Microbiological quality analysis of inoculants based on Bradyrhizobium spp. and Azospirillum brasilense produced “on farm” reveals high contamination with non-target microorganisms

The use of inoculants carrying diazotrophic and other plant growth–promoting bacteria plays an essential role in the Brazilian agriculture, with a growing use of microorganism-based bioproducts. However, in the last few years, some farmers have multiplied microorganisms in the farm, known as “on farm” production, including inoculants of Bradyrhizobium spp. for soybean (Glycine max L. Merrill.) and Azospirillum brasilense for corn (Zea mays L.) or co-inoculation in soybean. The objective was to assess the microbiological quality of such inoculants concerning the target microorganisms and contaminants. In the laboratory, 18 samples taken in five states were serial diluted and spread on culture media for obtaining pure and morphologically distinct colonies of bacteria, totaling 85 isolates. Molecular analysis based on partial sequencing of the 16S rRNA gene revealed 25 genera of which 44% harbor species potentially pathogenic to humans; only one of the isolates was identified as Azospirillum brasilense, whereas no isolate was identified as Bradyrhizobium. Among 34 isolates belonging to genera harboring species potentially pathogenic to humans, 12 had no resistance to antibiotics, six presented intrinsic resistance, and 18 presented non-intrinsic resistance to at least one antibiotic. One of the samples analyzed with a shotgun-based metagenomics approach to check for the microbial diversity showed several genera of microorganisms, mainly Acetobacter (~ 32% of sequences) but not the target microorganism. The samples of inoculants produced on farm were highly contaminated with non-target microorganisms, some of them carrying multiple resistances to antibiotics.


Introduction
Soybean (Glycine max L. Merr.) and corn (Zea mays L.) are the main Brazilian grain crops [1], with a production ~ 125 million tons in ~ 37 million hectares of soybean, and ~ 102.5 million tons in ~ 18.5 million hectares of corn [2]. The symbiosis between soybean and elite Bradyrhizobium strains can supply the most part of the required N via biological nitrogen fixation (BNF) [3] and grain yield increases by 8% due to inoculation [4]. In corn, yield increase due to inoculation with Azospirillum brasilense has been attributed to bacterial phytohormones [5,6]. Coinoculation of soybean with Bradyrhizobium spp. and A. brasilense has doubled the benefits compared with single inoculation [7,8].
Brazil has a long tradition in research with inoculants containing rhizobia and Azospirillum, and legislation for quality control of inoculants. According to the standards established by the Ministry of Agriculture, Livestock and Food Supply (MAPA), commercial inoculants must have the minimal concentration of 10 9 viable cells of Bradyrhizobium and 10 8 cells of Azospirillum per gram or milliliter of inoculant, no contaminants at the 10 −5 dilution, and must carry only elite strains with recognized agronomic efficiency [9,10].
The industrial production of inoculants is a complex process, but improvements in the last two decades have resulted in high-quality products in terms of cell concentrations, no contaminants, and very low cost, probably the cheapest inoculant in the world [11]. However, in the last five years, some farmers have tried to produce their own bioproducts, including inoculants in the farm, using simplified biofactories, known as "on farm" production. In most cases, the production system is rudimentary and varies in terms of installations, equipment, microbiological standards, and technical capacity. Very often the bioproducts are produced in fermenters, open tanks, or even water tanks, without appropriate control of contaminations, which may result in highly contaminated, non-effective products [12,13].
The objective of this study was to assess the microbiological quality of inoculants based on Bradyrhizobium spp. and A. brasilense produced on farm in Brazil, concerning the intended microorganisms, presence, and characterization of probable contaminants.

Sampling
Sampling and transportation kits containing Styrofoam box, sterile 50-mL Falcon-type conical tubes, sterile 30-mL disposable syringes, disposable gloves, Parafilm M® for sealing the tubes, and cooling packs were sent to farmers interested to know the microbiological quality of their inoculants produced on farm. The kit included a protocol for sampling, emphasizing aseptic procedures and an identification form. Immediately after sampling, two aliquots per tank or fermenter were packed with cooling packs in the Styrofoam box and sent back by express postal service or personally delivered in the Laboratory for Soil Biotechnology at Embrapa Soja. A total of 18 samples were obtained during 2019/20 cropping season, six aiming Bradyrhizobium and 12 aiming Azospirillum as target microorganisms (Table 1). These samples were obtained from five states: São Paulo (six), Bahia (two), Paraná (five), Rio Grande do Sul (three), and Mato Grosso (two). For comparative purposes, commercial inoculants containing A. brasilense strains Ab-V5 and Ab-V6 (C1, lot 1,108,718), B. diazoefficiens strain SEMIA 5080 and B. japonicum strain SEMIA 5079 (C2, lot 0,135,218),

Physical-chemical and organoleptic properties
The samples and the commercial inoculants were evaluated for pH using a pH-meter model FiveEasy Plus pH-meter FP20 (METTLER TOLEDO, Ohio, USA) and electrical conductivity in a digital conductivity-meter Tec-4MP (TEC-NAL, Piracicaba, Brazil). A sensorial analysis was based on the "odor wheel" described by McGinley and McGinley [14], which highlights eight categories of odors.

Isolation of morphotypes
Under aseptic conditions, serial dilutions were made in sterile 0.85% NaCl saline and 100-μL aliquots of the 10 −5 , 10 −6 , and 10 −7 dilutions were spread on five different culture media: modified YMA (Yeast Mannitol Agar) for Bradyrhizobium [15]; RC (Rojo Congo) [16] for Azospirillum; LB (Luria Bertani) [17]; NA (Nutrient Agar) [18]; and Sabouraud [19]. The different culture media aimed to check for occurrence of typical colonies of the target microorganisms, and increase the chance of obtaining as many as possible contaminating isolates able to grow in these culture media.
After spreading on each medium, plates were incubated at 28 ± 1 °C in the inverted position in a growth room and were daily observed for 7 days. The morphologically distinct colonies in each culture medium were streaked again on the same culture medium to select single colonies. To avoid morphologically distinct isolates due to the growth medium, all isolates were streaked on NA to standardize the morphology of colonies. Finally, morphologically distinct isolates in NA medium were cryopreserved in NA broth with 30% glycerol at − 80 °C for further analysis.
Prior to cryopreservation, all isolates were observed at 400 × magnification under an optical microscope (AxioLab A1, Zeiss) coupled to an AxioCam ERc 5 s digital video camera system (Zeiss) for recognition of typical yeast traits such as nucleus, vacuole, and cell dimensions. Isolates identified as yeasts were not submitted to further analysis.

Molecular identification of isolates
Total DNA of morphologically distinct isolates was extracted with the DNeasy Blood and Tissue Kit (Qiagen), according to the manufacturer's instructions. After extraction, the integrity of DNA was verified by electrophoresis in 1% agarose gel. The 16S rRNA gene was amplified as described [20] with universal primers fD1 (5′-AGA GTT TGA TCC TGG CTC AG-3′) and rD1(5′-AAG GAG GTG ATC CAGCC-3′) for phylogenetic studies of bacteria, flanking nearly the entire region of the 16S rRNA gene (~ 1,500 bp) [21]. The PCR products were purified with the PureLink™ Quick PCR Purification Kit (Invitrogen), according to the manufacturer's instructions. Sequencing was performed in an ABI3500xL analyzer (Applied Biosystems) as described [22]. Fragment sequences ranging from 484 to 1139 bp were analyzed using the software Bionumerics version 7.6 and identification was based on comparison with the NCBI GenBank database using the BLAST tool for nucleotides (https:// blast. ncbi. nlm. nih. gov/ Blast. cgi).

Metagenome analysis
To have a broader view of the diversity of microorganisms that might not have grown on the culture media, or occurring at low concentrations in the sample, metagenomic analysis was performed in sample 10, from Palotina, PR. We used the shotgun approach, sequencing all DNA fragments extracted from the sample, without previous amplification of any specific region, as described before [23]. The shotgun approach detects higher diversity in a sample as well as microorganisms in all domains of life and, if required, can also be used for functional analysis. For the metagenomics analysis, total DNA was extracted with the DNeasy blood and tissue kit (Qiagen) and used to build the library with the Nextera XT kit, according to the manufacturer's procedure. The library was processed on the MiSeq platform (Illumina) at Embrapa Soja, and the sequences were assembled with the A5-miseq pipeline (de novo) version 20,140,604. The sequenced fragments were uploaded to the MG-RAST v.4.0.4 (RASThttp:// metag enomi cs. anl. gov) and submitted to automatic annotation in the server based on the NCBI BLAST and SEED databases [24].

Susceptibility to antimicrobials
After molecular identification, isolates belonging to potentially pathogenic genera were subjected to evaluation of susceptibility to antimicrobials by the Disk-Diffusion Test [25]. Cells grown for 24-48 h on NA medium were suspended in sterile saline (0.85% NaCl) until a turbidity compatible with the McFarland scale 0.5 (~ 1.5 × 10 8 CFU mL −1 ). The suspension was then inoculated on the Müeller-Hinton [26] agar plate using a sterile swab. Then, paper disks impregnated with antimicrobials were added, as indicated in the annual updates of the Clinical and Laboratory Standards Institute (CLSI) [27].

Physical-chemical and organoleptic properties
The physical-chemical and organoleptic properties, type of equipment used for multiplication (open tanks or fermenters), and growth time (from inoculation up to sampling) of the 18 samples are shown in Table 1. The pH ranged from 3.6 (sample 6) to 7.2 (sample 7), the latter was the only with slightly alkaline pH, whereas the others were acidic, below pH 6.0. The electrical conductivity ranged from 800 (sample 5) to 8390 μS cm −1 (sample 10). Among the commercial inoculants, pH was slightly alkaline and the one containing A. brasilense presented the highest electrical conductivity. The cell concentration in the commercial inoculant C1 (A. brasilense Ab-V5 and Ab-V6) was 1.01 × 10 9 CFU mL −1 ; in C2 (Bradyrhizobium spp. SEMIA 5079 and SEMIA 5080) was 6.30 × 10 9 CFU mL −1 ; and in C3 (B. elkanii SEMIA 587 and SEMIA 5019) was 8.47 × 10 9 CFU mL −1 . No contaminants were found in the commercial inoculants.
In the sensorial analysis [14], only two samples were classified as "yeast" (samples 13 and 14), whereas the others presented odors classified as "offensive," which might be attributed to putrefaction processes. The commercial inoculants, however, presented odors classified as "vinegar" and "yeast" for Azospirillum and Bradyrhizobium, respectively (Table 1). Among 18 samples, three were declared as multiplied in fermenters, 13 in open tanks, and two were not informed. The growth time ranged from 4 h (sample 5) to 10 days (sample 3).

Bacterial isolation and molecular identification
The plating for isolation in culture media indicated a variety of colony morphotypes, as exemplified in Fig. 1A, suggesting occurrence of contaminants, as they differed from typical colonies of Bradyrhizobium (Fig. 1B) and Azospirillum (Fig. 1C).
A total of 84 morphologically distinct isolates were obtained from the 18 samples (Table 2). Sequencing of 16S rRNA gene resulted in sequences that ranged from 484 to 1140 bp, most of them above 1000 bp. Comparisons of sequences in the GenBank showed 44 isolates with similarity ≥ 99% and 28 between 99 and 97.2% with deposited sequences, and coverage between 95 and 100%. Finally, 12 isolates were identified as yeasts based on the cell morphology (size, presence of nucleus, and budding) and were not sequenced.

Metagenome analysis
The shotgun approach of the sample no. 10 revealed a total of 2,467,209 sequences. After removal of the lowquality sequences and artificial duplicate reads, a total of 679,917,634 bp with average length of 276 bp was obtained. The rarefaction curve indicated that the number of sequences submitted was capable of detecting the existing diversity in the sample (not shown). Among the good-quality sequences, 1% contained ribosomal RNA genes, 90.68% encoded for proteins with known functions, and 8.14% proteins with unknown functions. Considering the automatic annotation in the MG-RAST v.4.0.4 server, the taxonomic classification of all shotgun sequences indicated that 99.23% belonged to the domain Bacteria, 0.2% to Eukaryota, 0.01% to Archaea, and 0.56% to Viruses (not shown). Among the 14 predominating genera identified in the sample, Acetobacter and Leuconostoc represented more than 50% of the sequences in the microbiome, whereas Azospirillum, the target microorganism in that sample, was not found (Fig. 2).

Susceptibility to antimicrobials
The test of susceptibility to antimicrobials was carried out according to [73][74][75] only with 36 isolates considered of clinical relevance. Considering the CLSI protocol, 12 isolates presented no resistance to at least one antibiotic; six presented intrinsic resistance to at least one antibiotic; and 18 isolates presented single or multiple resistance (Table 3). Noteworthy, some isolates showed multiresistance to antibiotics, e.g., isolates 1.5 and 2.4, which showed high 16S rRNA gene homology with Enterococcus faecalis, and showed resistance to all and to five tested antibiotics, respectively.

Discussion
Among 84 isolates, 25 genera were identified, 44% of which are known to harbor potential human pathogens, whereas only one isolate (5.2) showed 16S rRNA gene homology with the target microorganism A. brasilense. That was a case in which the sample was taken only 4 h after the tank had been inoculated with a commercial inoculant. Thus, the isolate probably originated from the commercial inoculant used as inoculum, not from the multiplication, since the short time between the addition of inoculum and the sampling may still have allowed the microorganism to survive. No other sample provided colonies identified as Azospirillum, showing that the target microorganism is eliminated or suppressed as the growth media become dominated by contaminating microorganisms. In addition, among the six samples aiming to multiply Bradyrhizobium, no isolate corresponded to the target bacteria.
Multiplication of microorganisms must assure several minimal microbiological procedures to guarantee that the target microorganism prevails in the culture medium. In the case of Azospirillum and mainly Bradyrhizobium, a slowgrowing bacterium [15], several other microbial contaminants dominate the culture medium as they have shorter    [70] generation times, i.e., higher growth rates than the target bacteria. In many cases, the carbon source in the culture medium used for on farm production is not appropriate. For example, the use of sucrose provided as molasses for growth of Bradyrhizobium is not appropriate, as the preferred carbon sources are glycerol or mannitol [15]. Besides competition with contaminating microorganisms, the physical-chemical characteristics in the culture medium are also inappropriate for growth of the target microorganisms. For example, the adequate range of pH for Bradyrhizobium and Azospirillum is between 6.8 and 7.0 [15,16,76]; however, 94.4% of the samples had pH ranging from 3.6 to 5.9. The low pH can also favor the growth of contaminating microorganisms adapted to low pH and thus contributing to suppress the target microorganisms. The lack of standardization in the incubation time is another problem in the samples taken from on farm production in this study. The average growth time of the recommended Bradyrhizobium strains to reach the ideal concentration (at least 1 × 10 9 cells mL −1 ) in the inoculant is approximately 7 days [76][77][78][79]. Similarly, A. brasilense has a growth time of about 5 days to reach at least 1 × 10 8 cells mL −1 [80]. In contrast, many contaminants have much shorter generation times, and dominate the culture medium * DNA fragment (bp) sequenced; Coverage: percentage of the sequence of interest aligned with a sequence deposited at GenBank; identity: maximum identity obtained with the highest alignment scores ** The isolates identified as "yeast" under microscope observation were not subjected to molecular identification *** The isolates were numbered using the sample numbering as received in the laboratory followed by the number of the isolated colony. For example, isolate 1.5 is the 5th isolate of the sample 1 Commercial inoculants: C1, Azospirillum brasilnse (strains Ab-V5 and Ab-V6); C2, Bradyrhizobium japonicum (SEMIA5079) and B. diazoefficiens (SEMIA5080); C3, B. elkanii (SEMIA587 and SEMIA5019), respectively Azospirillum, the target microorganism, was not detected in the sample Table 3 Antimicrobial resistance test applied to isolates with pathogenic potential to humans obtained from samples of inoculants produced on farm aiming the multiplication of Bradyrhizobium spp. and -GEN, LNZ, AMP, STR, VAN, PEN in less than 24 h. Contaminating microorganisms compete for resources in the growth medium that becomes nutritionally poor and can also release inhibiting byproducts [81]. Therefore, it is reasonable to conclude that the high multiplication rates of the contaminating microorganisms, in addition to the low growth rates of the target microorganisms, result in the rapid depletion of the culture medium and enrichment with metabolites that inhibit the development of slow-growing microorganisms, like Bradyrhizobium and Azospirillum.
Multiplication of microorganisms without strict quality control can be risky to humans, animals, crops, and environment. Many contaminants are potentially pathogenic to humans and may cause various diseases, posing risks to the health of individuals who handle these products, or even final consumers if applied to products consumed in natura. Although potentially pathogenic microorganisms are found in the environment, they usually do not cause risk due to the low potential of inoculum in the environment. However, the multiplication of this microbial population in contaminated culture media could also magnify risks of infections or contaminations. For example, microorganisms from genera like Enterococcus, for which similar sequences were found in 61.1% of the samples, are frequently related to bacteremia, septicemia, urinary tract infections, abscesses, meningitis, and endocarditis [32,[82][83][84]. Some isolates also presented high genetic similarity with Citrobacter freundii [85], Enterobacter cloacae [86,87], and Paenibacillus polymyxa [88], which are also potentially pathogenic to plants [86][87][88].
The possibility to carry genes of resistance to antimicrobials is a further concern in magnifying the population of potentially pathogenic contaminants in the on farm production. The spread of such genes in the environment may restrict the resources to fight infections. Some opportunist pathogens like Stenotrophomonas maltophilia are intrinsically resistant to several antimicrobials and collaborate to spread genes of resistance in the environment [70]. In this study, 12 isolates presented non-intrinsic resistance to antimicrobials, and 10 isolates presented resistance to two or more antimicrobials (1.1, 1.2, 1.3 Isolates identified microscopically as yeasts were not sequenced for genetic comparisons with sequences deposited in ribosomal databanks. However, some genera of yeasts can also cause injuries to humans and animals. Although yeasts are used in the manufacture of breads and beer, without any risk to humans and animals, like Saccharomyces cerevisiae, the genus Candida is the main pathogenic yeast and comprises approximately 200 species [89].
The approach based on metagenome for sample no. 10 showed that only contaminating microorganisms prevailed in the on farm sample. Although four morphologically distinct  colonies were isolated from that sample based on the culture medium approach, the metagenome approach revealed more than 10 genera, including the ones isolated based on the cultivation method. This indicates that the amount of contaminating microorganisms in the on-farm multiplications can be far higher than revealed by the culture-based method. In addition, even using a more sensitive method, the target microorganism was not found in that sample. Studies on inoculants produced on farm and their impacts on production systems and potential risks to public health are scarce. However, our findings corroborate previous studies on bioinsecticides produced on farm, which revealed low concentration or absence of the target microorganisms Bacillus thuringiensis [12], and absence of Chromobacterium subtsugae and Saccharopolyspora spinosa [13]. However, there was high prevalence of contaminants in the samples, some of them potentially pathogenic to humans [12,13].
The negative effect of low-quality bioproducts produced on farm goes beyond the risk to Brazilian quality of agricultural products, crops, and environment, because the benefits to the crops cannot be reached with its use. The lack of effect for not containing the target microorganism might put in doubt consolidated technologies that are important to the sustainability of cropping systems like the BNF in soybean by inoculation with Bradyrhizobium [3,4], and more recently inoculation of grasses and co-inoculation of soybean with Azospirillum [7,8,11].
In conclusion, the samples of inoculants produced on farm assessed in this study were highly contaminated with several non-target microorganisms, whereas the target microorganisms Azospirillum and Bradyrhizobium were not detected in the great majority of the samples. In addition, the occurrence of contaminants presenting high genetic similarity with potentially pathogenic microorganisms, some of them carrying non-intrinsic resistance or multiresistance to antimicrobials, may indicate risk to human health.