The experiment was conducted at three desert stations, namely: (a) Cele Desert Research Station (CL): Located on the southern border of the Tarim Basin at coordinates 37°00′57″N, 80°43′45″E, with an altitude of 1,318 m. The average yearly temperature is 11.9°C, annual precipitation averages 35.1 mm, and potential evaporation amounts to 2595.3 mm. There are approximately 20 days per year when dust storms occur, and dust is present for a total of 240 days. (b) Turpan Desert Botanical Garden (TLF): Situated in the arid expanse of Central Asia, within the Eurasian hinterland, at coordinates 42°51′N, 89°11′E. This botanical garden is unique as it is located at an elevation ranging from −105 m to −76 m, making it the world’s lowest. The region receives an average annual rainfall of 16.4 mm, while the annual evaporation rate is 3,000 mm. The mean yearly temperature is recorded as 13.9°C, and the frost-free period spans 265.6 days. (c) Mosuowan Desert Research Station (MSW): Positioned at 45°01′N, 89°11′E, this station is characterized by a continental arid climate. It sits at an altitude of 346 m, with an average yearly temperature of 6.6°C. Precipitation averages around 117 mm annually, and evaporation measures approximately 1979.5 mm. A frost-free period extends from 135 to 205 days (Figures 1, 2).

For each of the three study fields, encompassing CL, TLF, and MSW, a total of four homogeneous quadrats were chosen. Each quadrat, measuring approximately 30 m × 30 m, featured thriving natural populations of the desert plant A. sparsifolia, serving as representative samples. In total, 12 research blocks were selected, distributed across the three study fields. Throughout the spring, summer, and autumn of 2022, comprehensive sampling was undertaken from the phylloplane and leaf endosphere at these well-established, long-term, in-situ observation stations (No samples were collected in winter because the leaves of the plants were all dead). Stringent hygiene practices were adopted during the sampling process, with researchers wearing masks and sterile gloves. Disease-free leaves were chosen from all cardinal directions (east, west, north, and south) of A. sparsifolia within each quadrat, with a total of 60 leaves. Sterile swabs were used to clean the leaves, and subsequently, the cleaned leaves were carefully placed into sterile centrifuge tubes. Extra leaves from each cardinal direction were stored separately to maintain sample integrity. To prevent potential cross-contamination, gloves were changed, and hands were sanitized using alcohol-soaked cotton balls between each sampling instance. Upon arrival at the laboratory, the swabs and leaves within the centrifuge tubes were temporarily stored in an icebox maintained at a temperature of 4°C. During this period, they underwent pre-treatment for a duration of 2 to 5 h. This pre-treatment involved immersing the leaves in 95% alcohol, oscillating them for three cycles of 15 s each. Subsequently, the samples underwent three washes with sterile water and were placed in sterile centrifuge tubes at −80°C for the extraction of leaf endosphere bacteria.

Each set of samples comprised four biological replicates, resulting in a total of 72 samples collected throughout 2022, accounting for two compartments (phylloplane and leaf endosphere), three basins (CL, TLF, and MSW), three seasons, and four replicates. Genomic DNA extraction procedures were performed, and the resultant samples were stored at −80°C for subsequent analysis.

To extract complete genomic DNA from 500 mg of phylloplane and leaf endosphere, the DNeasy Power Soil DNA Isolation Kit (Qiagen, Inc., Netherlands) was employed, adhering to the guidelines provided by the manufacturer. PCR (Polymerase Chain Reaction) amplification of bacterial 16S rRNA genes V3-V4 region and fungal ITS 1–5 (Internal Transcribed Spacer) region was performed using universal primers 341 (5′-CCTAYGGGRBGCASCAG-3′) and 806 R (5′-GGACTACNNGGGTATCT AAT-3′); ITS 5-1737F (5′-GGAAGTAAAAGTCGTAACAAGG-3′) and ITS 2-2043R (5′-GCTGC GTTCTTCATCGATGC-3′), respectively (Nossa et al., 2010). Amplicon quality was determined with gel electrophoresis, purified with Agencourt AM-Pure XP beads, and amplified PCR (only one round). Qubit dsDNA assay kit (Life Technologies, United States) was used to quantify the final amplicon. Equal amounts of purified amplicon were pooled and sequenced using the Illumina HiSeq X Ten, PE150 platform.

The Uparse software (v7.0.1001) was used to generate operational taxonomic units (OTUs) with a 97% similarity threshold (Edgar, 2013). The QIIME (v1.91) was used to select the representative read for each out (Caporaso et al., 2010). Bacterial OTUs were classified using the RDP classifier (with a confidence threshold of 70%) according to either the Silva database (version 132) or Greengenes (Wang et al., 2007). To ensure bacterial and fungal (leaf phyllosphere) sequence uniformity among all samples, the minimum number (36,551 and 38,560) of sequences was used as the depth to filter other samples to generate in a filtered ASV table {This result removed plant sequences (“mitochondria “, “chloroplasts”) in fungal results, however, bacteria do not remove plant sequences}. The raw sequencing data are available at the National Center for Biotechnology Information (NCBI) Short Read Archive, Bio Project ID PRJNA1024038.

The leaf organic carbon (LOC) was determined using the K₂Cr₂O₇-H₂SO₄ oxidation method, and total nitrogen (TN) concentration was measured using the Kjeldahl Nitrogen Analyzer (K1160, Jinan Hanon Instruments Co. Ltd., China) (Bao, 2000). After digestion in concentrated HNO₃, the alkali hydrolyzable method was used to measure the available nitrogen (AN), while the Inductively Coupled Plasma-Optical Emission Spectrometer (iCAP 6,300, Thermo Elemental, United States) was employed to determine the total phosphorus (TP) and total potassium (TK) (Hooper and Vitousek, 1998). Available phosphorus (AP) was extracted using HCl/NH₄F through colorimetric analysis on a continuous-flow autoanalyzer (Autoanalyzer 3, Bran and Luebbe, Germany) employing ascorbic acid molybdate (Lu, 1999). To determine the availability of potassium (AK), the NH₄OAc extraction technique was utilized (Warra et al., 2015). Using a pH and EC meter (PHSJ-6 L and DDSJ-319 L, respectively) made by INESA Scientific Instrument Co. Ltd. in China, the acidity level and electrical conductivity of the leaf were determined. The pH was measured with a soil/water ratio of 1:2.5 (w/v), while the EC was measured with a soil/water ratio of 1:5 (w/v).

The leaf area was determined by employing the ImageJ program (Softonic International, Spain) following the established protocol. Afterward, the samples were placed in an oven at a temperature of 75°C for 24 h to determine their leaf dry weight (LDW). The specific leaf area (SLA) was calculated by dividing the leaf area by the LDW. However, the specific leaf weight (SLW) was calculated by dividing the LDW by the leaf area (Zhang et al., 2020). The physiochemical properties and morphological characteristics of leaves are shown in Supplementary Table S1.

The experiment collected meteorological data including the temperature, air relative humidity, atmospheric pressure, precipitation, wind speed, wind direction, direct radiation, indirect radiation, and diffuse radiation from the National Meteorological Science Data Center (http://data.cma.cn/) at the CL, TLF, and MSW of the Chinese Academy of Sciences (Supplementary Table S2).

The phylloplane and leaf endosphere were combined into phyllosphere microorganisms for analysis. To obtain a quantitative multi-functionality of the phyllosphere (bacteria and fungi) of A. sparsifolia from the different seasons (spring, summer, and autumn) and regions (CL, TLF, and MSW), two distinct and complementary multi-functionality approaches were used: (i) the average multi-functionality approach provides an easy-to-interpret and straight forward measure of the ability of an ecosystem to simultaneously sustain multiple functions, and this approach is widely used by current multi-functionality studies, (ii) the assessment of multi-dimensional functioning through the extraction of fundamental axes of ecosystem variation from principal coordinate analysis, serving as a representation of overall ecosystem functioning (Manning et al., 2018; Delgado-Baquerizo et al., 2020). The following calculation methods were used:

(i) Selected eight α-diversity indexes that were closely related to species diversity (Observed, Chao1, ACE, Shannon, Simpson, InvSimpson, Fisher phylogenetic diversity, and Goods coverage) to represent multi-functionality. First of all, these α-diversity indexes are normalized, and then their Z-scores are calculated (Maestre et al., 2012). The calculation formula is:

Where Z is the Z-score of the α-diversity index, and x_i is the value of the α-diversity index parameter; λ_i is the average value of the α-diversity index. δ_i is the standard deviation of the α-diversity index. Averaging the Z-score of α-diversity is called average multi-functionality (Maestre et al., 2012). The calculation formula is:

(ii) The α-diversity indices (Observed, Chao1, ACE, Shannon, Simpson, InvSimpson, Fisher phylogenetic diversity, and Goods coverage) were normalized and principal component analysis was performed. The interpretation rate of the first two axes is greater than 90%. Therefore, the first two axes are selected to evaluate the multi-functionality.

The statistical analysis was performed using R4.1.0 (R Core Team, 2021). The phylloplane and leaf endosphere were combined into phyllosphere microorganisms for analysis. To evaluate the influence of season and sampling location on α-diversity measures (Richness, ACE, Shannon, Simpson, Fisher phylogenetic diversity, and Goods coverage), a one-way analysis of variance (ANOVA) was utilized. The 1% level was used to determine statistical significance, and p-values were adjusted by applying the false discovery rate (FDR) (Benjamini et al., 2006). To identify the sources of variation, the agricolae package’s LSD test function was utilized (Mendiburu, 2020). The niche width of bacteria and fungi in the phyllosphere was calculated using the Spaa package (method = Levins) (Jiao et al., 2020), and then visualized using the ggplot2 package (Ginestet, 2011). The Mantel test with 999 permutations was employed to assess Spearman’s correlations between the multi-functionality of bacteria and fungi and various factors, including α-diversity, leaf physicochemical properties, morphological characteristics, and climate. Additionally, using the vegan package (Oksanen et al., 2017), redundancy analysis (RDA) models were utilized to determine the significant influence of leaf physicochemical properties, morphological characteristics, and climate factors on different phyllosphere bacteria and fungi at the phyla-level communities. The structural equation model (SEM) implemented in Amos-24 software was utilized to investigate the causal relationships among climate factors, leaf physicochemical properties, morphological characteristics, bacteria and fungi richness, and phyllosphere bacterial and fungal multi-functionality. Additionally, the Co-occurrence networks were constructed in Gephi,¹ using a Spearman’s correlation matrix with an absolute correlation coefficient (r > 0.7) and FDR-adjusted (p < 0.001).

The α-diversity of phyllosphere bacteria, as measured by various indices including ACE, Shannon, Simpson, Fisher, Coverage, and Richness, did not exhibit significant differences among three distinct regions (CL, MSW, and TLF) (Figures 3A–F; p > 0.05). Moreover, the ACE and Richness indices showed no significant variation across spring, summer, and autumn (Figures 3G, L; p > 0.05). The phyllosphere bacteria exhibited notably decreased values for the Shannon, Simpson, and Fisher phylogenetic diversity during the spring season when compared to both summer and autumn (Figures 3H–J; p < 0.05). In spring, the Goods coverage index (GCI) of phyllosphere bacteria showed notably higher values compared to both summer and autumn (Figure 3K; p < 0.05).

The ACE, Fisher, Coverage, and Richness indexes of phyllosphere fungi exhibited no significant differences among the three typical regions (CL, MSW, and TLF) in terms of diversity (Figures 3M,P,Q,R; p > 0.05). However, the Shannon and Simpson indexes of phyllosphere fungi were significantly higher in CL compared to TLF (Figures 3N,O; p < 0.05). Additionally, the Simpson index of phyllosphere fungi did not show any significant variations across spring, summer, and autumn (Figure 3U; p > 0.05). In spring, the phyllosphere fungi exhibited significantly higher ACE, Shannon, Fisher phylogenetic diversity, and Richness indexes compared to summer and autumn (Figures 3S,T,V,X; p < 0.05). In contrast, the phyllosphere fungi exhibited a notably decreased GCI during the spring season in comparison to both summer and autumn (Figure 3W; p < 0.05).

The multifunctionality of phyllosphere bacteria and fungi does not exhibit significant differences across the three typical regions (CL, MSW, and TLF) (Figures 4A,D; p > 0.05). During spring, the multi-functionality of phyllosphere bacteria was significantly lower compared to summer and autumn, while the multi-functionality of phyllosphere fungi displayed the opposite pattern (Figures 4B,E; p < 0.05). Furthermore, with the increase in richness, the multi-functionality of phyllosphere bacteria and fungi increased, indicating robust stability (Figures 4C,F; p < 0.001) (Equations 1, 2).

The first two axes of multi-functionality were selected by principal component analysis (PCA). In terms of the multifunctionality exhibited by phyllosphere bacteria, no significant differences were observed across three representative regions (CL, MSW, and TLF) (Figures 5A,B; p > 0.05). It was discovered that the multi-functionality of phyllosphere microorganisms in the primary dimension was diminished in spring in contrast to summer and autumn, whereas the multi-functionality in the secondary dimension was elevated in spring as opposed to summer and autumn (Figures 5C,D; p < 0.05). No significant variation in the multifunctionality of phyllosphere fungi was detected across the three typical regions (CL, MSW, and TLF) on the first axis and different season (spring, summer, and autumn) on the second axis (Figures 5E,H; p  > 0.05). Moreover, the multi-functionality of phyllosphere fungi in the second axis was significantly lower in CL compared to TLF (Figure 5F; p < 0.05). Furthermore, the multi-functionality of phyllosphere fungi in the first axis exhibited higher values during spring as opposed to summer and autumn (Figure 5G; p < 0.05).

The network analysis conducted on phyllosphere bacteria and fungi across various seasons and regions revealed that during spring, the nodes, edges, and average degree of phyllosphere bacteria were comparatively lower than those observed in summer and autumn, and that of phyllosphere fungi showed the opposite trend. Moreover, in the MSW region, the nodes, edges, and average degree of phyllosphere bacteria were higher than those observed in the CL and TLF regions, but phyllosphere fungi showed the opposite trend. The network structure of phyllosphere bacteria exhibits greater complexity in various seasons or regions compared to that of fungi. In particular, in the springtime, the lowest number of positive and negative connections between the borders of phyllosphere bacteria is noticed, while the highest number of positive and negative connections between the borders of phyllosphere fungi is observed (Figures 6A,B; Supplementary Table S3).

The niche width of phyllosphere bacteria did not exhibit significant differences among three distinct regions (CL, MSW, and TLF) (Figure 7A; p > 0.05). In contrast, the niche width of phyllosphere fungi was significantly higher in CL compared to TLF (Figure 7A; p < 0.05). In spring, the niche width of phyllosphere bacteria notably lower values compared to both summer and autumn (Figure 7B; p < 0.05), however, fungi showed the opposite trend.

There exists a noteworthy association between the multi-functionality of phyllosphere bacteria and α-diversity indexes such as observed, Shannon, and Inv Simpson, as well as a significant correlation with the total potassium content of leaves (Figures 8A,C; p < 0.05). Similarly, the multi-functionality of phyllosphere fungi demonstrates a significant correlation with observed, Chao1, ACE, and Fisher phylogenetic diversity, along with a significant correlation with leaf electrical conductivity (Figures 8B,D; p < 0.05). However, the multi-functionality of both phyllosphere bacteria and fungi does not exhibit a significant correlation with climatic factors (Figures 8E,F; p > 0.05).

Redundancy Analysis (RDA) revealed the influential factors on microbial communities extracted from a total of 10 leaves and 9 climate parameters. The analysis showed that the diversity of phyllosphere bacteria displayed clustering patterns during various seasons at the three sampling locations, suggesting a greater degree of multi-functionality (Figure 9A). In contrast, the diversity of phyllosphere fungi was more scattered, especially in the spring (Figure 9C). The diversity of phyllosphere bacteria was notably affected by L-SLA (p < 0.05, R² = 0.072) and temperature (p < 0.05, R² = 0.035; Figure 9B). The diversity of phyllosphere fungi was greatly affected by L-EC (p < 0.001, R² = 0.262), L-LOC (p < 0.05, R² = 0.093), and horizontal radiation (p < 0.001, R² = 0.116; Figure 9D). Moreover, during the spring, a direct relationship was observed between horizontal radiation and the diversity of phyllosphere fungi (Figure 9C). The results indicate that the influence of leaf morphological characteristics and climatic conditions on the diversity of phyllosphere fungi is more significant compared to phyllosphere bacteria (Figures 9A–D).

The findings from the analysis of SEM demonstrated that a combination of climate, leaf physicochemical properties, leaf morphological characteristics, and bacterial richness together explained 42% of the variations in multi-functionality observed in phyllosphere bacteria. Leaf morphological characteristics and bacterial richness were found to have a direct impact on the multi-functionality of phyllosphere bacteria. Furthermore, the influence of climate on the multi-functionality of phyllosphere bacteria was found to be indirect, mediated through alterations in the physicochemical properties and morphological characteristics of the leaf. The impact of leaf morphological characteristics and bacterial richness on the multi-functionality of phyllosphere bacteria was greater than that of climate and leaf physicochemical properties. The indirect and total effects of climate on the multi-functionality of phyllosphere bacteria were observed to be greater than those of leaf physicochemical properties, leaf morphological characteristics, and bacterial richness (Figure 10A).

However, the climate, leaf physicochemical properties, leaf morphological characteristics, and fungal richness together explain 17% of the multi-functionality changes of phyllosphere fungi. Specifically, the physicochemical properties and morphological characteristics of leaves, as well as the richness of fungi, directly impact the multi-functionality of phyllosphere fungi. Additionally, climate indirectly affects the multi-functionality of phyllosphere fungi by altering the physicochemical properties and morphological characteristics of the leaf. The impact of leaf physicochemical properties, leaf morphological characteristics, and fungal richness on the multi-functionality of phyllosphere fungi was greater than that of climate. Conversely, the indirect and total effects of climate on the multi-functionality of phyllosphere fungi were greater than physicochemical properties, morphological characteristics, and fungal richness of the leaf (Figure 10B).

Finally, the potential influence of leaf physicochemical properties on the multi-functionality of phyllosphere bacteria may be ignored, and the negative direct effect on the multi-functionality of phyllosphere fungi is greater. Leaf morphological traits exhibited a positive direct impact on the multi-functionality of phyllosphere bacteria, albeit to a lesser extent than the positive direct impact on the multi-functionality of phyllosphere fungi (Figures 10A,B).