Clinically relevant pathogens on surfaces display differences in survival and transcriptomic response in relation to probiotic and traditional cleaning strategies

Indoor surfaces are paradoxically presumed to be both colonized by pathogens, necessitating disinfection, and “microbial wastelands.” In these resource-poor, dry environments, competition and decay are thought to be important drivers of microbial community composition. However, the relative contributions of these two processes have not been specifically evaluated. To bridge this knowledge gap, we used microcosms to evaluate whether interspecies interactions occur on surfaces. We combined transcriptomics and traditional microbiology techniques to investigate whether competition occurred between two clinically important pathogens, Acinetobacter baumannii and Klebsiella pneumoniae, and a probiotic cleaner containing a consortium of Bacillus species. Probiotic cleaning seeks to take advantage of ecological principles such as competitive exclusion, thus using benign microorganisms to inhibit viable pathogens, but there is limited evidence that competitive exclusion in fact occurs in environments of interest (i.e., indoor surfaces). Our results indicate that competition in this setting has a negligible impact on community composition but may influence the functions expressed by active organisms. Although Bacillus spp. remained viable on surfaces for an extended period of time after application, viable colony forming units (CFUs) of A. baumannii recovered following exposure to a chemical-based detergent with and without Bacillus spp. showed no statistical difference. Similarly, for K. pneumoniae, there were small statistical differences in CFUs between cleaning scenarios with or without Bacillus spp. in the chemical-based detergent. The transcriptome of A. baumannii with and without Bacillus spp. exposure shared a high degree of similarity in overall gene expression, but the transcriptome of K. pneumoniae differed in overall gene expression, including reduced response in genes related to antimicrobial resistance. Together, these results highlight the need to fully understand the underlying biological and ecological mechanisms for community assembly and function on indoor surfaces, as well as having practical implications for cleaning and disinfection strategies for infection prevention.


INTRODUCTION
Indoor surfaces harbor pathogens and mediate the transmission of infectious disease. At the same time, these environments have been called a "microbial wasteland" where microbes are "dead, dying, or dormant." Following this model, surface-associated microbial communities are almost entirely driven by the opposing forces of dispersal and death. Surface cleaning and disinfection practices seek to promote microbial cell death, thereby preventing infection, especially in healthcare settings. While cleaning using antimicrobials can effectively reduce microbial load on contaminated surfaces in health care facilities 1 , evidence of their effects on preventing disease transmission and reducing healthcareacquired infections remain limited 2 . Moreover, because antimicrobials can cause both acute and chronic health impacts 3 and lead to antimicrobial resistance [4][5][6] , there is a strong drive to develop alternative cleaning strategies.
Probiotic cleaning seeks to take advantage of competitive exclusion and other ecological principles to inhibit viable pathogens on surfaces 4,7 . Members of Bacillus, Bifidobacterium, Lactobacillus, Rhodopseudomonas, and Saccharomyces are the most frequently used microorganisms in these microbial based cleaning products 8 . While probiotic organisms may persist on surfaces, competitive exclusion is not compatible with the "wasteland" model, which stipulates no growth or metabolic activity. Previous studies of probiotic cleaners have shown that probiotic cleaning decreases the recovery of cultivable bacterial pathogens 7,9 , and even inactivates viruses 10 . Competitive exclusion is an unlikely mechanism of action against viruses, but even for bacteria there is limited direct evidence documenting competitive exclusion 11 . Rather, culture-based results support the hypothesis that probiotic Bacillus outcompete pathogens in some environmental conditions but are in fact out-competed by resident organisms when the latter are present in biofilms 11 . Moreover, there is a dearth of molecular evidence supporting whether surviving probiotic organisms and healthcare-associated pathogens are engaged in activities such as nutrient scavenging, stress response, or microbe-microbe interactions 11 . To this end, studying gene expression levels instead of cultivable colonies or gene presence in both pathogens and probiotic microorganisms is necessary to unravel the mechanisms of probiotic cleaning. Transcriptomics allows quantitative determination of changes in regulation of genes and pathways of the microorganisms upon application of probiotic cleaners under different conditions, revealing changes in function and metabolic activity. Despite these advantages, the application of transcriptomics in indoor surface microbiome studies has been sparse 12 . Current indoor surface microbiome studies only focus on taxonomic profiling and functional potentials that can be recovered from (meta)genomics analysis; the importance of gene expression levels is overlooked or yet to be addressed 13,14 . Transcriptomics analysis from this study can provide insights on the response of pathogens to tested cleaning scenarios on indoor surfaces.
Acinetobacter baumannii and Klebsiella pneumoniae have been recognized as a major source of healthcare-associated infections associated with high morbidity and mortality rates due to their resistance to many antibiotics including carbapenems 15,16 . These organisms not only persist on surfaces for up to several months 17 but are commonly found in healthcare environments, especially in low-and middle-income countries. K. pneumoniae have been found on surfaces in hospitals in Brazil 15,16 , Tunisia 17 , Russia 18 , China 19 and Ghana 20 . A. baumannii was found on surfaces in hospitals in Brazil 21 , Turkey 21 , and Saudi Arabia 22 and on the surfaces of medical devices in Iran 23 . It was also found on surfaces in buildings surrounding hospitals in the US 24 . Both A. baumannii and K. pneumoniae were found on surfaces in neonatal intensive care units in Egypt 25 and hospitals in Algeria 26 . While these organisms likely reach surfaces via dispersion from human occupants, they can also be transferred back to humans, particularly the very vulnerable. K. pneumoniae persists in neonatal intensive care units and transfers to neonates' gut 27,28 and possibly also the nares 29 . A high degree of relatedness has been observed between genotypes of A. baumannii from patients and their surrounding rooms 30 , and bidirectional transfer of A. baumannii between the environment and adult patients has also been observed 31 . Notably in this last study, A. baumannii then persisted in the environment despite terminal disinfection 31 . More concerning than individual transfer events or colonization, surface-associated organisms can cause or exacerbate outbreaks within a hospital. For example, surface-associated K. pneumoniae was associated with an outbreak in Vietnam 32 , and A. baumannii on bed railings was associated with an outbreak in Saudi Arabia 33 .
To understand how these organisms survive on surfaces, sometimes in defiance of cleaning efforts, we sought to observe changes in gene expression in A. baumannii and K. pneumoniae on surfaces with and without cleaning. Sequencing of low biomass samples collected from the built environment often suffers from a low signal-to-noise ratio 34 . To obtain a fundamental mechanistic understanding of indoor cleaning strategies and their impacts on microbial survival and interactions, we utilized microcosm chambers containing stainless steel sheets to simulate a simplified built environment with greater than usual microbial load. Because antimicrobial resistance and hospital-acquired infections disproportionately affect low-and middle-income countries, many of which are located in tropical climates, we also evaluated the effects of elevated temperature and relative humidity (RH). To evaluate general ecological phenomena, and particularly competitive exclusion, governing survival in mixed communities, we further exposed these organisms to a probiotic cleaner containing several strains of Bacillus.

RESULTS AND DISCUSSION
Survival of A. baumannii and K. pneumoniae under four cleaning scenarios and two temperature/humidity conditions We quantified the CFUs of A. baumannii and K. pneumoniae to examine their survival under ambient and humid conditions in combination with four cleaning strategies (Fig. 1, Table 1). Over the course of 3 days, we observed a maximum log 10 Fig. S1), Probiotic Cleaner cleaning at 24 h showed 1.43 greater log 10 reduction compared to Cleaner Only (Fig. 2a). As there was no significant difference between Probiotic Cleaner and Cleaner Only, the reduction of A. baumannii was largely attributed to the chemical component of the cleaning product. Differences in the survival of K. pneumoniae between Probiotic Cleaner, Cleaner Only and No with four types of inocula containing pathogen (A. baumannii or K. pneumoniae) and Bacillus spp. and sealed sterile microcosm chamber maintained under ambient temperature and relative humidity (c). The elevated temperature and humidity condition was maintained by adding 10 mL of sterilized miliQ water into the sealed chamber.
Cleaning scenarios were not statistically significant (p > 0.05, Fig.  S1), with only 0.77 greater log 10 reduction in Probiotic Cleaner compared to Cleaner Only at 24 h. The fact that K. pneumoniae was unable to persist on the surface in No Cleaner group under ambient conditions indicated that chemical components in the cleaning product only had marginal contribution to K. pneumoniae inactivation.
Surfaces cleaned with Probiotic Cleaner were also gradually dominated by Bacillus spp. due to the decreasing absolute abundance of the pathogen population. This persistence of Bacillus spp. on surfaces was observed under both ambient and elevated humidity and temperature (Fig. 2). In contrast, the germinated Bacillus spp. spores introduced as vegetative cells were gradually taken over by the A. baumannii population; for K.  pneumoniae Germinated Probiotic group the cleaned surface was still dominated by Bacillus spp. over K. pneumoniae (Fig. 2a). Without the use of chemical-based surface cleaner (No Cleaner and Germinated Bacillus), A. baumannii showed comparable CFUs at all time points (p > 0.05).
Throughout the experiments, the minimum log 10 reduction was observed under elevated temperature and humidity for both A. baumannii and K. pneumoniae. Their survival over time improved under elevated temperature and humidity with or without Probiotic Cleaner (Fig. 2), although A. baumannii exhibited higher viable populations (p < 0.01, Fig. S1A) while K. pneumoniae showed low or no significance (Fig. S1B) when comparing No Cleaner (Humid) to No Cleaner (Ambient) (paired t test was performed for all time points). Higher viability of bacteria is observed under higher RH, which is related to the stress caused by concentration of solutes and desiccation as droplets dry out 35 . Genes csrA, which is related to carbon storage, and bfmR (rstA), which is related to stress response, play a major role in desiccation tolerance in A. baumannii 36,37 . Both genes were downregulated under the humid condition in A. baumannii No Cleaner groups compared to the ambient condition, and csrA gene was also downregulated under the humid condition in K. pneumoniae Probiotic Cleaner group compared to the ambient condition (log 2 fold change value < −1, adjusted p value < 0.05). This result suggests greater potential health risks and infectious disease burdens in developing regions with a humid and hot climate and limited climate control 38 .
Overview of whole genome sequencing and metatranscriptomic data Sequencing of the A. baumannii isolate generated 3.72 million reads with an average read length of 136 bp. Reads were assembled into contigs with a N50 of 141,238 bp. Sequencing of K. pneumoniae generated 4.4 million reads with an average length of 145 bp. Reads were assembled into contigs with a N50 of 113,698 bp. Sequencing of the All-Purpose Probiotic Cleaner resulted in 16.49 million reads with an average read length of 139 bp, after quality trimming and contaminant removal. Reads were assembled into contigs with a N50 of 143,313 bp. Open reading frame (ORF) identification recognized 4039, 5823, and 22,429 coding regions for A. baumannii, K. pneumoniae, and Probiotic Cleaner, respectively, 2253, 4052, and 13,339 of which were functionally annotated (not as "hypothetical protein"). Among these annotated regions for A. baumannii, K. pneumoniae, and Probiotic Cleaner assemblies, 2150, 3675, and 11,594 were classified into one or more KEGG Orthologs respectively. Contigs from Probiotic Cleaner reads were binned into five groups with completeness > 90%, and contamination <5%. All five bins were classified as Bacillus species as shown in Table 2.
RNA samples containing A. baumannii or K. pneumoniae were collected for up to 72 h of surface inoculation. Biological triplicates were collected for each condition and an average of 12.73 ± 3.83 and 11.03 ± 4.98 million reads with an average read length of 49 bp were recovered for A. baumannii and K. pneumoniae experiments, respectively, after removing low-quality reads, contaminated sequences, adapters, and rRNA. The average mapping rate using HISAT2 was 98.20% ± 1.06% for experiments with A. baumannii and 96.40% ± 10.24% for experiments with K. pneumoniae (Fig. S2). Percentages of reads mapped onto A. baumannii and Bacillus genomes vary across four cleaning scenarios; Germinated Bacillus samples containing vegetative Bacillus spp. on average had 40.53% ± 6.07% reads mapped onto Bacillus reference genomes, whereas this number for Probiotic Cleaner samples was only 0.56% ± 0.75%. A similar trend was observed in experiments with K. pneumoniae. Samples containing K. pneumoniae and vegetative Bacillus spp. on average had 40.52% ± 16.83% reads mapped onto Bacillus reference genomes; however, this number was only 1.00 ± 1.08% for Probiotic Cleaner samples. Thus, these data suggest that the majority of Bacillus spp. in the probiotic product likely remained as spores throughout the incubation. Due to this insufficient coverage of Bacillus transcriptome in Probiotic Cleaner samples, our transcriptomic analysis is composed of quantitative comparisons for the A. baumannii and K. pneumoniae transcriptomes and qualitative evaluations for Bacillus spp.
Transcriptomes of A. baumannii and K. pneumoniae clustered into distinct groups determined by cleaning and physical conditions Prior to conducting principal component analysis (PCA), read count data were transformed to remove the dependence of the variance on the mean. PCA of the transformed (variance stabilization) read counts belonging to A. baumannii indicated that the expression of A. baumannii clustered into three distinct groups based on cleaning scenarios and physical conditions (Fig. 3a). Clustering of read counts from K. pneumoniae is more distinct for different time points and physical conditions ( Fig. S2) compared to cleaning scenarios and physical conditions (Fig. 3c). PERMANOVA (9999 permutations) revealed that temperature and humidity (p < 0.001), cleaning (p < 0.001), and time (p < 0.001) explained 12%, 30%, and 9% of the variance for A. baumannii read counts, and 29%, 10%, and 12% for K. pneumoniae read counts, respectively. Gene expression profiles of A. baumannii with detergent cleaning (Probiotic Cleaner and Cleaner Only) are more similar to each other than they were to those without detergent (Germinated Bacillus). Temperature and humidity conditions also affected the expression of A. baumannii under the no-cleaning condition, but this impact was minimal when Probiotic Cleaner was used. On the contrary, the temperature and humidity strongly affected the gene expression profiles of K. pneumoniae, and the impact was more pronounced as the cleaning time extended from 3 h to 24 and 72 h.
Using No Cleaner samples as the reference transcriptome, genes with absolute log 2 fold change values greater than or equal to 1 as calculated by DESeq2 were considered as differentially expressed genes (DEGs, Benjamini-Hochberg adjusted p value ≤ 0.05). 414, 522, and 634 genes of A. baumannii and 578, 145, and 57 genes of K. pneumoniae were found to be differentially expressed in Cleaner Only, Probiotic Cleaner, and Germinated  Fig. 3d). Since the transcriptome of K. pneumoniae under Probiotic Cleaner scenario was fairly similar to the No Cleaner scenario, their similar response to elevated temperature and humidity could be expected. Samples collected at various time points throughout the microcosm inoculation also showed that much of the temporal variation in gene expression occurred within 24 h of surface cleaning for A. baumannii (Fig. 3b), K. pneumoniae (Fig. 3d), and vegetative Bacillus (Fig. S4) baumannii with Germinated Probiotic). These phenomena support the hypothesis that the A. baumannii genome encodes survival mechanisms for persistence in replete environments, while K. pneumoniae, although capable of short-term response, may be less equipped for long term survival on nutrient poor surfaces. The A. baumannii transcriptome shifted under treatment conditions and 26 KEGG pathways broadly related to metabolism, energy production, translation, and membrane transport were enriched. Significant enrichment of translation machinery under Germinated Probiotic treatment, suggests that protein synthesis in A. baumannii was likely promoted to achieve a large demand for surface competitions. On the other hand, three pathways were positively enriched in samples with chemical-based detergent cleaning, indicating that the A. baumannii population overall repressed metabolic activities and adaptively reduced synthesis in response to environmental constraints associated with detergent cleaning (Fig. 4a). As A. baumannii is notoriously associated with the hospital environment 40 , it stands to reason that it may remain constitutively competent to respond to such replete environments and have acquired additional adaptive traits to quickly acclimate when additional stressors, e.g., chemical cleaners, common in the hospital environment arise. Normalized enrichment score (NES) accounts for differences in gene set size and were used to compare enrichment results across KEGG pathways; a positive NES suggests an overall positive upregulation of genes belonging to its corresponding pathway [70][71][72] . *p < 0.05, **p < 0.01, and ***p < 0.001.
In contrast, K. pneumoniae responded with far greater heterogeneity at the transcript level. Carbon metabolism genes in K. pneumoniae were mutually upregulated in every cleaning condition; however, specific carbon metabolism pathways were induced in a treatment-dependent manner. For instance, TCA cycle, pyruvate, starch and sucrose metabolism genes were upregulated by treatment with Germinated Bacillus or Cleaner Only but not in samples treated with Probiotic Cleaner. This suggests that vegetative Bacillus and cleaning products induce alternative though overlapping carbon metabolism networks in K. pneumoniae. Indeed, increased expression of biosynthetic pathways (glyoxylate cycle, carbon fixation) suggests vegetative Bacillus may promote anabolic competency in K. pneumoniae. Perhaps this broad range of metabolic activity is the result of slight alterations in the K. pneumoniae panic response, as the population tries to survive in a "hostile" environment, unlike that of its native host.
While fewer pathways (11) were uniquely enriched in response to Germinated Bacillus, every treatment stimulated some unique pathway upregulation in K. pneumoniae, unlike the uniform pattern observed in A. baumannii (Fig. 4d). Genes involved in two-component systems and membrane transport (ABC) were upregulated in samples treated with Probiotic Cleaner or Cleaner Only but not in samples containing Germinated Bacillus, indicating that chemical detergents, rather than Bacillus, induce environmental sensing and response mechanisms in K. pneumoniae. Surprisingly, no pathways were mutually and specifically enriched by Germinated Bacillus or Probiotic Cleaner, which may be influenced by the diminished transcriptional landscape of Bacillus spores (Fig. S2). Alternatively, the divergent transcriptomic profiles induced by each cleaning condition indicates that K. pneumoniae's strategy for surviving on surface environments vastly differs from A. baumanii and further suggests that the A. baumanii response system to built environment surfaces may be more robust than K. pneumoniae's. Indeed, evolutionary studies modeling extended resource exhaustion in Escherichia coli have shown that while stress responses dominate in the short-term, in the long-term, genetically-encoded adaptations to replete environments sweep the population via clonal expansion 41,42 . One could thus hypothesize the dissimilar fitness levels and phenotypic responses (e.g., enrichment patterns) of K. pneumoniae 43 and A. baumanii in this study may be the results of their unique evolutionary history and ecological niches.
Cleaning products may represent a source of nutrients in an otherwise replete environment. Both A. baumannii and K. pneumoniae exhibited increased transcription of genes related to metabolism, although not the same pathways and not under all conditions. Our results suggest that vegetative Bacillus survive poorly on surfaces and become a source of nutrients to more resilient A. baumannii. K. pneumoniae enhances transcription of various metabolic pathways, but none of the responses enhance its survival. The specific metabolic activity of organisms on surfaces is highly dependent on both the organism and the resource availability.
Different cleaning scenarios induced differential expression of genes associated with stress response and competition We examined gene expression of important interspecies competition pathways to understand the potential interactions between probiotic bacteria Bacillus spp. and pathogens A. baumannii or K. pneumoniae. The two pathogens responded differently toward cleaning scenarios where A. baumannii exhibited more pronounced differential gene expression compared to K. pneumoniae.
Iron acquisition through siderophores is believed to be one of the key mechanisms of inter-species competitive exclusion 44 . Differential expression of the siderophore and iron acquisition genes indicated that siderophore cheating is unlikely to happen under our studied conditions. In A. baumannii experiments, several genes responsible for siderophore synthesis, export, and reception (e.g., aerobactin, anguibactin, and enterobactin) were downregulated in the Germinated Probiotic group (Fig. S5A). Although bacteria with multiple siderophore receptors can gain competitive advantages in social competition through siderophore cheating 45 , the downregulation of siderophore-related genes in these samples was likely due to utilization of iron originating from lysed cell debris, especially from Bacillus spp., rather than active competition through siderophore cheating, given the inability of vegetative Bacillus spp. to persist on surfaces (Fig. 2a). For K. pneumoniae experiments, no siderophore-related gene was differentially expressed under the three cleaning scenarios (Fig.  S6A, iutA, fepA, and fepC log 2 fold change absolute values <1). Gene expression levels of Bacillus spp. (Figure S7) further precluded siderophore cheating as the competing mechanism on surface as genes related to the synthesis, and export of Acinetobacter-and Klebsiella-associated siderophores (e.g., aerobactin, anguibactin, and enterobactin) were only occasionally expressed (Fig. S7).
In addition to iron acquisition, genes related to Type VI Secretion Systems (T6SS), Type IV or Type 1 Pili, and biofilm formation were shown to have important implications in interspecies competition, virulence expression, and starvation response in A. baumannii or K. pneumoniae 44,46,47 . In A. baumannii experiments, genes associated with T6SS, Type IV Pili, poly-beta-1,6-N-acetyl-D-glucosamine (PGA) production and biofim formation were upregulated in Probiotic Cleaner, Cleaner Only, and Germinated Probiotic samples with varying statistical significance (Fig. S5). The pgaABCD operons, which had increased expression in Germinated Probiotic group, play roles in surface binding and maintaining biofilm structure stability 48,49 . This result suggested greater potential for PGA, a biofilm adhesin polysaccharide 48,49 , production and subsequent biofilm activities when A. baumannii and Bacillus spp. co-existed. In contrast, K. pneumoniae exhibited negligible differences in expression levels for genes related to T6SS and Type 1 Pili had under different cleaning scenarios for (Fig. S6BD). Additionally, only a few biofilm formation related genes were differentially expressed (Fig. S7C). Lack of shared differentially expressed genes across cleaning scenarios indicated nonspecific response of K. pneumoniae biofilm formation towards different stressors.
Similar to what was observed for KEGG pathways, antibiotic resistance gene expression of Probiotic Cleaner and Cleaner Only treated A. baumannii exhibited high similarity (Fig. 4b) while K. pneumoniae responded very differently to Probiotic Cleaner and Cleaner Only treatments (Fig. 4e) in. In the Germinated Probiotic groups, we observed an increased expression of abaF, a gene encoding fosfomycin resistance, in A. baumannii while no antibiotic resistance gene was upregulated (log 2 fold change ≥ 1) in K. pneumoniae. Interestingly, no genetic determinants of fosfomycin biosynthesis were identified from Bacillus spp. reference genomes. In addition, it was reported that the disruption of abaF in A. baumannii not only resulted in an increase in fosfomycin susceptibility but also a decrease in biofilm formation and virulence 50 . Therefore, the upregulation of abaF in A. baumannii may not be induced by the fosfomycin secretion from Bacillus spp. but rather a stress response to Bacillus spp. existence exerting enhanced biofilm formation.
In A. baumannii experiments, pronounced downregulation of genes related to sulfur metabolism, particularly those induced by sulfur starvation, was observed for groups where chemical cleaning detergent was present (Probiotic Cleaner and Cleaner Only groups for the ambient condition, Fig. 4c; Probiotic Cleaner group for the humid condition, Fig. 5c). These genes include ssuEADCB and tauABCD operons known to assimilate sulfur from aliphatic sulfonates 51 and facilitate the utilization of taurine as an alternative sulfur source in sulfate-deprived environments 52 .
Owing to the presence of anionic surfactants in liquid detergent products, such as sodium lauryl ether sulfate (SLES) 53 , these sulfate starvation genes were likely suppressed in Cleaner Only and Probiotic Cleaner groups. This observation is consistent with studies that illustrated the overexpression of tauABCD and ssuEADCB genes induced by sulfate shortage 54 . Additionally, bacterial utilization of SLES facilitated by Citrobacter braakii 55 and a consortium of Acinetobacter calcoacetiacus, Klebsiella oxytoca, and Serratia odorifera 53 has also been reported in previous studies and may contribute to the persistence of microbial communities associated with hospital sinks 56 . On the contrary, no differential expression was observed for sulfur metabolism genes in K. pneumoniae Probiotic Cleaner group under both ambient and humid conditions (Figs. 4f and 5f).
In K. pneumoniae experiments, Cleaner Only and Germinated Probiotic groups showed similarities in gene expressions while Probiotic Cleaner appeared to have very limited effects on K. pneumoniae gene expression. It is possible that the dormant Bacillus spores in Probiotic Cleaner were buffering the stress caused by chemical cleaner while the vegetative Bacillus cells introduced competition pressure on the surfaces. Thus, it is important to understand how spores and vegetative probiotic microorganisms interact with and affect the performance of chemical surface cleaning reagents, which can alter the response landscape of the surface associated pathogens. Since the results from our study showed species-specific responses, more pathogen species should be investigated for their interspecies interactions with probiotic microorganisms to observe general patterns. This will set the foundation and facilitate the development of targeted cleaning and disinfection strategies.
Effects of elevated temperature and humidity on pathogen cell motility and biofilm formation Although over 1000 differentially expressed genes were identified from A. baumannii or K. pneumoniae transcriptomes between the ambient and humid conditions, the majority of the KEGG pathways were composed of genes that were both upregulated and downregulated, rendering non-significant results for most pathways.
Under the humid condition (elevated temperature and humidity), 16 KEGG pathways related to metabolism (10), membrane transport (3), biosynthesis (2) and translation (1) were enriched in K. pneumoniae experiments (Fig. 5d, Probiotic Cleaner Humid vs No Cleaner Humid) while only 4 KEGG pathways related to metabolism (2), biosynthesis (1) and energy production (1) were enriched in A. baumannii (Fig. 5a, Probiotic Cleaner Humid vs A. baumannii Humid) with Probiotic Cleaner. Overall, more KEGG pathways were affected by elevated temperature and humidity in K. pneumoniae than A. baumannii ( Fig. 5d and a). This result corresponded to the PCA and PERMANOVA results where temperature and humidity affected the transcriptome profiles of K. pneumoniae more than A. baumannii, but not reflected in the survival of the two pathogens under elevated temperature and humidity assessed by viable CFU counts.
Without Probiotic Cleaner, cell motility appeared to be affected differently by increased moisture availability for A. baumannii and K. pneumoniae. A. baumannii exhibited upregulation of type IV pili genes (Fig. 5b) whereas K. pneumoniae type 1 pili genes remained unaffected, which was also the case under ambient condition (Fig.  S6D). Nevertheless, 19 genes related to biofilm formation were differentially expressed (Fig. 5e, K. pneumoniae Humid vs K. pneumoniae Ambient), and the quorum sensing pathway, which affects biofilm formation of K. pneumoniae 57 was enriched under elevated temperature and humidity (Fig. 5d, K. pneumoniae Humid vs K. pneumoniae Ambient). This indicated that elevated temperature and humidity could assist the persistence of K. pneumoniae on surfaces (Fig. 2b) by altering the biofilm formation related gene expression.
For both A. baumannii and K. pneumoniae under Probiotic Cleaner scenario, sulfur metabolism pathway was downregulated under elevated temperature and humidity, which is consistent with the sulfur metabolism-related gene expression results discussed in the previous section when chemical detergent was present. The effects of Probiotic Cleaner on K. pneumoniae were likely masked by the changes due to higher temperature and humidity ( Fig. 5e and f), meaning the benefits of probiotic cleaning products could be more limited in tropical areas for certain types of pathogens.
Our results demonstrated the persistence of probiotic Bacillus included in the All-Purpose Probiotic Cleaner up to 72 h after cleaning. However, this persistence did not result in more efficient reduction of pathogen biomass. A. baumannii on surfaces cleaned using chemical-based detergent with and without probiotic Bacillus contained a comparable amount of viable pathogens. The transcriptome of A. baumannii with and without probiotic addition shared a high degree of similarity in overall gene expression. In contrast, the transcriptome of K. pneumoniae with Probiotic Cleaner showed a high degree of differences compared to samples with chemical Cleaner Only or Germinated Bacillus, including reduced response in genes related to antimicrobial resistance, sulfur metabolism and biofilm formation. We thus conclude Bacillus spores indeed survive unfavorable conditions, but that the small percentage of germinated Bacillus cells and/or slow germination rate on surfaces preclude interspecies interactions like competitive exclusion. Elevated temperature and humidity resulted in prolonged persistence of viable pathogens on surfaces. These conditions also had a pronounced effect on K. pneumoniae overall gene expression regardless of which cleaner was used. These results suggest that vigilant indoor climate control could contribute to infection prevention. Especially in tropical environments, our results suggest that reducing the temperature and relative humidity is a more important strategy for infection prevention than cleaning product selection.
The design of more effective probiotic cleaning products and strategies to promote "healthy" microorganisms will require a more nuanced understanding of the relevant environmental conditions and how they impact activity. Our study used microcosms to model environmental conditions in situ and enable the manipulation of individual parameters. While our microcosms do not represent the life histories of organisms in situ (e.g., previous exposure and selection on a surface), this type of work forms an important bridge between experiments in culture media, which do not reflect in situ conditions, and field studies, which preclude replication and manipulation. We further demonstrate that transcriptomics can be used to assess activity on surfaces. We advocate for the inclusion of such assessments in field studies, which are historically typically DNA-or culture-based, especially as RNA sequencing methods improve for low biomass samples. The assessment of cleaning and disinfection is currently based on log reductions of viable biomass. Using these tools to progress beyond taxonomic and viability profiling will allow us to include more nuanced endpoints, such as the prevention of difficult-todisinfect phenotypes like biofilm formation or antimicrobial resistance.

Microcosm set-up
A commercially available All-Purpose Probiotic Cleaner (referred to as Probiotic Cleaner herein) was obtained from Graz, Austria. Clinical isolates of A. baumannii (ABBL18) and K. pneumoniae (CRE231) were provided by Dr. Alan Hauser, Northwestern University Feinberg School of Medicine. To simulate a typical high-touch surface in a healthcare environment, 24 G stainless steel sheets were cut into "2 x 2" squares. A. baumannii and K. pneumoniae isolates were grown in 40 mL tryptic soy broth (TSB) for 23 h at 30 o C with continuous shaking to its mid log phase. The Probiotic Cleaner was filtered through a 0.22 µm syringe filter to create an Cleaner Only control without probiotic materials. Sterility of the Cleaner Only control was assessed by plating on tryptic soy agar (TSA). Bacillus spp. were germinated by mixing 0.1% Probiotic Cleaner with TSB for use as a model Germinated Bacillus control where cells were vegetative. This culture was incubated at 25 o C for 23 h with continuous shaking. Cells (Bacillus spp., A. baumannii and K. pneumoniae) were harvested and washed in phosphatebuffered saline (PBS) before inoculation. A total of six testing scenarios for each pathogen (twelve groups in total) were specified as shown in Table 1: Each "2x2" stainless steel surface was first inoculated with 50 µL A. baumannii (corresponding to 8.5 × 10 7 CFU) or 50 µL K. pneumoniae (corresponding to 1.0 × 10 8 CFU). 50 µL sterilized PBS buffer, filtered Probiotic Cleaner, Probiotic Cleaner, or vegetative Bacillus spp. inoculum was added to create No Cleaner control samples (A. baumannii or K. pneumoniae only), Cleaner Only groups (chemical cleaning samples), Probiotic Cleaner groups (probiotic cleaning samples), or Germinated Bacillus groups (vegetative cleaning samples), respectively. Following the introduction of bacterial inoculum, all surfaces were transferred into autoclaved mason jars (referred to as microcosm chambers) under two treatment conditions: (1) ambient temperature (25 o C) and relative humidity (20%), and (2) elevated temperature (37 o C) and relative humidity (>90%) with 10 mL of autoclaved milliQ water as in Hu (2019) 13 . The relative humidity was measured using Fisherbrand™ Traceable™ Humidity-On-A-Card™ Humidity Monitor (Waltham, MA, USA). Each microcosm chamber contained one surface sample. All microcosm chambers were stored in incubators maintained at 25 o C or 37 o C. Samples were sacrificed and biomass recovered by swabbing at three specific timepoints: 3 h, 24 h, and 72 h following the inoculation and cleaning. Triplicate samples were prepared for each combination of cleaning, timepoint, and incubation conditions (Fig. 1).

Swab sample collection
Swab samples were collected through a combination of dry and wet swabbing 13 . Each surface was first dry swabbed from left to right for seven times and the swab was swirled into an aliquoted PBS buffer or RNAprotect Bacteria Reagent (Qiagen), for colony forming units (CFUs) enumeration and RNA extraction, respectively. Surfaces were then wet swabbed for a total of 20 s, including 5 seconds of rinse and rewet. Finally, the swabs were left in 15 mL conical tubes and vortexed for 10 s. For total enumeration, 100 µL swab suspension along with its dilutions were spread onto TSA. Additionally, the same volume of swab suspension and dilutions were spread onto CHROM A. baumannii selective agar or on MacConkey agar (as K. pneumoniae selective media 58 ). Under the circumstances where Bacillus were introduced through Probiotic Cleaner and Germinated Bacillus cleaning, total CFUs were quantified using TSA. All plates were incubated at 37 o C. Numbers of CFUs were counted after three days.
DNA extraction, sequencing, and construction of reference genomes To obtain draft genomes for RNA read mapping, 5 mL A. baumannii or K. pneumoniae and 15 mL Bacillus spp. DNA was extracted using the DNeasy PowerSoil Pro Kit (Qiagen) following the manufacturer's instructions. DNA samples were submitted to the Microbial Genome Sequencing Center (MiGS; Pittsburgh, PA, USA) for library construction and short-read sequencing on an Illumina platform. KneadData v0.7.10 (https://bitbucket.org/ biobakery/kneaddata/) was then used for raw sequence quality control and contaminant removal with default parameters. Short reads were assembled into contigs using SPAdes v3.14.1 59 and each reference genome was annotated using Prokka v1.14.6 60 . Additional annotation of antibiotic and antimicrobial resistance genes were conducted using the Comprehensive Antibiotic Resistance Database (CARD) 61 . Assembled contigs from All Purpose Probiotic Cleaner were also binned into five metagenome-assembled genomes (MAGs) using metaWRAP 62 . CheckM 63 was used to assess the quality of microbial genomes. Taxonomic classification of each draft genome was determined by GTDB-Tk 64 , which is an algorithm designed to classify MAGs based on the Genome Taxonomy Database.
RNA extraction, sequencing, and data processing Swab samples were stored in RNAprotect Bacteria Reagent under -80 o C until extraction. RNA was extracted using the RNeasy Mini Kit (Qiagen) following its recommended "Protocol 2: Enzymatic lysis and mechanical disruption of bacteria", consisting of a 10 min treatment with 15 mg/mL lysozyme and 10 min beadbeating. Libraries were constructed at Northwestern NUSeq using the Illumina Total RNA Prep with Ribo-Zero Plus, according to manufacturers' protocol. A total of 110 samples (108 samples collected at various time points and cleaning scenarios, one kit control, and one sample prepared for between-batch normalization) were sequenced using an Illumina HiSeq4000 with 50 bp single end reads.

DATA AVAILABILITY
Raw sequencing data are available in NCBI Sequence Read Archive (SRA) under BioProject accession number PRJNA818256.