NASA has maintained cleanrooms at the Johnson Space Center (JSC) for curating extraterrestrial samples from the moon, meteorites, cosmic dust, asteroids, comets, solar wind particles, and micrometeorite impacts on space-exposed hardware starting with the lunar samples from the Apollo missions in 1969 (McCubbin et al., 2019). Oligotrophic, low humidity conditions, regular cleaning and air filtration render these facilities inhospitable to microbial life. Despite these fastidious controls, the cleanrooms contain bacteria and fungi (Regberg et al., 2018), which could alter the composition of astromaterials and confound searches for extraterrestrial life (Rummel, 2001). The presence of bacteria and fungi in the curation cleanrooms is acceptable because current astromaterials collections do not have biological contamination control requirements; however, future sample collections, such as the Mars sample return missions, will have these requirements (Carrier et al., 2021). To prepare for these missions, NASA has developed a routine microbial monitoring program for existing collections (McCubbin et al., 2019). Bacillus species comprise, on average, 45% of the microbes cultured in this curation program at the JSC (Mazhari, 2021) and are frequently isolated from cleanrooms at the NASA Jet Propulsion Laboratory (JPL) (Tirumalai et al., 2013; Tirumalai et al., 2018).

Strain-level identification of microbes recovered from cleanrooms is important for developing a robust microbial source tracking program and overall contamination control strategy (Song et al., 2024); however, 16S rRNA sequencing, which is widely used to identify bacterial isolates (Church et al., 2020), lacks the resolution required to differentiate closely related Bacillus species. For example, 16S rRNA gene sequences of many Bacillus species, that occupy fundamentally different environmental niches, are over 99% identical (Yamada et al., 1999). This limits the utility of 16S rRNA gene sequencing for monitoring diversity within curation cleanrooms (Espariz et al., 2016). Whole genome sequencing (WGS) has emerged as the definitive method for microbial identification (Kwong et al., 2015) and is widely used for tracking food-borne pathogens and disease outbreaks (Brown et al., 2019). However, building a library for microbial identification by WGS costs at least $400 per isolate (Brown et al., 2021) and requires highly trained personnel to generate and interpret the data.

Matrix-assisted laser desorption – time of flight mass spectrometry (MALDI-TOF MS) systems provide strain-level identification of microbes, for less than a dollar an isolate, in seconds (Gagné-Bourque et al., 2015). For example, MALDI-TOF can differentiate strains with different functional properties that produce and are resistant to antibiotics (Flores-Treviño et al., 2019) and differentiate between pathogenic and virulent strains of Bacillus species (Celandroni et al., 2016). MALDI-TOF MS systems use pattern matching between mass spectra generated from isolates and mass spectra generated from reference strains (Singhal et al., 2015). This high-throughput method is cheaper and more accurate than conventional biochemical systems for bacterial identification (Lévesque et al., 2015; Seng et al., 2009) and for identifying Bacillus species (Lasch et al., 2009) and related genera (Celandroni et al., 2016). MALDI-TOF MS appears comparable to WGS for identification of the specific pathogenic bacteria (da Silva et al., 2020; Nithimongkolchai et al., 2023; Rudolph et al., 2019; Werinder et al., 2021); however, these systems have limitations (Haider et al., 2023; Rychert, 2019). For example, MALDI-TOF MS systems can struggle to differentiate some species related to Bacillus cereus (Muigg et al., 2022) and lack of suitable reference spectra has limited the application of MALDI-TOF MS outside of the field of clinical microbiology (Rahi et al., 2016). Further, variation between strains of bacteria, can influence the performance of MALDI-TOF systems, so there is a need to evaluate this approach for bacteria isolated from different environments (Emami et al., 2016; Topić Popović et al., 2023). This need is particularly pressing for extreme environments (Kopcakova et al., 2014), including cleanrooms (da Costa et al., 2022) and similar facilities (Seuylemezian et al., 2018).

In this study, we compared the resolution of MALDI-TOF MS to WGS for a set of bacteria isolated from cleanrooms at the Johnson Space Center. Draft genomes of 14 bacteria were assembled with a hybrid of Illumina and Oxford Nanopore sequencing and genomic relationships were characterized using an estimated maximum-likelihood phylogenomic tree. The resolution of WGS, in terms of average amino acid identity, was then compared to MALDI-TOF MS with custom scripts following LaMontagne (LaMontagne et al., 2021). MALDI-TOF MS showed species-level resolution, which is comparable to WGS.

The optimization loop showed a modal relationship between Jaccard coefficients and the number of peaks detected. The highest Jaccard coefficients were observed at 163 peaks (Supplementary file S8).

From 15 isolates, 14 draft genomes were de novo assembled with greater than 98% mean completion (Table 1). GC contents for these isolates ranged from 34% to 49%, which is consistent with the nucleotide compositions typical for these genera (Lightfield et al., 2011) and mesophilic bacteria in general (Hu et al., 2022). Genome sizes ranged from 7.1 to 3.7 Mbp, which is typical for the phylum Bacillota (Martinez-Gutierrez and Aylward, 2022). Isolate 1781tsa1, which was classified as a Paenibacillus species (Figure 1), showed the largest genome size; isolate 1370ba1, which was classified as a strain of Bacillus safensis (Figure 1), showed the smallest genome size.

Bacterial identification by WGS generally agreed with identifications by the MALDI system at the genus level (Table 2); however, two isolates were classified as different genera. Isolate 1480ba3 was identified as Priestia flexus and isolate 2069sda1 was identified as Peribacillus frigoritolerans by WGS. These two isolates were both classified as Bacillus species by the MALDI system (Table 2). This reflects the recent assignment of Bacillus flexus to the novel genera Priestia (Gupta et al., 2020) and the proposal of Peribacillus as a novel genus (Patel and Gupta, 2020). Apparently, these changes in nomenclature have not been updated in the proprietary Bruker MALDI Biotyper^® database.

The MALDI system and WGS identified 13/15 and 9/14 at the species level, respectively, (Table 2). For 9 isolates identified at the species level by WGS, only 4 identifications were shared with identifications by MALDI-TOF MS. For example, isolate 2987tsa1 was identified as B. amyloliquefaciens and B. velezensis and isolate 2069sda1 was identified as B. simplex and Peribacillus frigoritolerans by MALDI-TOF MS and WGS and respectively (Table 2). In contrast, 16S rRNA gene taxonomy did not identify any of the isolates at the species level (Mazhari, 2021).

The topology of a phylogenomic tree generated from multiple single-copy core genes appeared largely coherent in terms of isolate identifications (Figure 1). That is, isolates generally formed monophyletic clades with representatives of their genera and the topology of this tree received strong bootstrap support (0.915–1.00). In contrast, a tree generated from 16S rRNA gene sequences (Supplementary file S9) showed several clades with relatively weak bootstrap support (0.24–0.41). Also, clades containing reference genomes for Bacillus wiedmannii were polyphylogenetic. This species clustered with two different Bacillus clades. Isolates did not appear to cluster by sample location.

Several clades in the phylogenomic tree showed that bacteria isolated from cleanrooms at JSC appeared closely related to bacteria isolated from cleanrooms at JPL, or similar built environments, like the International Space Station (ISS). For example, isolates identified as Paenibacillus species (2941sda1 and 1781tsa1), isolated from the floor of the Lunar curation lab at JSC, clustered with two strains Paenibacillus xylanexedens including a strain (PL-73) isolated from an air sample collected in a cleanroom at the JPL (GCA_019749275). Similarly, isolate 2090tsa1, which was isolated from a table in the Hayabusa lab at JSC, showed similarity to a strain (I1-R4) of Bacillus cereus isolated from an air sample collected at JPL. Further, Bacillus species (2987tsa1, 1943r2a1, 1735sda2, and 1370ba1) isolated from swabs of the cleanroom floors from the JSC Cosmic Dust, Lunar, Hayabusa, and Meteorite labs respectively, clustered with seven Bacillus strains isolated from spacecraft assembly cleanrooms at JPL and the ISS with strong bootstrap support (> 0.95, Figure 1).

The topology of a dendogram generated by mass spectra appeared coherent for isolates that were reliably identified at the species level by the MALDI system (Figure 2). In other words, representatives of the same species clustered together with strong bootstrap support (Figure 2). These clusters also appeared in the phylogenomic tree generated by WGS (Figure 1). In particular, edge 3 contained four isolates that classified as Bacillus cereus (Figure 2, Table 2). This clade corresponded to a clade, generated by WGS, that contained B. cereus and related species (Figure 1). Similarly, edge 5 contained two isolates that were classified as Bacillus flexus (Figure 2, Table 2). This clade corresponded to a clade, generated by WGS, that contained Priestia flexa and related species (Figure 1). A clade (edge 4) that contained two isolates that were classified as Paenibacillus species also received strong bootstrap support. Clades containing singletons generally appeared unreliable; however, edge 6 contained two Bacillus species that appeared closely related to each other in terms of mass spectra (Figure 2) and WGS (Figure 1).

Partitioning of isolates into MTUs, by MALDI-TOF, generally agreed with partitions defined by WGS. The adjusted Rand coefficient, for congruence of microbial typing method, was 0.95 and the adjusted Wallace coefficient was 0.90, with a 95% confidence interval of 0.80 to 1.00.

Pairs of isolates that were classified as members of the same species by the MALDI system, showed high similarity to each other in terms of WGS (Figure 3). This supports the hypothesis that MALDI-TOF MS and WGS are comparable. Specifically, pairwise similarity of the amino acid sequences from single-copy core genes, predicted from the draft genomes, ranged from 93% to 98% AAI for pairs of isolates from the same species. These pairs showed cosine similarities, generated from mass spectra, that ranged from 0.83 to 0.90 with an average of 0.88 (Figure 3). Pairwise comparisons of mass spectra between members of the same species were significantly (p = 0.0001) larger than pairwise comparisons for members of different species (Figure 3, Supplementary file S7).

Pairwise similarities of mass spectra at the genus level were on average lower (0.42); however, the pair of Paenibacillus species showed a high similarity to each other in terms of cosine similarity of mass spectra (0.85) but a relatively low AAI (Figure 3). Similarly, three Bacillus isolates (1370sda2, 1735sda2, and 1943r2a1), which were classified as different species by both WGS sequencing and mass spectra (Table 2), showed relatively high AAI similarities (93–95%). These three isolates showed cosine similarities of 0.68 to 0.74, which were lower than expected given trends in Figure 3.

Pairwise comparisons of isolates that were assigned to different families (Paenibacillaceae and Bacillaceae) of the order Bacillales, and different genera (Lysinibacillus and Bacillus) of the family Bacillaceae, were resolved by AAIs but not by cosine similarities (Figure 3). These comparisons at the order and family level averaged 57 (56–57)% and 61 (60–62)%, respectively, for AAIs and 0.40 (0.33–0.47), 0.41 (0.35–0.46) respectively for cosine similarities (Supplementary file S7).

At a cosine similarity of 0.8, which appeared to resolve genera and species of Bacillus (Figure 3), MALDI-TOF MS detected 10 MALDI-TOF taxonomic units (MTUs) in a library of 15 isolates. This library contained three MTUs with more than one isolate and seven singletons. This corresponds to a coverage of 59%; extrapolation to a library of 28 isolates would give 80% coverage (Supplementary file S10).

MALDI-TOF showed a resolution comparable to that of WGS. Both MALDI-TOF MS and WGS identified Bacillus strains consistently at the genus level (Table 2) and the relationship between isolates as assessed by phylogenomic analysis (Figure 1) agreed with clustering generated by mass spectra (Figure 2). Comparison of the similarity between isolates as assessed by WGS, suggested a threshold for identifying species of Bacillus isolates by MALDI-TOF MS. Pairs of strains that showed cosine similarities of mass spectra of >0.80, were reliably identified at the species level by the MALDI system and showed AAI, as assessed by WGS, of >92% (Figure 3). The high cosine similarities for comparisons within Bacillus species relative to comparison between Bacillus species (Figure 3) confirms that MALDI-TOF MS provides an accurate representation of species diversity and can differentiate between related Bacillus species (Celandroni et al., 2016; Hotta et al., 2011; Lasch et al., 2009). This suggests MALDI-TOF provides a quick, cost-effective, and accurate method of identifying microbes that contaminate astromaterials curation facilities.

This threshold for defining a species by MALDI-TOF MS agrees with the cosine similarity threshold of 0.79 set by Strejcek et al. (2018). This proposed threshold of cosine similarity threshold (0.80) could result in the misclassification of Paenibacillus strains, which appear to be different species as assessed by AAI (Figure 3). AAI levels of >92%, observed for within-species comparisons is lower than the widely used threshold for defining a species for prokaryotes by AAI of 95% (Konstantinidis and Tiedje, 2005); however, this higher threshold for AAI levels may be biased towards clinical isolates (Rosselló-Mora, 2005).

Phylogenomic analysis showed several isolates clustered with Bacillus spp. isolated from other studies of cleanrooms and the ISS (Figure 1). This suggests there is a cosmopolitan class of Bacillus strains associated with cleanrooms and similar built environments, including the ISS (Blaustein et al., 2019). Microbes in the ISS and cleanrooms face similar selective pressures, including low nutrient availability and humidity. These “extreme environments” (Checinska Sielaff et al., 2019; La Duc et al., 2003) may share many microbes; however, this biogeographical pattern was not consistent. For example, the isolate identified as Lysinbacillus endophyticus (2047tsa1) clustered with a Lysinbacillus species isolated from the ant microbiome and the isolate that classified as Bacillus infantis (2933tsa1) clustered with a Bacillus species isolated from a marine system. Clearly, these species are not exclusively observed in cleanrooms; however, their detection can be explained. Lysinibacillus species have been isolated from air samples (Wong et al., 2019) and novel species, that are closely related to Bacillus infantis, have been isolated from cleanrooms (Seiler et al., 2013).

For routine monitoring, this proteomics approach could replace Vitek2 and 16S rRNA sequencing, which are widely used in environmental monitoring programs. The resolution of MALDI-TOF MS could also help investigators prioritize isolates for deeper analysis by WGS and biochemical tests. For example, genetic and functional characteristics of MTUs could be used to identify novel species with unique metabolic capabilities. The high resolution of MALDI-TOF also allows for tracking the source of microbial contamination, as has been applied to aquatic systems (Giebel et al., 2008; Siegrist et al., 2007).

MALDI-TOF MS can efficiently identify species of the genus Bacillus that are frequently isolated from facilities where astromaterials are curated, and similar built environments, with a resolution comparable to WGS. Implementation of this proteomics approach would require development of database of mass spectra.