Seed Surface Sterilization Can Alter Root Microbiomes, Increase Endophyte Diversity and Enhance Plant Growth

Featured Application

Plant microbiome researchers should take into account that surface sterilizing their seed before planting could be creating experimental artifacts.

Abstract

With the goal of bioprospecting for growth-promoting endophytes that could become yield-enhancing inoculants in maize agriculture, we isolated 129 endophytic bacteria from 22 wild plants growing in a nature preserve and an urban park in Cali, Colombia. These strains were put through a bioassay with surface-sterilized seeds of perennial ryegrass (Lolium perenne) grown in sealed tubes, and growth promotion assessed by measuring plant fresh weight. The top two strains Pseudomonas delhiensis and Serratia marcescens, along with five different subcultured root endophytic communities, were put into a secondary screen along with two uninoculated controls of untreated and surface-sterilized seed of both the turfgrass and a commercial maize hybrid. Impact on plant microbiomes was assessed using molecular fingerprinting and high-throughput sequencing. This second bioassay indicated that plant growth promotion was corelated not with inoculation but with seed surface sterilization which shifted root microbiomes, increased endophyte diversity and probably eliminated pathogens. Inoculating maize (but not ryegrass) seed with either pure bacteria or microbial communities was also able to shift the root microbiome. Because the majority of plant microbiome researchers employ seed surface sterilization as a method to standardize their experiments, they could be inadvertently studying unusual plant phenotypes and microbiomes; a possible reason why field trials correlate poorly with those of lab tests.

1. Introduction

Corn (Zea mays ssp. mays) is a giant grass which was domesticated about 9000 years ago in southern Mexico [1] and which has become the third-most-important crop in the world [2]. Maize is not only vital for human consumption, but is also used in animal feed, biofuel production and in the manufacture of industrial products [3]. The top global producers of maize are the United States, China and Brazil, while Colombia comes in at 50th [4], fulfilling only 26% of its demand through domestic production, while being the largest importer in South America and the 7th in the world [5]. Colombia’s deficit is mostly due to low productivity, with half of farmers planting landraces in small plots in a traditional manner with little or no inputs and obtaining yields of 2 t/ha. Technologically intensified Colombian farmers meanwhile obtain maize yields of about 5 t/ha, which is still less than half (11 t/ha) that of the average grower in the United States [5].

Although further adoption of modern agricultural technologies such as elite hybrid seed, fertilizers, pesticides, machinery, etc., will no doubt play an important role in increasing Colombian maize production, development of homegrown microbial technologies could also could contribute to sustainably increasing yields [6,7]. To be clear, the main purpose of our research was to bioprospect in Colombian plants for endophytes that could be used to develop microbial inoculants for improving maize production. The application of beneficial bacteria to plants can stimulate growth through the production of phytohormones, fixing nitrogen endophytically, reducing plant stress by reducing ethylene production, upregulating plant stress resistance genes and solubilizing minerals in the rhizosphere, as well as suppressing pathogens and reducing plant diseases. Wild plants have co-evolved with such microbes for hundreds of millions of years [7] and continue to depend on them for their help in growth and survival. It has been shown that during angiosperm evolution, plant microbiomes have been changing [8,9], and maize in particular has been shown to have lost or changed some of its associated microbiota during its domestication, breeding and dispersal all around the planet [10,11,12,13]. Bringing back some of these lost microbes which still exist in wild plants (termed by some as “back to the roots”) may be a powerful and sustainable way to enhance the productivity of modern agriculture while further valorizing plant biodiversity [14]. Considering only plants, Colombia is the second-most-biodiverse country on Earth with at least 28,947 species [15] that must likewise harbor a vast, undiscovered diversity of plant-endosymbionts, which, with a newfound national peace, there is greater opportunity to study [16].

Microbial biostimulants promote crop productivity and plant nutrition by boosting nutrient utilization efficiency, priming the plant to resist stress, optimizing plant physiology (e.g., stimulating root growth), improving product quality (e.g., grain chemistry) and increasing the quantity and availability of soil nutrients [17,18]. Some examples of growth-promoting and yield-enhancing bacterial inoculants for maize include Pivot Bio’s PROVEN 40 (endophytic and diazotrophic Klebsiella variicola), Indigo Agriculture’s Biotrinsic W12 which stimulates root growth and drought tolerance (endophytic Bacillus simplex isolated from maize plus the endophytic fungus Coniochaeta nivea) and TerraMax’s MicroAZ-IF Liquid™, which contains Azospirillum brasilense for root growth stimulation and nitrogen fixation. Many other examples of maize growth-promoting and yield-enhancing endophytes have been reviewed elsewhere [19,20]; however, to our knowledge no growth-promoting endophytes have been discovered and developed for maize production in Colombia.

Ideally, screening of yield-enhancing microbial inoculants would be performed directly on the intended crop variety growing under agricultural conditions; however, this is usually impractical or impossible due to budgetary and logistical considerations, as well as high levels of variation complicating field data. In other words, few labs have the funding or space to treat thousands of seeds with hundreds of microbes and grow them to harvest for one entire agricultural cycle (i.e., 5 months for maize); thus, initial rounds of screening are often performed in smaller and quicker bioassays under controlled laboratory conditions. We have previously had success screening bacterial endophytes for plant growth potential using a gnotobiotic bioassay involving tissue-cultured potato explants growing in tubes with sterile Murashige and Skoog agar, allowing the plant to grow together with the bacteria for several weeks and develop an exaggerated phenotype [10]. However, a theoretical problem with this approach in our search for maize inoculants is that host preference may cause many monocot-adapted endophytes to behave badly in a dicot bioassay (potato). We thus elected to adapt an existing growth promotion bioassay for our purposes, screening endophytes from wild monocots on a small turfgrass, Lolium perenne also known as perennial ryegrass (hereafter referred to as Lolium, turfgrass or ryegrass), grown from surface-sterilized seed inside glass tubes on sterile Murashige and Skoog agar [21].

This research sought to discover maize growth-promoting and yield-enhancing bacterial inoculants through the bioprospection of endophytic bacteria in wild Colombian monocots. After endophyte isolation and identification, we screened them for growth promotion potential using a small, rapid, economical and controlled turfgrass-based bioassay in our laboratory in Cali, Colombia. Promising candidates were put through a second round of the bioassay, which included a commercially relevant variety of maize grown under similar conditions. Curious about the possibility of transplanting entire communities, endophytic root microbiomes of several wild monocots were also subcultured and inoculated onto tube-grown turfgrass and maize. In order to verify strain colonization and observe the dynamics of bacterial communities, we extracted DNA for molecular fingerprinting (terminal-fragment length polymorphism or TRFLP) and high-throughput sequencing (Illumina Miseq) of the bacterial 16S diversity in the perennial ryegrass and maize. Although our search for growth-promoting endophytes did not yield significant results, our inclusion of uninoculated and unsterilized controls allowed us to observe that the seed surface is an important part of microbiome transmission from seed to plant. The implications of the seed surface being involved in vertical transmission have potentially important implications for experimental science and plant microbiome engineering, which we will discuss.

2. Materials and Methods

2.1. Bioprospecting for Plants

For the collection of monocotyledonous plants, we selected two parks in or near the city of Cali, Colombia: a national nature reserve and an urban park. The first site was PNN Farallones de Cali, located at coordinates 3.3299212898168777 Latitude, −76.65360468628405 Longitude. This national park located in the occidental cordillera west of Cali contains a great diversity of vegetation thanks to its watersheds, abundant rainfall and protected forest. The second site is Ingenio park within the city of Cali and located at coordinates 3.3867858490791756 Latitude, −76.53074931880121 Longitude. This urban park is recognized as an important green space within the city, through which runs the river Melendez and which receives a significant level of pollution and disturbance due to human activities. Eleven samples of monocotyledonous plants were collected from the nature preserve and fourteen from the urban park.

2.2. Sample Processing and Isolation of Endophytic Bacteria

The collected plants were cleaned and surface sterilized following a protocol somewhat similar to previously established guidelines [22]. A total of 10 g of root tissue, 10 g of stem and 10 g of leaves were taken from each plant. These were subjected to a rigorous cleaning and disinfection process, including an initial wash in tap water, shaking in 2.5% sodium hypochlorite for 5 min, a 20 min wash in 70% ethanol and finally 3 rinses with sterile water.

Adapting a protocol used for isolating endophytes from seeds [23], the surface-sterilized plant tissue samples were placed in 15 mL conical tubes with three banded tungsten carbide beads and 3 mL of sterile water. The tubes were agitated at maximum speed in a FastPrep-24™ homogenizer for 60 s. The ground tissue slurry was serially diluted 4 times in trypticase soy broth (TSB) and 0.1 mL of each was sown on soybean trypticase agar (TSA). After 5 days of incubation at 25 °C, Petri dishes contained a variety of microbes. The bacterial colonies were characterized by their morphotype and purified using the agar striation depletion technique on fresh TSA. After streaked bacterial colonies were confirmed pure, the purified strains were preserved at −80 °C in a mixture of 30% glycerol diluted with sterile TSB.

2.3. Amplification and Sequencing of 16S rDNA from Bacterial Isolates

For the molecular identification of the purified bacteria, the 16S rDNA of each was amplified using colony PCR [24] and sequenced. Template DNA was obtained by boiling a pipette tip full of each colony in 100 μL of sterile water for 5 min. A PCR protocol similar to one we used previously [10] was carried out with a Mastercycler Nexus Thermal Cycler (Eppendorf, Framingham, MA, USA), containing 0.5 μL of dNTPs at 25 mM, 2.5 μL of Thermopol Reaction Buffer, 0.3 μL of Taq ThermoPol DNA polymerase (New England Biolabs, Ipswich, MA, USA), 0.5 μL of two universal 16S bacterial primers at 10 mM: 27-F (5′-AGAGTTTGATCMTGGCTCAG-3′) and 1492R (5′-GGTTACCTTTTTACGACTT-3′), 19.7 μL of Milli-Q water and 1 μL of template DNA. PCR conditions included denaturation at 95 °C for 30 s, followed by 35 cycles [95 °C for 20 s, 50 °C for 30 s, 68 °C for 1 min] and a final step of 68 °C for 3 min. The resulting PCR product was carried forward into a semi-nested PCR of our own design consisting of 30 cycles using the 27-F and 1387-R primers (5′-GCCCGGGAACGTATTCACCG-3′) and the same conditions as the previous reaction. The expected amplicon of this second reaction was screened for as a 1360 bp band visualized with UV light in a 1% agarose electrophoresis gel containing GelRed. PCR products were purified by precipitation with ethanol and sodium acetate, and quantified on the NanoPhotometer™ NP80 (Thermo Fisher Scientific, Waltham, MA, USA) equipment. The products were sent to Genewiz in the USA for SANGER sequencing. Sequence data were screened by BLASTing against the National Center for Biotechnology Information (NCBI) database, according to the best match with a homology greater than 97%.

2.4. Taxonomic Identification of Wild Plants

Taxonomy was gleaned from photograph analysis of plants in situ and later as samples in the lab. Identities were established by iterative queries posed first of photos submitted to the Android app PlantNet (Version 3.23.0), then asking Microsoft Copilot (May 2025 release) what species it could be if sampled in the park near Cali, Colombia. A phylogenetic tree of plants sampled was drawn to the level of genus with the software Powerpoint (Microsoft^® Word LTSC Version 2108, Redmond, WA, USA) based on established plant taxonomy.

2.5. Using a Plant Bioassay to Screen Endophytes for Growth Promotion Potential

To screen purified endophytes for potential benefit to maize growth, we decided to adapt the turfgrass bioassay methodology suggested by others [21]. Perennial ryegrass (Lolium perenne) seed of the Embassy variety was purchased locally from the Anasac International Corporation (Lampa, Chacabuco, Chile). As the assay calls for growth in sterile tubes, seeds were rinsed in water with 0.01% Tween 80 for 10 min, then transferred to two successive 3% sodium hypochlorite soaks for 10 min. After bleaching, seeds were washed with 70% ethanol for 10 min and finally the seeds were thoroughly rinsed six times with sterilized distilled water.

To verify surface sterility, five seeds per batch were rolled around on 25% TSA Petri dishes, which were incubated at 25 °C for 8 days. Surface-sterilized seeds were germinated on moist, sterilized paper towels inside Petri dishes and incubated for 8 days at 25 °C in the dark.

Endophytes were reactivated from glycerol and cultured in 25% TSB at 25 °C with agitation for 24 h. For each suspension, the optical density was adjusted to OD₆₀₀ = 0.2 and then used to soak germinated seeds of the perennial ryegrass. As a positive control, the growth-promoting strain Burkholderia phytofirmans PsJN [25] was included, while as a negative control we included the biocontrol strain Burkholderia gladioli 3A12 [26]. Uninoculated seedlings were instead dipped in sterile 25% TSB.

For each treatment, nine similar Lolium seedlings were immersed in either bacterial suspension or sterile broth. A total of 132 treatments were included, corresponding to 129 endophytes isolated from monocotyledonous plants, plus 1 positive control, 1 negative control and 1 uninoculated control. After immersion, 3 seedlings were transferred to a 15 cm × 25 cm glass tube containing 10 mL of sterile Murashige and Skoog Basal Medium (MS) agar at pH 5.8. There were 3 tubes per treatment and they were incubated in racks within a F-740 Plant Growth Chamber (Taiwan Hipoint Corporation, Kaohsiung City, Taiwan) at 30 °C and 16 h of white fluorescent light (Philips F72T8/TL841/HO 65 W, 115–145 μmol m⁻² s⁻¹). After 2 weeks, the plants were removed from agar, rinsed in sterile water, blotted on sterile paper towel and air-dried for 20 min. Each seedling was then weighed on a scientific scale.

2.6. Transplanting Wild Plant Microbiomes and Inoculating Pure Bacteria into Lolium and Maize

Lolium and FNC 8134 yellow maize (FENALCE, Bogota, Colombia) were grown for a month in pots with soil from the urban park, then harvested, surface sterilized, triturated, mixed with sterile glycerol and frozen at −80 °C for later. Similarly, 3 new monocotyledonous plants were collected from the urban park, surface sterilized and triturated, mixed with sterile glycerol and frozen at −80 °C. Later, the root microbiome glycerol stocks were reactivated by inoculating sterile TSB in test tubes and incubated for 24 h at 25 °C. In preparation for inoculation, broth in each test tube had its optical density adjusted to OD₆₀₀ = 0.2.

As before, seeds from new batches of both Lolium and FNC 8134 yellow maize were surface sterilized (except for unsterilized controls) and germinated. With 5 repetitions per treatment, each seed was handled with sterile tweezers and immersed in 2.5 mL of microbial inoculum before being placed in a 15 cm × 25 cm glass tube (C5916, Sigma, St. Louis, MO, USA), with a plastic lid (C5791, Sigma, St. Louis, MO, USA). Control seeds (uninoculated) were soaked in sterile TSB. For maize each tube contained 2.5 g of autoclaved peat moss and 7.5 g of three-fold autoclaved sand (a total of 10 g per tube), while Lolium was grown in tubes in 10 mL of sterile MS agar. The tubes were incubated under the previously specified conditions and allowed to grow for 8 days for maize (until some plants were pressing on the cap) or 15 days for perennial ryegrass. Plants were harvested and blotted dry before measuring fresh weight. Roots were rinsed with sterile water and frozen for later extraction of DNA.

2.7. Terminal Restriction Fragment Length Polymorphism Analysis (TRFLP)

For TRFLP, we employed a protocol we had previously developed [27]. Total DNA was extracted from broth or root samples using DNeasy Plant Mini Kits (Qiagen, Germantown, MD, USA) and quantified with a NanoPhotometer™ NP80 (Implen, Westlake Village, CA, USA). To conduct TRFLP of 16S rDNA a first round of PCR was prepared thusly mixing 5 μL of standard Taq buffer (NEB, Ipswich, MA, USA), 1 μL of dNTP mixture (25 mM), 1 μL of 63F-Degen primer (10 mM, CAGGCCTAAYACATGCAA), 1 μL of 1389R primer (10 mM, HGGHTACCTTGTTACGACTT), 0.6 μL of Taq (NEB, USA), 25 ng of template DNA and double-distilled water up to a total volume of 50 μL. Amplification was performed on a Mastercycler Nexus Gradient Eco thermal cycler (Eppendorf, Framingham, MA, USA) with the following program: 95 °C for 30 s, followed by 35 cycles (95 °C for 15 s, 55 °C for 30 s, 68 °C for 1 min) and a final extension at 68 °C for 10 min.

Under the same conditions, a semi-nested PCR was performed using 1 μL of the previous PCR product as a template, with fluorescently labeled primers: 799f (AACMGGATTAGATACCCKG) labeled with 6FAM and 1389R (HGGHTACCTTTTTACGACTT) labeled with Max (Integrated DNA Technologies, Coralville, IA, USA). After visually checking for successful amplification in an agarose gel, 1.5 μL of the labeled PCR product was mixed with 1U Dde I and 1x Buffer 3 (NEB, USA) and incubated at 37 °C for 16 h in the dark. The fragments were separated and visualized on a 3730xl Genetic Analyzer, along with GeneScan 1200 LIZ Size Standards (Applied Biosystems, Foster City, CA, USA).

2.8. Metagenomic Library Preparation and Sequencing

In order to prepare 16S amplicons for sequencing on the Illumina MiSeq platform, a two-step PCR strategy was employed, first amplifying all Lolium and maize root DNA extracts with bacterial 16S primers, before dual labeling them with index sequences. The protocol used for amplification, labeling and sequencing is identical as that used in our previous study [28].

2.9. Bioinformatics

TRFLP results were analyzed with GeneMarker software version 2.2 (SoftGenetics, State College, PA, USA), using the default TRFLP setting with a modified fragment peak height cut-off of 25 fluorescence units for 6FAM, of 100 fluorescence units for Max and manual verification for peak identification. The verified 6FAM and Max signals with fragment sizes of 30 to 800 bp were exported to Microsoft Excel for further analysis. The 6FAM and Max peak height data were combined for each sample and transformed into binary form (0 = absence or 1 = presence) to reduce the experimental noise inherent in TRFLP analysis [29]. The statistical analysis of the molecular fingerprints was performed using covariance principal component analysis (PCA) using XLStat software version 2019.1 (Addinsoft, Paris, France).

MiSeq data were demultiplexed by the commercial sequencing facility and received as one FastQ file per sample. Further sequence processing was performed using USEARCH 11 using the recommended settings. Briefly, paired-end reads were aligned and merged to form full-length sequences called “Uniques,” while quality filtering was performed to remove unmatched and low-quality reads. Next, the program binned these full-length reads together at a similarity threshold of 97% and formed a reference sequence for each bin referred to as an operational taxonomic unit (OTU). Only OTUs represented by two or more raw reads were used for analysis. Bacterial 16S OTUs were assigned a taxonomic identity by Geneious (version 7.0) trained on 16S-ITGDB [30]. OTU counts were exported to Excel (Microsoft, USA) for further analysis and visualization, with statistics achieved by XLSTAT (version 2018.1, Addinsoft, Paris, France). Based on taxonomic annotation, OTUs with lower than 15% identity to a target sequence were hand checked and excluded if they were chloroplasts, mitochondria, plant ribosomes, protists or other non-target sequences. OTU counts were normalized by transformation into proportional abundance as recommended elsewhere [31].

3. Results

3.1. Monocot Sampling and Identification

To build an endophyte library for further assays, we went to two parks in or around Cali, Colombia, and sampled 22 different plants that had a monocot-like morphology. These plants were labeled either as Farallones (National Natural Preserve) or Ingenio (urban park) #1–11 and each was identified taxonomically by image analysis (Supplemental Table S1). A phylogenetic tree was drawn to the genus level to relate these to one another and to the assay organisms Lolium and Zea (Figure 1).

In the national nature preserve, 11 plants were sampled, among them 1 fern (Asplenium serratum—the bird’s nest spleenwort), 1 philodendron, 1 palm, 1 orchid, 1 spiral ginger, 1 heliconia, 1 sedge, 2 bamboos, 1 panicgrass and 1 cool season grass. The latter, Festuca arundinacea, is native to Europe and North Africa [32], however, all other plants in the national nature preserve were native to Colombia.

Of the 11 plants sampled within the urban park, 9 were exotics introduced to Colombia from China, Asia, North Africa or Europe. Festuca arundinacea was also found growing in the urban park, along with four different bamboos native to China, India or Southeast Asia including Phyllostachys vivax (Chinese Timber Bamboo), Phyllostachys sulphurea (sulfur bamboo), Phyllostachys aureosulcata (yellow-groove bamboo) and Bambusa vulgaris (common bamboo). Only two monocots sampled in the urban park were native to Colombia; Gynerium sagittatum (arrow cane—a vigorous grass that grows up to six meters high along river banks) and Tradescantia zebrina (spiderwort, inchplant or wandering jew). For later experiments in microbiome transplants, we also harvested from the urban park some Panicum dichotomiflorum (fall panicgrass or autumn millet—native to Eastern USA and Canada), Coix lacryma-jobi, (Job’s tears—native to India and Southeast Asia) and Dracaena sanderiana (lucky bamboo native to Central Africa), although these are not included in the phylogenetic tree.

3.2. Endophyte Isolation from Wild Plants

Root, stem and leaf samples were crushed, diluted with water and spread out onto TSA in Petri dishes. Seventy different endophytic bacteria were isolated from the plants of the national park, and fifty-nine were isolated from the plants sampled from the urban park (Figure 2). Ingenio11 (Phyllostachys aureosulcata) yielded no culturable endophytes. The sequences of each of the endophytic bacteria with their respective taxonomic identification are given in Supplemental Table S2 with accompanying Genbank IDs ranging from PV638890 to PV639007.

The majority (60%) of cultured endophytes were identified belonging to the phylum Proteobacteria, followed by Firmicutes (23%), Actinobacteria (12%) and finally Bacteroidetes (4%). Of the 129 cultured endophytes, 4 were not able to be identified. There were some differences in proportions of endophytes isolated from the urban versus the national park; for example, 17 versus 9% Actinobacteria, 2 versus 6% Bacteroidetes, 31 versus 16% Firmicutes and 69 versus 50% Proteobacteria, respectively,

The top seven genera of isolated bacteria were, in order of abundance, Bacillus (23), Pseudomonas (16), Enterobacter (9), Microbacterium (8), Pantoea (8), Burkholderia (7) and Rhizobium (4). Conversely, the majority (24) of genera isolated were rare, occurring only once across all plants sampled from both parks.

3.3. Screening of Pure Bacterial Strains for Plant Growth Ability in a Ryegrass-Based Bioassay

To screen endophytes for their ability to promote plant growth, 128 of the strains (missed testing CI47) were inoculated onto surface-sterilized Lolium perenne seeds and co-cultivated with three seedlings for 3 weeks inside glass tubes containing sterile growth media [21]. After 2 weeks of growth, plantlets were harvested, blot dried and fresh weights measured (Figure 3). Excluding the positive control inoculant Burkholderia phytofirmans PsJN (which non-significantly increased plant biomass), ten (8%) of the strains tested were able to increase Lolium fresh weight in this bioassay, although none of these results were statistically significant (using the Kruskal–Wallis test). The majority (92%) of the strains reduced turfgrass biomass and 78 of these did so significantly (using the Kruskal–Wallis test) including the negative control Burkholderia gladioli 3A12 which weighed on average 40% less. Of the beneficial strains, the top three were CI90, CI15 and CI13, which increased turfgrass biomass relative to the uninoculated controls (average weight = 0.028 g), with fresh weights that were 13%, 11% and 11% higher. CI90 was isolated from roots of Farallones9 (the native palm Geonoma undata) and was identified as Pseudomonas delhiensis. CI15 and CI13 were both isolated from roots of Ingenio3 (the introduced grass Festuca arundinacea) and were identified as Serratia marcescens and Bacillus subtilis, respectively.

3.4. Impact of Seed Sterilization and Inoculation on Tube-Grown Turfgrass and Maize

3.4.1. Plant Biomass

In order to repeat the bioassay with top performing endophytes and to observe the impact of different seed treatments on the plant microbiome, the two top strains CI90 and CI15 were inoculated onto surface-sterilized Lolium perenne seeds and unto surface-sterilized maize seeds. Similarly, subcultured root endospheres from five different plants growing in urban park soil (Lolium perenne, Zea mays ssp. mays, Coix lacryma-jobi, Dracaena sanderiana, Panicum dichotomiflorum) were inoculated onto surface-sterilized turfgrass seeds and unto surface-sterilized maize seeds. As controls for inoculation, uninoculated sterilized and unsterilized Lolium and maize seeds were included in the experiment. Seeds were planted one per tube in five tubes per treatment and left to grow for either 2 weeks in agar for Lolium (note that this was fewer plants grown for less time than the previous assay) or 1 week in sterile peat moss (maize), before harvesting, weighing, DNA extraction, TRFLP and microbiome sequencing on the Illumina Miseq platform. Plant fresh weights and representative photos of each treatment are shown in Figure 4.

Both Lolium perenne (Figure 4A) and maize (Figure 4B) tended to increase in size thanks to seed sterilization and inoculation, with plants growing from untreated seed being among the smallest. Looking at average fresh weights, unsterilized, uninoculated Lolium (Figure 4C) were the second smallest plants (after maize microbiome treated turfgrass with 0.019 g/plant) with an average of 0.021 g/plant. Surface sterilization (without inoculation) of Lolium increased plant weight by 47% to an average of 0.032 g and only inoculation with Serratia marcescens CI15 and the subcultured Dracaena microbiome was able to promote further increases to 0.033 and 0.036 g, respectively. Other fresh weight averages varied from 0.03 g/plant for sterilized seed inoculated with CI90, to 0.031 g/plant for sterilized seed re-inoculated with Panicum endosphere microbiome—less than plants grown from uninoculated, surface-sterilized seed. Shannon diversity index (SDI) calculated with Miseq data generally correlated with increasing plant biomass, going from 1.5 from unsterilized seeds (the second lightest plants) up to 2.2 and 2.3 for surface sterilized seeds without inoculation and with CI15 inoculation, the third and second heaviest plants respectively). Of note, the average fresh weight for sterilized/uninoculated plants in this experiment was 12% larger than it was during the first set of bioassays reported in Figure 3. Meanwhile, the weight of sterilized/CI90 or CI15 inoculated plants turned out similar to that observed in the first set of bioassays.

Seed treatments of maize resulted in similar changes to plant biomass as those observed in turfgrass. Average fresh weights of maize grown from unsterilized, uninoculated seedlings (Figure 4D) were middle of the pack, coming in at 1.26 g/plant. Seed surface sterilization alone increased maize weight by 13.7% to an average of 1.43 g and only inoculation with the subcultured turfgrass microbiome was able to further increase it to 1.44 g. Other fresh weight averages varied from 1.29 g/plant for sterilized seed inoculated with Serratia marcescens CI15, to 1.39 g/plant for sterilized seed re-inoculated with Panicum endosphere microbiome. Maize grown from surface-sterilized seed and inoculated with subcultured Dracaena, Coix and maize microbiomes had biomasses lower than even unsterilized seed, presumably due to the activity of antagonists in the undefined microbial inoculum. Shannon diversity indexes calculated with Miseq data generally correlated with increasing maize biomass, going from 0.8 for the smallest plants (sterilized seeds inoculated with Dracaena microbiomes) to 1.3 for unsterilized seeds and 2.9 for surface-sterilized seeds without inoculation (the second heaviest plants).

3.4.2. Using TRFLP of Root DNA Extracts to Monitor Microbial Dynamics

To understand what changes in endophyte populations might accompany variations in plant growth phenotype, we generated TRFLP-based molecular fingerprints for each sample, transformed them into presence/absence data (Supplemental Table S3), then did multivariate statistics using PCA. Comparing the sterile peat moss on which maize was grown to the pure strain inoculants and the complex, subcultured root endosphere inoculants, the largest differences observed were between inoculants and the sterile substrate (Figure 5A). Being relatively close to the origin (0, 0) this result confirms that the sterile peat moss is the material with the least diversity of 16S rDNA. Furthest away from the peat samples is a cluster of the subcultured root endosphere microbiome inoculants, all of which presumably contain a large diversity of different bacteria. The pure strain inoculants were placed midway between the sterile substrate and the subcultured root endospheres, although CI90 clustered somewhat with the mixed community inoculants, suggesting possible contamination.

Looking at TRFP data from the first round of screening, the greatest difference visible between the microbiomes of 2-week-old turfgrasses are those of plants coming from unsterilized/uninoculated seed versus all other treatments (Figure 5B). The rest of the turfgrass samples forming a cluster in the bottom right, all derived from surface-sterilized seed, suggesting that inoculation with either pure strains of endophytic bacteria or complex subcultured root microbes did not significantly alter Lolium root microbiomes.

Tube-grown maize TRFLP samples (Figure 5C) organized into a pattern very similar to those of turfgrass, with plants coming from unsterilized/uninoculated seed having microbiomes significantly different from those of the rest coming from sterilized seed. Unlike turfgrass, inoculation of surface-sterilized maize with either a pure strain or a subcultured root endosphere appears to have shifted their microbiomes noticeably away from that of surface-sterilized/uninoculated seeds. That is to say, three different clusters emerged for root microbiomes of tube-grown maize plants: that from unsterilized/uninoculated seed, that from sterilized/uninoculated seed and that from sterilized/inoculated seed.

3.4.3. Using High-Throughput Sequencing of Root DNA Extracts to Monitor Microbial Dynamics

To help understand the taxonomy of the most abundant bacteria in the turfgrass and maize microbiomes, the same plant DNA extracts previously studied with TRFLP were also used for PCR of 16S rDNA and sequenced on the Illumina Miseq platform (Supplemental Table S4). Figure 6A shows proportional averages of the top 24 most abundant bacterial 16S OTUs observed in tube-grown turfgrass. Looking across the nine sample types, almost all OTUs occurred in all samples, with little variation in abundance suggesting that the dominant bacteria making up Lolium root microbiomes are transmitted internally through seed. The most abundant OTU was #2 representing a strain of Klebsiella which made up between 37 and 67% of the bacterial population. We expected to see elevated levels of Serratia CI15 and Pseudomonas CI90 in plants that had been inoculated with these pure strains. The OTUs 21 and 33 matching these bacteria’s 16S rDNA were observed in all Lolium plants. The highest levels of Serratia OTU21 was 9.7% in plants which had previously inoculated with the bacteria, followed by 6.4% in plants coming from surface-sterilized/uninoculated seed; the plants with the lowest average abundance of OTU21 came from surface-sterilized seed inoculated with Lolium root endosphere at 0.9%. Pseudomonas OTU33 likewise appeared in all treatments, but turfgrass seed that was inoculated with this bacterium resulted in plants with the lowest levels of the strain, at only 0.1%. For contrast, the plants with the highest levels came from surface-sterilized seed inoculated with Coix microbiomes at 0.7%.

Figure 6B shows proportional averages of the top 24 most abundant bacterial 16S OTUs observed in tube-grown maize. Like Lolium, almost every OTU was found in all samples suggesting seeds are transmitting these bacteria internally. Contrary to turfgrass, there was a lot of variation in OTU abundance between samples, suggesting that maize seed surfaces transmit more of the root microbiome and are perhaps better suited for microbial inoculation. The most abundant OTU was again #2-Klebsiella, which was lowest in unsterilized/uninoculated plants (6.3%) and highest in plants that had been inoculated with subcultured root endosphere microbiomes, representing up to 85.7% of the endophytic bacteria detected. OTU7, representing a Stenotrophomonas, was the most abundant OTU detected in unsterilized/uninoculated maize at 50.6%, being much lower in plants whose seed had been surface sterilized and inoculated, representing as little as 0.6% of the sterilized/SerratiaCI15 inoculated microbiome. Seed that was surface-sterilized but not inoculated resulted in plants with the highest levels observed of 28.1% for OTU42 representing a strain of Pseudomonas aeruginosa. We expected to see elevated levels of Serratia CI15 and Pseudomonas CI90 in plants that had been inoculated with these pure strains, and that is what we saw: the highest levels of Serratia OTU21 were 7.9% in plants which had previously inoculated with the bacteria, and the highest levels of Pseudomonas OTU33 appeared in roots grown from seed inoculated with this at 32.6%. Serratia OTU21 was observed in all plants at levels well below 1%, except for Pseudomonas CI90 inoculated seeds where it was found at 6.8% of the total, suggesting possible cross contamination. Curiously, SerratiaCI15 treated plants had the highest levels of Achromobacter brasilense represented by OTU27 at 30.2%.

Using Miseq data to calculate the Shannon diversity index (SDI), maize root bacterial diversity went from 0.8 to 3.0, which was more variable compared to Lolium which only varied from 1.3 to 2.3; this suggests that maize seed surfaces can transmit a greater diversity of bacteria to their roots (Figure 4). In both species, surface sterilization of seeds appeared to be the most important way to increase bacterial SDI, taking it from 1.5 and 1.3 in Lolium and maize, respectively, to 2.2 and 2.9. In some instances, this increase in SDI was related to increasing numbers of OTUs, such as from unsterilized/uninoculated to sterilized/uninoculated maize roots going from 111 OTUs up to 149. In other cases, such as unsterilized/uninoculated to sterilized/uninoculated Lolium roots, SDI increased because evenness went up while OTU number decreased from 74 to 49. Meanwhile, surface sterilization followed by microbial inoculation had variable effects on SDI resulting in both the highest and lowest numbers for both plant species.

4. Discussion

The main purpose behind this experiment was to discover endophytes with the potential to aid Colombian maize agriculture. Why look for these bacteria in wild plants? Evolving, growing and reproducing without the help or interference of humans for hundreds of millions of years, wild plants have had to learn to cooperate with microbes to aid them in nutrient acquisition, defending against pathogens and overcoming abiotic stress [7]. Bioprospecting for these endophytes in the nature preserve/national park allowed us to access a variety of native plants with most of their ecological webs intact, while in the urban park, we expected to find robust invasive plants thriving under stressful conditions in a disturbed/polluted environment. Because we are hoping to develop maize inoculants and previous studies have suggested that endophytes can suffer host range limitations, we opted to sample monocots only.

As expected, the majority of the monocotyledonous plants sampled in the national nature preserve were native (except for non-native Festuca arundinacea), while the reverse was true in the urban park, where 86% of species were exotic. Some of these species from the Farallones de Cali national park such as Guadua angustifolia, Chusquea albilanata, Geonoma undata and Heliconia griggsiana are large plants which might be expected to contain growth-promoting endophytes. Similarly, in the urban park there were many species which grew to great size, including the canes (Gynerium sagittatum and Saccharum spontaneum) and the bamboos (Phyllostachys aureosulcata, P. sulphurea, P. vivax and Bambusa vulgaris)—these giant plants may contain growth-promoting endophytes which help them develop a large size. With an alternative hypothesis that invasive plants contain endophytes which help them become successful weeds [33,34], the weedy species Festuca arundinacea, Cynodon dactylon and Oplismenus hirtellus were promising sources for growth-enhancing endophytes.

It was not surprising to isolate the majority of the endophytes from root tissues, as this is generally the most heavily colonized part of the plant [35]. It was likewise expected to see that Proteobacteria would be the most common phylum of endophytes, as its been observed to makeup the core of the seed transmitted plant microbiome [28]. There are many reports of the genera Bacillus, Pseudomonas, Microbacterium, Enterobacter and Pantoea containing examples of plant growth promoters [36]. While we did observe that many strains (10) tend to improve turfgrass biomass, none, not even the positive control B. phytofirmans PsJN, generated a statistically significant increase; was there a problem with our screening system? For context, a previous bioassay that we conducted with 91 maize seed-derived endophytes inoculated into tissue-cultured potatoes found only 2 of these could increase root and/or shoot fresh weight [10]. Having speculated that perhaps using a dicot bioassay (potato) to screen monocot (maize)-derived endophytes resulted in lower numbers of positive hits due to host incompatibility issues, our colleagues developed a turfgrass-based method for the screening of monocot-derived endophytes [21]. Seventy-five of the same maize seed endophytes from the aforementioned study were retested on annual ryegrass and only one strain was able to significantly increase plant biomass. This was not a greater hit rate than what was seen before using the “incompatible” dicot system to screen monocot-derived endophytes, and frustratingly, the maize seed-derived endophytes Hafnia alvei and Burkholderia phytofirmans which had promoted potato growth, did not also stimulate annual ryegrass growth. We were expecting between three or four Lolium perenne growth-promoting endophytes when screening this collection of 129 monocot-derived endophytes. Perhaps Lolium perenne is less sensitive to microbial inoculation than ideal for this type of screen, and in retrospect, Enterobacter cloacae 3D9 would have been a better choice of positive control since it stimulates annual ryegrass growth [21] and indeed has been patented for its plant growth-promoting abilities [37].

In any case, we moved the top two hits (CI90 and CI15) from this bioassay onwards to a second round of testing, repeating the inoculation experiment with Lolium perenne as well as a commercial variety of Colombian maize. We also subcultured root endospheres from urban park soil grown plants, and applied those as complex, undefined inocula to surface-sterilized seeds with the idea that it may be possible to transplant more than one wild endophyte into a new crop at the same time. Other publications have reported that inoculating seeds (either surface-sterilized or not) with large quantities and diversity of bacterial endophytes can significantly change the plant microbiome, thus we had expected to observe similar results [38]. We hypothesized that inoculating surface-sterilized seed with its own soil grown root endosphere may restore the microbiome to the same state as the unsterilized control.

The most striking phenotypic change observed in Figure 4 is the dramatic impact that surface sterilization had on both Lolium and maize fresh weight, explaining most of the increases observed in heavier inoculated plants (all inoculated seeds were first surface-sterilized). Passage through animal intestinal systems has been observed in hundreds of different plant species to increase germination and vigor by digesting seed mucilage and scarifying the seed coat [39] so perhaps this could partially explain our results. Seeds passaged through bird intestines and excreted in feces have also been shown to enjoy increased seedling survival and disease resistance, although it has not yet been studied whether the plant microbiome is involved [40]. Surface sterilization of both maize and Lolium seeds also strongly increased the SDI of bacterial OTUs, perhaps reducing bacterial load while increasing the evenness of different strains and ensuring a better balance between microbes. This apparent correlation between bacterial alpha diversity and plant biomass has previously been reported between soil microbes and lettuce above ground biomass, although the authors were not able to discover what specific mechanisms might have been involved [41]. Another paper suggests that the correlation between soil bacterial diversity and plant productivity is related to plant disease inhibition [42], although in our experiment, surface-sterilized seeds shouldn’t have been exposed to environmentally transmitted pathogens.

An alternative way that seed surface sterilization could have promoted plant growth was by removing pathogens from the microbiome. This phenomenon might be similar to the doubling of tree growth after planting in fumigated soil that has been depleted of plant pathogens, nematodes and insect pests [43], or the significant increases in the number of tomato leaves and fruit when planted in solarized soil that has been depleted of Fusarium oxysporum f. sp. radicis lycopersici, nematodes and parasitic branched broomrape [44]. During germination in Petri dishes, we observed fungi growing on unsterilized maize and Lolium seed, overwhelming and killing many seedlings before transplant. Lolium perenne seed can be difficult to clean of pathogenic fungi, disqualifying it as a model plant for axenic bioassays [21], although we were satisfied that our surface sterilization was effective. Maize seed also contains harmful fungal pathogens which require varying levels of surface sterilization combined with antibiotics before having pathogen-free seedlings [45]. Seed transmission of human pathogens to the plant microbiome is a significant problem in the fresh sprout industry, with chemical or physical sterilization of seed surfaces being important to avoid contamination [46,47]. Similarly, most commercial crop seeds are chemically treated, usually with pesticide coatings before planting in order to reduce diseases during germination and later on in the plant’s life [48]. Inoculating surface-sterilized Arabidopsis seed with root-derived fungi or oomycetes generally reduced shoot biomass or killed the plant outright (again showing the benefits of removing them from infected seeds), while inoculating with bacteria generally increased plant fresh weight and protected against filamentous microbial eukaryotes [49]. Conversely, maize seeds that have been surface-sterilized (removing many beneficial epiphytes as well as any pathogens) are more susceptible to colonization and attack from environmentally transmitted fungi, germinating and growing at a reduced rate [50].

Our attempts to increase plant biomass via microbial inoculation were not consistent or translatable from Lolium to maize. Inoculating Lolium with Serratia CI15 or the subcultured Dracaena root microbiome increased plant biomass, while in maize they had the opposite effect. This phenomenon of variable plant responses to microbial inoculation is often observed. For example, inoculating various maize genotypes with different growth-promoting endophytes showed that a variety of interactions are possible, with some bacteria promoting growth of all plant genotypes, some with only particular maize genotypes, and others negatively impacting certain corn varieties [51]. Our bioassay suggested that the majority of strains reduce plant biomass, removing them as growth-promoting candidates and perhaps identifying them as interesting for other applications like herbicides or pesticides. After all, the strain we used as a negative control has been patented for use as a biopesticide of turfgrass [52]. While our results were disappointing, many of these strains may nevertheless have potential that needs to be assayed under more agronomically relevant conditions—as we all know, lab results rarely ever translate to the field [53]. Perhaps repeating these assays in a greenhouse system with field soil would increase our positive hits—a meta-analysis of 70 publications testing yield-enhancing microbial inoculants, representing 576 unique treatment–control pairs, found that while greenhouse assays gave higher, more variable results, all tested strains increased yields in the field as well [54].

We had expected that inoculating seeds with microbe(s) would install new members into the community and increase the diversity of bacteria in the root, perhaps shifting the majority of the native microbiome. This has been documented in other research; for instance, after tissue-cultured plantlet inoculation with seed endophyte Xathomonas sacchari, soil-grown rice root, shoot and rhizosphere microbiomes were significantly shifted, with an elevated SDI [55]. In another experiment inoculating wheat seeds with 219 different soil microbial communities, it was found that on average about 50% of the variable root bacterial population came from the soil inoculant while the rest were seed transmitted [56]. In our experiment, we found that inoculation of maize seed with either pure strains or subcultured microbial communities was able to shift the root microbiome away from that of uninoculated controls, but this effect was not evident in Lolium. In both plant species, any sort of inoculation usually reduced the SDI relative to surface-sterilized/uninoculated controls. Inoculating with pure strains of Serratia and Pseudomonas increased dramatically the abundance of that particular bacteria in the resulting maize root microbiome and less so in Lolium. However, based on PCA of TRFLP data, inoculation with pure strains was indistinguishable from inoculation with complex microbiome inoculants that may have contained hundreds of different bacterial species. Perhaps the process of freezing and subculturing the root extracts dramatically reduced species diversity to only a few different strains? Our results suggest that maize seed surfaces are better suited to transmit microbes to the root endosphere than are Lolium perenne seed surfaces, but why this might be is a mystery.

The most significant discovery of these experiments is that surface sterilization of seeds can result in significant shifts in bacterial populations within plant microbiomes. Although it is well known that seed surface sterilization or treatment with pesticide is important to control fungal/oomycete pathogens [48] and we have previously shown that spermospheres also contain a rich diversity of seed transmitted bacteria [28], we believe that this is the first study to provide evidence that seed surfaces transmit a significant portion of the plant microbiome to the developing seedling. Our observation contrasts to the finding of a similar study which concluded that maize seed surface sterilization did not significantly shift the root microbiome or bacterial diversity of 5-day-old axenically grown seedlings, although there were some differences in individual bacterial OTUs, a reduction in cultivatable bacteria and an increased susceptibility to fungal pathogens [50]. A possible explanation between the discrepancy in our conclusions is that our statistical techniques analyzed binary TRFLP data which we have previously found to be at least as sensitive as high-throughput sequencing [57]. Because bacterial members of the plant microbiome have been shown to be important in defending against the negative effects of fungal pathogens [49], it is possible that these plants with altered microbiomes would suffer more disease when grown under normal conditions in microbe-rich soil. It is also possible that when grown under natural conditions this missing seed surface microbiome could be replaced and rebalanced by environmental microbes.

Most laboratory studies of the plant microbiome surface-sterilize seeds as a matter of course, reasoning that this is a necessary step in standardizing results, presuming that the seed surface is a chaotic and unpredictable mix of contaminating microbes. A recent review on seed microbiomes says, for example, “A primary step conducted in assays seeking to gain insights into the diversity of the endophytic microbiota consists of sterilizing the surface of the seeds.” [58]. We have reasoned instead that similar to animal microbiomes that are mostly transmitted via surface contact with the mother [59], seed surfaces are important and natural reservoirs for the plant microbiome. The only imaginable times this seed surface microbiome might be removed is when it is first passaged through an animal’s intestinal system (in the case of wild plants) or if a farmer pays for seed surface disinfestation with physical or chemical methods like hot steam or coating with antibiotics. Scientists planning to study the plant microbiome should ask themselves whether it is more important for them to include all natural microbial reservoirs in their experiment, or sterilize the seed surface in order to remove a possible source of biological variation.

5. Conclusions

This study uncovers a bit of the hidden diversity of endophytic bacteria in wild plants from underexplored ecosystems in Colombia. While our search for growth-promoting potential among our endophyte isolates proved ultimately inconclusive, we have identified some candidates which warrant study in field trials; we also observed the ability of select strains and microbial consortia to reshape maize root microbiomes. Most importantly, we have also shown that seed surface sterilization has a significant effect on root microbiome composition and plant phenotype—probably through the removal of seed-borne pathogens and increasing the evenness of endophytic bacterial species.

These results challenge conventional methodologies in plant microbiome research, suggesting that routine sterilization practices may inadvertently distort natural microbial assemblages and plant responses, potentially explaining discrepancies between laboratory and field outcomes. From an evolutionary perspective, the seed surface microbiome may be a resilient and adaptable subset of the mother plant’s own microbiome that has contributed to plant health and reproduction over hundreds of millions of years. As most plant microbial inoculants are intended for deposition on seeds, the members of this surface microbiome may already be preadapted for this type of agricultural application and warrants further study. Testing our endophytes, as well future isolates from seed surface microbiomes, is an important endeavor for future field trials with maize under realistic agronomic settings. Likewise, the impact of seed surface sterilization on plant health, performance and microbiomes should be studied under field conditions, where we can better understand the seed surface microbiome’s importance in vertical transmission and potential as a leverage point in plant microbiome engineering.