs) present in ≥80% of all samples within a species group and with a mean relative abundance >1%, was reconstructed at the genus level. At the phylum level, four phyla were consistently dominant: Firmicutes (32.83–55.13%), Bacteroidota (21.18–34.12%), Proteobacteria (2.81–24.01%), and Campylobacterota (1.12–5.73%) (Figure 1A,B). Within the broad phylum Firmicutes, members of the genus Bacillus were a consistent and identifiable component of the core microbiome across all plant species. Alpha-diversity indices (Chao1, Shannon) confirmed that CM seeds had the highest bacterial richness and diversity, while ZX had the lowest (

Table 3. Comparison of alpha diversity of endophytic bacteria from 12 species of seeds.

Samples	ACE Index	Chao1 Index	Simpson Index	Shannon Index	Coverage
AC	2074.37 ± 63.39 b	2052.52 ± 61.33 b	0.99 ± 0 b	8.44 ± 0.03 b	0.99
CA	1745.63 ± 30.5 cd	1732.84 ± 29.37 cd	0.99 ± 0 b	8.31 ± 0.03 c	0.99
CL	2144.43 ± 254.56 b	2134.26 ± 254.73 b	0.99 ± 0 c	8.48 ± 0.12 b	0.99
CM	2943.73 ± 212.41 a	2913.42 ± 215.84 a	1.00 ± 0 a	9.47 ± 0.12 a	0.99
CS	1444.37 ± 13.99 fg	1438.11 ± 14.51 fg	0.99 ± 0 d	7.66 ± 0.02 g	0.99
HB	1645.34 ± 29.38 de	1636.92 ± 28.61 de	0.98 ± 0 de	7.78 ± 0.02 f	0.99
HSA	2091.44 ± 26.71 b	2079.52 ± 25.54 b	0.99 ± 0 c	8.13 ± 0.03 d	0.99
NT	1508.41 ± 32.22 ef	1497.4 ± 30.81 ef	0.99 ± 0 b	8.19 ± 0.03 d	0.99
PT	1819.69 ± 41.2 c	1807.93 ± 38.94 c	0.99 ± 0 c	8.01 ± 0.03 e	0.99
SA	1340.5 ± 22.57 g	1335.57 ± 21.94 g	0.98 ± 0 ef	7.57 ± 0.02 h	0.99
TA	2211.72 ± 51.84 b	2196.22 ± 50.69 b	0.99 ± 0 c	8.16 ± 0.02 d	0.99
ZX	1393.86 ± 16.56 fg	1387.95 ± 15.49 fg	0.98 ± 0 f	7.6 ± 0.03 gh	0.99

Note: Different letters in the same column indicate significant differences between seeds (p < 0.05).

). Phenotype prediction using BugBase indicated a high relative abundance of traits such as anaerobic metabolism, Gram-positive cell structure, and biofilm formation (Figure 1E).

3.2. Diversity of Seed Endophytic Fungi in Desert Plants

Sequencing of the fungal ITS region identified communities from 17 phyla across the 12 seed sample sets. Seeds of CL contained a significantly higher (p < 0.05) number of fungal taxa, followed by CM. CA seeds had the lowest fungal richness (

Table 4. Taxonomic statistics of endophytic fungi of 12 desert plant species.

Sample	Kindom	Phylum	Class	Order	Family	Genus	Species
AC	1	13	46	102	221	433	631
CA	1	11	33	66	133	236	325
CL	1	17	58	139	317	763	1268
CM	1	16	53	113	262	549	851
CS	1	13	41	93	176	325	435
HB	1	13	44	95	200	386	583
HSA	1	11	42	95	197	380	588
NT	1	12	38	78	165	288	397
PT	1	14	46	96	199	383	547
SA	1	11	37	83	158	275	358
TA	1	13	36	81	166	288	377
ZX	1	11	37	85	171	285	375
Total	1	17	61	153	370	992	1901

).

At the phylum level, the dominant lineages were Ascomycota (36.76–75.60%), Basidiomycota (10.67–45.64%), unclassified Fungi (4.19–30.56%), and Mortierellomycota (1.66–6.09%) (Figure 2A,B). At the genus level, 101 genera were shared across all samples. CL seeds contained 215 unique genera, far exceeding other species (Figure 2C). Dominant shared genera (relative abundance >1%) included unclassified Fungi, unidentified, Aspergillus, Mortierella, Fusarium, Cladosporium, and unclassified Basidiomycota (Figure 2D).

Alpha-diversity analysis confirmed CL seeds had the highest fungal richness and diversity, while CA had the lowest richness (

Table 5. Comparison of the alpha diversity of endophytic fungi from 12 species of seeds.

Samples	ACE Index	Chao1 Index	Simpson Index	Shannon Index	Coverage
AC	760.96 ± 525.00 bc	760.78 ± 525.08 bc	0.94 ± 0.05 ab	7.09 ± 1.07 ab	0.99
CA	252.39 ± 107.43 c	252.03 ± 107.05 c	0.98 ± 0.01 a	7.11 ± 0.6 ab	0.99
CL	1816.15 ± 951.41 a	1808.06 ± 945.75 a	0.99 ± 0.00 a	8.46 ± 0.22 a	0.99
CM	1069.53 ± 256.23 b	1066.11 ± 254.26 b	0.96 ± 0.01 a	6.76 ± 0.63 bc	0.99
CS	461.860 ± 325.96 bc	461.36 ± 325.85 bc	0.91 ± 0.09 ab	6.24 ± 1.60 bc	0.99
HB	591.96 ± 145.59 bc	590.49 ± 145.24 bc	0.98 ± 0.00 a	7.37 ± 0.22 ab	0.99
HSA	538.53 ± 39.31 bc	538.43 ± 39.50 bc	0.79 ± 0.01 c	5.25 ± 0.07 c	0.99
NT	354.690 ± 295.47 c	354.14 ± 295.42 c	0.95 ± 0.02 ab	6.43 ± 0.93 bc	0.99
PT	467.97 ± 513.01 b	467.63 ± 512.63 b	0.98 ± 0.01 a	7.38 ± 0.37 ab	0.99
SA	420.10 ± 301.78 c	420.04 ± 301.70 c	0.98 ± 0.00 a	7.29 ± 0.87 ab	0.99
TA	313.49 ± 70.06 c	313.38 ± 69.88 c	0.88 ± 0.13 b	6.02 ± 2.14 bc	0.99
ZX	394.39 ± 216.77 c	394.25 ± 216.79 c	0.99 ± 0.00 a	7.61 ± 0.56 ab	0.99

Note: Different letters in the same column indicate significant differences between seeds (p < 0.05).

). Functional guild prediction (FUNGuild) suggested dominant roles as undefined saprotrophs, plant pathogens, and animal pathogens (Figure 2E).

3.3. Isolation and Plant Growth-Promoting (PGP) Traits of Culturable Endophytes

Based on ecological dominance and sequencing data indicating high bacterial abundance, HB and HSA were selected for bacterial isolation. A total of 22 strains were obtained (Table 1); 16S rRNA sequencing identified isolates belonging to the genera Bacillus (15 strains), Priestia (3 strains), Peribacillus, Kocuria, Terribacillus, and Paenibacillus (Figure 3A).

In vitro screening revealed multiple strains exhibited PGP traits, including IAA production, nitrogen fixation, phosphate solubilization, and siderophore production (Figure 3B–D). Strain HB-4 showed the highest IAA production. Based on robust growth and PGP efficacy, strains HB-4 and HB-9 were selected for pot experiments with wheat.

3.4. Promotion of Wheat Growth by Inoculation with HB-4 and HB-9

Wheat seedlings inoculated with HB-4 or HB-9 showed enhanced growth compared to the non-inoculated control (CK) after 30 days (Figure 4A). Inoculation with HB-4 significantly increased seedling fresh weight by 120.33% and dry weight by 69.19% compared to CK (Figure 4B). HB-9 inoculation significantly increased fresh weight but not dry weight.

Total chlorophyll content in leaves was increased by both inoculants compared to CK (Figure 4C). Leaf malondialdehyde (MDA) content, an indicator of oxidative stress, was reduced by both strains (Figure 4D). Activities of the antioxidant enzymes catalase (CAT) and peroxidase (POD) were elevated in leaves from both treatment groups (Figure 4E,F). Proline content increased in the HB-4 treatment but decreased in the HB-9 treatment compared to CK (Figure 4G).

3.5. Screening and Validation of Salt- and Alkali-Tolerant Strains

In vitro screening identified four bacterial strains (HB-2, HB-4, HB-5, HSA-2) demonstrating tolerance to high NaCl concentration (15% w/v) (Figure 5A). The same four strains also showed tolerance to high alkalinity (pH 11) (Figure 5B).

In pot experiments, wheat growth was progressively inhibited with increasing concentrations of either NaCl (Figure 5C) or NaHCO₃ (Figure 5D). Inoculation with any of the four tolerant strains alleviated this growth inhibition under both salt and alkali stress (Figure 5E). Under stress conditions, inoculation significantly increased seedling fresh weight, dry weight, and root length compared to the stressed, non-inoculated control (Figure 5F,G).

3.6. Physiological Responses to Inoculation Under Stress

Under salt or alkali stress, chlorophyll content (Figure 6A), CAT activity (Figure 6B), and POD activity (Figure 6C) were generally higher in inoculated plants compared to the stressed control. The effect on proline content varied by strain (Figure 6D). The response of leaf MDA content to inoculation was also strain-dependent (Figure 6E).

4. Discussion

This study integrates microbial community profiling, functional microbiology, and plant physiology to evaluate the potential of desert seed endophytes as bio-inoculants for salinity stress mitigation. While our sequencing data confirmed the previously reported dominance of bacterial phyla like Firmicutes (notably Bacillus) and Ascomycota in seeds [16,36], we extended these observations by directly linking this taxonomic prevalence to cultivable, functionally characterized isolates and their physiological impact on a non-host crop [37].

The isolation of 15 Bacillus strains from the seeds, corresponding to the dominant phylum Firmicutes, provides a cultivable and functionally relevant link to the sequencing data. Bacillus species are well-documented for their resilience and multifaceted PGP mechanisms, including the production of IAA, siderophores, and antioxidant enzymes [14,17]. Our in vitro assays confirmed these traits in our isolates. The pot experiment results offer strong correlative evidence that these traits contribute to stress alleviation in planta. Specifically, the high IAA-producing strain HB-4 was particularly effective in promoting growth under both control and stress conditions. IAA is known to stimulate root development, increasing the surface area for water and nutrient uptake, which is critical under stress [38]. The significant reduction in MDA content, coupled with elevated activities of CAT and POD in plants inoculated with HB-4 and HB-9, strongly suggests that these strains enhance the host’s antioxidant defense system, thereby mitigating salt-induced oxidative damage—a key mechanism of microbial-mediated stress tolerance [12,39].

It is noteworthy that not all four tolerant strains improved every physiological parameter to the same extent. This strain-specificity underscores the importance of multi-trait screening for selecting effective inoculants, as the primary mode of action may vary (e.g., dominant phytohormone production vs. strong antioxidant support) [40]. Our finding that the most effective in planta strains also exhibited strong in vitro IAA production and stress tolerance supports a plausible mechanistic link, although definitive causal proof would require mutant studies [41].

The successful application of these desert-derived endophytes to enhance salt tolerance in wheat, a non-host cereal crop, demonstrates their cross-kingdom beneficial potential. This suggests that the stress-adaptive traits of these microbes are not strictly host-specific but represent general biochemical mechanisms (e.g., antioxidant enhancement, osmotic adjustment) that can be transferred to agronomically important species [42]. This is significant for developing inoculants for use in arid and saline agroecosystems [22].

We acknowledge that our study demonstrates association rather than definitive causation. Future work involving genomics, transcriptomics, or the use of microbial mutants deficient in specific traits (e.g., IAA synthesis) would provide deeper mechanistic insights into the plant–microbe dialogue under stress [43,44].

In conclusion, we demonstrate that desert plant seeds are a rich reservoir of stress-adapted endophytes, with Bacillus being a prevalent and functionally versatile genus. By isolating strains that reflect dominant community members and linking their in vitro PGP and stress-tolerance traits to the physiological alleviation of salt stress in wheat, we provide a rational framework for screening and selecting microbial inoculants [45]. Strains like Bacillus sp. HB-4, which consistently enhanced host growth, antioxidant capacity, and oxidative stress mitigation, show direct promise for development as biofertilizers to improve wheat productivity in saline-affected soils [46].

5. Conclusions

This study characterized the diverse endophytic microbiomes in seeds of 12 desert plant species, revealing a core community dominated by Firmicutes and Ascomycota. We isolated 22 bacterial endophytes from HB and HSA seeds, predominantly belonging to Bacillus spp., which exhibited multiple PGP traits in vitro. Salt–alkali tolerance screening identified robust strains that significantly alleviated salt and alkali stress in wheat in subsequent pot experiments. Physiological analysis indicated that the most effective strains, such as Bacillus sp. HB-4, likely function by enhancing the host’s antioxidant enzyme activities (CAT, POD) and osmolyte (proline) accumulation, thereby reducing oxidative damage (MDA) and improving chlorophyll retention. These findings functionally link taxonomic dominance to stress-adaptive potential and provide concrete candidates along with a mechanistic rationale for developing microbial inoculants to enhance crop resilience in arid, saline environments.