Evidence for Environmental–Human Microbiota Transfer at a Manufacturing Facility with Novel Work-related Respiratory Disease

Abstract

Rationale: Workers’ exposure to metalworking fluid (MWF) has been associated with respiratory disease.

Objectives: As part of a public health investigation of a manufacturing facility, we performed a cross-sectional study using paired environmental and human sampling to evaluate the cross-pollination of microbes between the environment and the host and possible effects on lung pathology present among workers.

Methods: Workplace environmental microbiota were evaluated in air and MWF samples. Human microbiota were evaluated in lung tissue samples from workers with respiratory symptoms found to have lymphocytic bronchiolitis and alveolar ductitis with B-cell follicles and emphysema, in lung tissue samples from control subjects, and in skin, nasal, and oral samples from 302 workers from different areas of the facility. In vitro effects of MWF exposure on murine B cells were assessed.

Measurements and Main Results: An increased similarity of microbial composition was found between MWF samples and lung tissue samples of case workers compared with control subjects. Among workers in different locations within the facility, those that worked in the machine shop area had skin, nasal, and oral microbiota more closely related to the microbiota present in the MWF samples. Lung samples from four index cases and skin and nasal samples from workers in the machine shop area were enriched with Pseudomonas, the dominant taxa in MWF. Exposure to used MWF stimulated murine B-cell proliferation in vitro, a hallmark cell subtype found in the pathology of index cases.

Conclusions: Evaluation of a manufacturing facility with a cluster of workers with respiratory disease supports cross-pollination of microbes from MWF to humans and suggests the potential for exposure to these microbes to be a health hazard.

At a Glance Commentary

Scientific Knowledge on the Subject

Multiple lines of investigation support the idea that the environmental microbiota influences the human microbiota of those exposed and may exert significant health effects. However, studies have been limited by the lack of parallel detailed environmental and host microbiota sampling and difficulty assigning a pathogenic role when health conditions are commonly associated with multiple risks factors beyond the environment. Occupational-related diseases present a rare opportunity to study a defined environment to which workers are exposed. Workers exposed to metalworking fluid (MWF), a fluid highly colonized with microbes, are at risk for respiratory symptoms and diseases. A novel and distinct work-related pulmonary condition characterized by lymphocytic bronchiolitis and alveolar ductitis with B-cell follicles and emphysema was described among workers exposed to MWF at a manufacturing plant. We therefore evaluated whether exposure to microbes in MWF from the plant influenced workers’ respiratory tract microbiota and potentially stimulated a local B-cell response.

What This Study Adds to the Field

Using multiple environmental and human samples, we demonstrate cross-pollination of microbes from the MWF to exposed workers. This association includes the presence of MWF-characteristic bacteria in the lung tissue of workers with this novel work-related pulmonary condition. We further demonstrate that the microbial communities in the MWF can trigger in vitro proliferation of murine B cells, a hallmark cell subtype in this condition. Our findings indicate that the work-related microbial environmental exposure is a determinant for the human microbiota, with possible implications for pulmonary health.

Associations between specific environmental exposures and lung health have been well documented through epidemiological studies of exposure to specific toxins such as particulate matter ≤2.5 μm in aerodynamic diameter, cigarette smoke, and diesel fumes. Much less is known about the effects of environmental microbes. Salient examples involve known environmental pathogens such as those causing Anthrax and Legionnaires’ disease. Humans are exposed to complex microbial communities that we now conceptualize under the term “microbiome.” Recent investigations have shown that exposure to specific environmental microbiomes has significant effects on respiratory health. Epidemiologic studies found associations between the exposure to environments rich in microbes (e.g., farming and household pets) and asthma prevalence (1–3). Experimental data in horses support that different microbial environments impact bronchial airway reactivity and inflammation (4).

Different working conditions likely shape the microbial communities present in the occupational environment. However, the impact of workplace environmental microbiota conditions on the respiratory health of exposed workers is difficult to evaluate (5). In this current investigation, the U.S. National Institute for Occupational Safety and Health (NIOSH) studied a cluster of a novel lung disease distinct from asthma and hypersensitivity pneumonitis at a facility that manufactured industrial equipment in part by machining steel and aluminum using water-based metalworking fluids (MWFs). MWF is a cooling and lubricating fluid that, once in use, becomes frequently colonized with microorganisms (6). Microbial control is needed to allow for MWF reuse, and particular strategies include 1) using biocides to control microbial colonization in preserved MWF and 2) engineering the MWF to promote the growth of a particular organism, such as Pseudomonas pseudoalcaligenes, in nonpreserved MWF. In the latter strategy, growth of Pseudomonas promotes fluid stability by outcompeting the growth of other organisms that may break down the fluid, thus reducing the need for the addition of biocides (7). The MWF represents a potential source of microbiota exposure from the environment, where in a confined area, workers may be consistently exposed. In this cross-sectional study, we evaluated the microbial composition of environmental and human samples with the understanding that specific working conditions may impact the microbial environment and the associated cluster of respiratory disease at the facility.

Between 1995 and 2007, four workers at the facility developed symptoms consisting of insidious onset of cough, wheeze, and dyspnea on exertion that led to diagnostic lung biopsies and resulted in lung transplantation in one of them. All four participants were never-smokers and reported symptom exacerbations related to work. Lung function testing revealed obstruction and decreased diffusion capacity, and chest computed tomography showed mild to confluent centrilobular emphysema. An examination of lung tissue revealed a unique pattern of lymphocytic bronchiolitis, alveolar ductitis, and emphysema with B-cell primary lymphoid follicles involving both respiratory bronchioles and alveolar ducts (BADE). These results identified a novel, previously unrecognized occupational lung disease and were reported elsewhere (8). In addition, other workers at the facility suffered work-related respiratory symptoms (9). Outbreaks of occupational asthma and hypersensitivity pneumonitis have been recognized in association with exposure to water-based MWF (6, 10), and its association with microbial exposure has been suspected, but an etiological agent has not been clearly demonstrated. Furthermore, the initial recognition of B-cell proliferative disease in these workers (8) and the known link between specific microbial agents and gastric and conjunctival B-cell proliferation (11–13) led us to investigate the potential role of exposure to the environmental microbiota on the host microbiota and the possible association with this occupational disease. If exposure to microbes present in MWF played a possible pathogenic role, we hypothesized that there should be increased microbial similarities between the microbiota in the MWF and the human microbiota of those affected with this occupational disease and/or in direct physical contact with the MWF at this manufacturing facility. In addition, we evaluated whether the MWF used in this manufacturing facility led to in vitro activation of B cells, the hallmark cell subtype of this novel occupational lung disease (8).

Methods

Human Subjects

The investigation was conducted according to the NIOSH Institutional Review Board requirements for health hazard evaluations. Samples were obtained under two separate protocols approved by the Centers for Disease Control and Prevention (12-DRDS-08XP and 16-RHD-03XP). All subjects in whom we obtained samples signed the appropriate informed consent form.

Human Lung Tissue Samples

Paraffin-embedded lung tissue samples were obtained from all four workers who underwent lung biopsy to evaluate their respiratory disease (case samples) over a period of approximately 10 years before 2014 (8). In addition, we obtained paraffin-embedded lung tissue samples from 16 patients matched for age and biopsy date who did not work at the facility but had samples obtained and stored at the same hospital (control samples) and were matched by age at time of biopsy (within 5 yr) and biopsy date (within 12 mo). For each case, four control subjects were selected (two with lung cancer other than lymphoma or metastatic disease, one with interstitial lung disease other than hypersensitivity pneumonitis or sarcoidosis, and one with normal lung tissue).

Human Airway and Skin Samples

Then, we included samples from 302 current workers obtained during a health evaluation by NIOSH in 2016. From those participants, a total of 901 samples were collected, including skin swabs (n = 299) and nasal swabs (n = 302) using Catch-All swabs (Epicentre) and oral washes (n = 300). These samples were collected on a NIOSH visit on September 12–16, 2016, and classified by the location in the facility where workers spent the most time (administration, assembly, or machine shop; Figure E1 in the online supplement).

Environmental Samples

Environmental samples included 79 fluid samples and 194 air samples obtained during the same health evaluation by NIOSH in 2016. Fluid samples included in-use MWF (both preserved and nonpreserved MWF) as well as controls: unused MWF (referred as neat MWF), municipal water, tube controls, and DNA isolation controls). Air samples were collected using 37-mm/2-μm polycarbonate filters (SKC Inc.) from areas throughout the facility (administration, assembly, or machine shop; Figure E1). Control air samples included samples collected outside the facility. The air filters are sterile as per manufacturer description. Once used for air sampling, they were sterilely contained and shipped to the Segal Laboratory, where they were processed under sterile conditions. All samples were stored frozen at −80°C until processing and analysis.

Background Samples

To evaluate for possible background contamination, we obtained n = 49 background samples that consisted of multiple possible entry ways for contamination (Figure E2). Tube controls were tubes used for the collection of MWF but without MWF (n = 3; background tube). Air filter controls were filters used to capture air samples that did not undergo air sampling within the facility (n = 20; background air). Skin swab controls were sterile swabs used for skin sampling (n = 3; background skin). Finally, sample processing and sequencing controls were reagent controls, including water used to dilute MWF, elution buffer alone, DNA-free water passed through DNA isolation kits, and DNA water plus library preparation reagents (n = 23; background tech).

Microbiota Sample Processing

Samples were processed to isolate DNA and perform 16S ribosomal RNA (rRNA) gene amplicon sequencing. For the paraffin-embedded lung tissue samples, DNA extraction was performed using a BiOstic FFPE Tissue DNA Isolation Kit (MoBio) following the manufacturer’s instructions. For skin, nasal, and oral samples, DNA extraction was performed using a DNeasy Powersoil HTP DNA Isolation Kit (MoBio) following the manufacturer’s instructions. For environmental samples, DNA extraction was performed using a DNeasy Plant Mini Kit (Qiagen) for air filters and a QIAamp DNA Mini Kit (Qiagen) for the fluid samples following manufacturer’s instructions. High-throughput sequencing of bacterial 16S rRNA gene amplicons (V4 region) was performed as 150 bp reads with a paired-end protocol using the MiSeq platform (Illumina) (14). Reagent control samples and mock mixed microbial DNA were sequenced and analyzed in parallel. Each unique barcoded amplicon was generated in pairs of 25-μl reactions with the following reaction conditions: 11 μl PCR-grade H₂O, 10 μl Hot MasterMix (2200410; 5 Prime), 2 μl forward and reversed barcoded primer (5 μM), and 2 μl template DNA. Reactions were run on a C1000 Touch Thermal Cycler (Bio-Rad) with the following cycling conditions: initial denaturing at 94°C for 3 minutes followed by 35 cycles of denaturation at 94°C for 45 seconds, annealing at 58°C for 1 minute, and extension at 72°C for 90 seconds, with a final extension of 10 minutes at 72°C. Amplicons were quantified using the Agilent 2200 TapeStation system and pooled. Purification was then performed using Ampure XT (Cat#A63882; Beckman Coulter) per the manufacturer’s instructions.

Microbial Data Analysis

The obtained 16S rRNA gene sequences were analyzed using the Quantitative Insights into Microbial Ecology 1.9.1 package (15). Reads were demultiplexed and quality filtered with default parameters. We required more than 1,000 reads in any sample, a threshold that was achieved with all samples. Sequences were then clustered into operational taxonomic units (OTUs) using a 97% similarity threshold with UCLUST (16) and the Greengenes 16S rRNA gene reference dataset and taxonomy (17). For each sample, the proportion of reads at the genus level was used as a measure of the taxonomic relative abundance in a specimen. Permutational multivariate ANOVA (PERMANOVA, adonis) testing was used to compare the β diversity based on the Bray-Curtis dissimilarity index. To decrease the number of features, we only focused on major taxa and OTUs, defined as those having a relative abundance >1% in at least one sample. No OTU was removed from the analysis. We used the ade4 package in R to construct principal coordinate analysis plots based on the Bray-Curtis dissimilarity index (18). For comparisons of β diversity or taxonomy between groups, nonparametric tests were used (PERMANOVA and Mann-Whitney). To evaluate the differences between groups of 16S rRNA gene-sequencing data, we used differential gene expression analysis based on the negative binomial distribution the DESeq2 (version 3.5; R Bioconductor) package (19). Further analysis was also done to identify contaminants from the three different sequencing datasets. To perform this, a prevalence-based method using the R package decontam (version 3.11) identified the prevalence of each OTU across background samples and compared the prevalence in different subsets of study samples (20). All microbiota sequence data are publicly available in the Sequence Read Archive (PRJNA579296 and PRJNA635409). All codes used for the analysis included in this manuscript are available at https://github.com/segalmicrobiomelab/niosh_microbiome_project. The microbiota characterization was approved by the NIOSH Institutional Review Board. All participating workers provided written informed consent.

Whole-Genome Shotgun Sequencing

To further explore taxonomic annotation of taxa of interest, a subset of samples (three MWF samples and two lung tissue case samples) with high abundance of marker taxa was used for whole-genome shotgun (WGS) sequencing. To this end, isolated DNA was used for library preparation using a Nextera DNA Flex Library Prep Kit (20018704; Illumina). Total DNA input was 10 ng for MWF and 100 ng for lung tissue. Sequencing was performed on an Illumina NovaSeq sequencer (Illumina) with a paired-ends 150 bp approach. Raw shotgun sequencing data were processed using Kraken (21). Reads from tissue samples were further aligned to MWF reads annotated to P. pseudoalcaligenes with VSEARCH (22) using usearch_global and 100% identity parameters. Only reads with complete alignment over the whole sequence were considered as exact matches. WGS sequence data are available through the NIH Sequence Read Archive at PRJNA635410.

B-Cell Exposure to MWF

Given that B-cell lymphocytes dominated the histopathology of the workers’ lung disease, we investigated the effect of MWF exposure on B cells in an in vitro model. B cells were obtained from the spleens of wild-type C57BL/6J 8–10-week-old female mice (Jackson Laboratory). The spleen tissue was mechanically disrupted and strained using a 40-μm filter. B cells were isolated using the Dynabead mouse CD43 isolation kit (ThermoFisher). Cells were labeled with cell trace violet proliferation dye (ThermoFisher). The following bulk MWF samples collected at the facility were used for experimental conditions: preserved neat, preserved in-use, nonpreserved neat, and nonpreserved in-use. Fluids were sterilized using sequential filtration with a 40-μm filter followed by a 20-μm filter (Millipore). Half a million purified B cells were then plated in 0.5 ml activation media (RPMI; Corning) containing 15% fetal bovine serum, N-2-hydroxyethylpiperazine-N′-ethane sulfonic acid, L-glutamate, nonessential amino acids, sodium pyruvate, penicillin, streptomycin, and β-mercaptoethanol. Cells were then cultured with either phosphate-buffered saline as a negative control, 200 ng/ml BAFF (B-cell activating factor) (R&D Systems) as a positive control that promotes B-lymphocyte survival but not proliferation, 20 μg/ml LPS (a component of endotoxin; Sigma-Aldrich) as a positive control that promotes B-lymphocyte survival and proliferation, or 25 μl filter-sterilized MWF (1:40 dilution). On Day 2 of incubation, 0.5 ml of media with an appropriate concentration of phosphate-buffered saline, BAFF, LPS, or MWF was added to the lymphocytes. On Day 4, bright-field images were recorded using a 40× objective on the EVOS FL Cell Imaging System (ThermoFisher), and flow cytometry was performed on a BD LSR Fortessa Cell Analyzer (BD Biosciences). Cells were stained with the following antibodies before flow cytometry: IgM fluorescein isothiocyanate (Jackson ImmunoResearch), IgG1 phycoerythrin, B220 PerCPCy5–5, CD19 PeCy7 (eBioscience), and CD138 allophycocyanin (BD Biosciences). The Pacific blue channel was used to visualize the cell trace violet proliferation dye (ThermoFisher). The presence of terminally differentiated B cells (plasma cells) was assessed in the in vitro cultures using CD138 (Syndican 1; BD Biosciences) staining and flow cytometry.

Results

Cohort

For this investigation, we obtained invasive profiling of the microbiota in lung tissue samples from four workers who underwent lung biopsy for the diagnosis of respiratory disease. The pathologic assessment identifying findings of BADE were described elsewhere (8). Briefly, the histology was reviewed by a committee of lung pathologists organized by NIOSH. The histological abnormalities were consistent across samples and characterized as bronchiolocentric lymphoplasmacytic infiltrates with CD20-positive B-cell primary lymphoid follicles without germinal centers involving both bronchioles and alveolar ducts; scattered CD3-positive T cells predominantly cuffing the B-cell follicles; no appreciable interstitial or airway fibrosis or granulomas; and airspace enlargement with septal wall fragmentation, indicative of mild to moderate histological emphysema. These histologic features, although individually nonspecific, formed a pattern distinct from any well-recognized disease entity. All four workers were located in the facility’s assembly and/or machine shop, and none worked in the administration area. We compared the microbiota in these samples with 16 lung biopsy control samples (from nonworkers as indicated in the Methods). In addition, we performed noninvasive profiling of the microbiota in skin and upper airway samples obtained in 302 workers. Table 1 shows the demographic and clinical characteristics for this cohort considering the different areas where subjects work (administration, assembly, and machine shop). Of note, cough, sneezing, and blocked nasal passages were more prevalent among workers located in the assembly area.

Table 1.

Demographic and Clinical Characteristics of Machine-Manufacturing Workers by Location

		Location
Characteristics	Overall	Administration	Assembly	Machine Shop	P Value*
		Questionnaire, n	302	110		88	104	—
Age, median (interquartile range), yr	45 (36–54)	47 (37–55)	45 (35–53)	42 (33–54)	ns
Sex, M, n (%)	280 (92.7)	—	—	—	—
Antibiotic use in the last 4 wk, n (%)	19 (6)	7 (6)	4 (5)	8 (8)	ns
Cough usual, n (%)	28 (9)	5 (5)	14 (16)	9 (9)	0.023
Wheeze 12 mo, n (%)	58 (19)	18 (16)	20 (23)	20 (19)	ns
Asthma attack 12 mo, n (%)	7 (2)	3 (3)	3 (3)	1 (1)	ns
Nose symptoms 12 mo, n (%)	150 (50)	49 (45)	43 (49)	58 (57)	ns
Eye symptoms 12 mo, n (%)	100 (33)	34 (31)	25 (28)	41 (39)	ns
Skin problem 12 mo, n (%)	32 (11)	10 (9)	12 (14)	10 (10)	ns
Nasal congestion runny nose, n (%)	53 (18)	16 (15)	20 (23)	17 (16)	ns
Sneezing, n (%)	43 (14)	10 (9)	20 (23)	13 (13)	0.020
Blocked nose, n (%)	42 (14)	7 (6)	19 (22)	16 (15)	0.008
Loss of smell, n (%)	14 (5)	4 (4)	4 (4)	6 (6)	ns
Postnasal drip, n (%)	34 (11)	12 (11)	10 (11)	12 (12)	ns
Face pain or sinus pressure, n (%)	25 (8)	8 (7)	9 (10)	8 (8)	ns
Sore throat, n (%)	11 (3)	5 (5)	4 (4)	2 (2)	ns
Asthma-like symptoms, n (%)	74 (24)	20 (18)	28 (32)	26 (25)	0.085
Asthma-like symptoms work-related, n (%)	14 (5)	2 (1)	5 (6)	7 (7)	ns
Hay fever dx after hire, n (%)	20 (7)	7 (6)	8 (9)	5 (5)	ns
Sinus dx after hire, n (%)	35 (12)	10 (9)	10 (11)	15 (15)	ns
Eczema dx after hire, n (%)	14 (5)	6 (5)	3 (3)	5 (5)	ns
Ever asthma dx after hire, n (%)	6 (2)	1 (0.9)	3 (3)	2 (2)	ns
Chronic bronchitis dx after hire, n (%)	2 (0.9)	1 (0.9)	1 (1)	0	ns
Pneumonia dx after hire, n (%)	19 (6)	7 (6)	4 (5)	6 (6)	ns
Cough start after hire, n (%)	19 (6)	4 (4)	9 (10)	6 (6)	ns
SOB start after hire, n (%)	14 (5)	2 (2)	6 (7)	6 (6)	ns
Wheeze start after hire, n (%)	48 (16)	15 (14)	19 (22)	14 (13)	ns
Chest tight start after hire, n (%)	25 (8)	5 (5)	10 (11)	10 (10)	ns
Flu-like illness start after hire, n (%)	26 (9)	8 (7)	9 (10)	9 (9)	ns

Definition of abbreviations: dx = diagnosis; ns = not significant; SOB = shortness of breath.

“Administration” refers to office-based jobs, “assembly” refers to jobs in production areas where industrial machines were assembled from component parts, and “machine shop” refers to jobs in production areas where metal was machined to make component parts. Work-related symptoms were defined as symptoms that improved away from the facility. Asthma-like symptoms were defined as at least one of the following: current use of asthma medicine, wheezing or whistling in the chest in the last 12 months, awakening with a feeling of chest tightness in the last 12 months, or attack of asthma in the last 12 months.

^*P values are based on comparison of the three location groups using chi-square and Cochran-Armitage Trend tests for categorical variables and the Kruskal-Wallis test and ANOVA for continuous variables. P ≤ 0.05 was considered significant (italics); P values ≤ 0.10 are presented.

Microbiota Characterization

The environmental microbiota of the manufacturing facility was evaluated in air and MWF samples and compared with that of the human samples. For almost all samples, the number of reads per sample was >5,000 (median, 27,115; interquartile range, 18,593–35,191). For most sample types, differences in microbiota characterization were noted between subgroups. In-use MWF had differences in α diversity and taxonomic composition (β diversity) when preserved MWF was compared with nonpreserved MWF (PERMANOVA; P < 0.01; Figures 1A and E3). However, airborne microbiota were similar (in terms of α and β diversity parameters) among air samples from the administration, assembly, and machine shop areas (Figure E4). Among lung tissue samples, there were no differences in α diversity between samples from cases and those from control lung tissue (P = not significant) (Figure E5). However, β diversity analysis of lung samples showed significant differences in composition (PERMANOVA; P = 0.048; Figures 1A and E5). Among samples obtained in 302 workers, α diversity was reduced in nasal samples of workers located in the assembly area, and there were taxonomic compositional differences in skin and nasal samples among workers from different areas (as indicated by PERMANOVA analysis of β diversity; P < 0.05; Figures 1A and E5) but not in oral samples.

Figure 1.

Compositional similarities between samples. (A) The graph shows a principal coordinate analysis based on the Bray-Curtis dissimilarity index for each sample type, coded by location. Overall, in-use metalworking fluid (MWF) and lung tissue clustered more closely than the other fluids obtained. The calculation between sample types and within MWF, skin, and nasal samples demonstrates significant compositional differences (see inset). (B–E) Degree of similarity between MWF and human samples. Microbiota similarities based on the Bray-Curtis dissimilarity index were explored for MWF (both preserved and nonpreserved) and lung samples (B), skin samples (C), nasal swab samples (D), and oral wash samples (E) (P value is based on Mann-Whitney test). Assem. = assembly; NP = nonpreserved; ns = not significant; P = preserved.

Next, we compared the degree of similarity between MWF and human samples based on the Bray-Curtis dissimilarity index. Figures 1B–1E show the degree of similarity between the MWF (both preserved and nonpreserved) and different types of human samples (lung, skin, nasal, and oral samples). Lung tissue case samples demonstrated a greater degree of similarity to both preserved and nonpreserved MWF samples than lung tissue control samples (P < 0.001). Nonpreserved MWF had greater similarity to human samples (skin, nasal, and oral) from workers in the machine shop compared with the human samples from workers in the administration area. A similar trend was noted for skin and oral samples among subjects in the assembly area compared with subjects in administration area. These data suggest that a transfer of microbes may occur from the MWF to the workers surrounding the area where it is used (assembly and machine shop areas). In contrast, β diversity analysis suggested no clear pattern of similarity between the composition of the air microbiota and the human samples (Figure E6). Together, these data suggest that the samples obtained from workers in the machine shop area are influenced by the microbial composition of MWF most likely through contact with MWF rather than airborne transmission (e.g., via droplets).

We then used DESeq2 analysis to identify the top differential taxa enriched in human samples and environmental samples (Figures E7–E12). Figure 2A shows many taxa differentially enriched across different samples. Notably, an OTU annotated to Pseudomonas (Pseudomonas_813945) was differentially enriched in lung, skin, and nasal samples as well as in MWF. Specifically, the relative abundance of Pseudomonas_813945 was enriched in nonpreserved MWF, in lung tissue samples from cases (compared with control lung samples), and in skin and nasal swabs from workers in the machine shop (compared with samples from workers in the administration or assembly areas; Figure 2B). Using Basic Local Alignment Search Tool (23), the sequence annotated as Pseudomonas_813945 in the 16S rRNA gene-sequencing data most closely aligned with P. andersonii, P. mendocina, P. pseudoalcaligenes, and P. oleovorans.

Figure 2.

A Pseudomonas operational taxonomic unit (OTU) found to be differentially enriched in human and environmental samples. (A) Violin plots display the adjusted P values for the comparison of the relative abundance of taxa between different groups, using DESeq2, within each sample type (cases vs. control for lung tissue; administration vs. assembly vs. machine shop for skin, nasal, oral, and air samples; and preserved vs. nonpreserved for metalworking fluid). Dot sizes are proportional to relative abundance for each OTU. (B) An OTU annotated to Pseudomonas (Pseudomonas_813945) was identified based on DESeq2 and found to be higher in case lung samples and consistently enriched in skin and nasal samples from employees working in the machine shop area. Lastly, this Pseudomonas OTU was enriched in nonpreserved metalworking fluid. *P value < 0.05 and ***P value < 0.001. Admin = administration; Assemb = assembly; Mach Shop = machine shop; MWF = metalworking fluid; NP = nonpreserved; P = preserved.

To further evaluate the speciation of the Pseudomonas identified, we performed WGS on three MWF samples (all nonpreserved) and two lung tissue case samples with a high relative abundance of Pseudomonas_813945. The sequence depth was >60 million reads in all samples (Table E1). MWF samples were dominated by P. pseudoalcaligenes, with 32.6% mean abundance, with other Pseudomonas also being among the most abundant in these samples. Lung tissue samples were, as expected, dominated by reads that could not be mapped to bacteria (93.2%) and that potentially represented eukaryotic host DNA. P. pseudoalcaligenes was found in trace amounts in these samples (0.0000364%). However, we found that these reads could be perfectly matched (100% identity) to P. pseudoalcaligenes reads from MWF samples, potentially suggesting that the bacterial DNA found in tissue samples might have originated from the MFW.

It is, however, important to notice that the low concentration of material and the extremely low abundance of bacterial reads in tissue samples makes it difficult to rule out other potential sources for the observed P. pseudoalcaligenes reads, including contamination. To further address this question using the 16S rRNA gene-sequencing data, we used Decontam (version 3.11), a bioinformatic pipeline that compares the composition of background control samples with study samples to identify the identity of each OTU as likely of background contaminant source (20). Decontam did not identify the Pseudomonas_813945 taxa as belonging to the contaminant group by prevalence (Figure E13A) or by relative abundance (Figure E13B and Table E2). Thus, the Pseudomonas_813945 likely represents a true taxon found in the tissue case samples of subjects who developed BADE as well as in the skin and nose samples of subjects working in this facility. Figure E13C shows that the relative abundance of Pseudomonas_813945 in negative control samples was low compared with that in other study samples.

In Vitro Exposure of Murine B Cells to MWF

To explore whether exposure to MWF could stimulate B cells, the hallmark cell of the peribronchiolar infiltrates detected in the four subjects that underwent lung biopsy (8), we exposed murine B cells to filter-sterilized MWF samples collected at the machining facility. Compared with neat MWF, B-cell exposure to in-use MWF led to enhanced survival and visible activation and proliferation of the B cells in vitro (P = 0.0003 for preserved; P = 0.0026 for nonpreserved) (Figure 3). These data support the idea that exposure to microbial products present in the in-use MWF might contribute to the histopathological derangements noticed in these workers.

Figure 3.

In vitro evaluation of murine B-cell proliferation after exposure to metalworking fluid (MWF). Murine B cells were exposed to different MWFs because this cell subtype is found as a hallmark cell for the histological abnormalities identified among subjects from this facility who underwent lung biopsy because of respiratory symptoms and were found to have of lymphocytic bronchiolitis, alveolar ductitis, and emphysema with B-cell primary lymphoid follicles involving both respiratory bronchioles and alveolar ducts. Top: Representative bright-field microcopy images of murine B cells after 4-day in vitro culture with media supplemented with either BAFF (B-cell activating factor) or LPS as positive controls or with one of the four MWF samples (neat preserved, in-use preserved, neat nonpreserved, and in-use nonpreserved). In the absence of additional stimuli, signaling via BAFF in naive cells promotes survival but not proliferation. LPS (positive control) induces the proliferation of B cells via TLR4 signaling. Larger blasting cells are evident in wells where the proliferation of B cells is induced, whereas only small dying cells are found in the wells with MWF neat conditions. Scale bars, 100 μm. Bottom panels (gated on live single cells) show representative fluorescence-activated cell-sorting analysis of cell staining with intracellular proliferation dye (cell trace violet). The dye stains free amines inside the cells and is diluted as cells proliferate. The percentage of the population that underwent proliferation is indicated. Cells exposed to BAFF are used as a reference population and are shown in gray.

Discussion

We recently reported BADE, a novel occupational lung disease characterized pathologically in part by expansion of peribronchiolar B-cell populations (8). Using samples obtained from subjects working from this facility, we have now found that the microbial environment present in MWF was associated with changes in the upper airway and lung microbiota. These analyses demonstrated differences in the bacterial populations in lung samples of subjects with histological findings of BADE, notably enrichment with Pseudomonas_813945. Further microbial differences were noted in skin and nasal samples from the employees working in different locations. Based on their work location, these samples had different degrees of similarity with the MWF microbiota. Importantly, among workers located in the machine shop area, where higher exposure to MWF is expected, the airway microbiome was consistently enriched with Pseudomonas_813945, a dominant OTU in MWF. These data suggest that a transfer of microbes occurs from the environment to workers exposed to MWF. Finally, as a proof-of-concept for the interaction between the immune system and MWF, we demonstrated that exposure to in-use MWF stimulates murine B-cell proliferation in vitro.

Evidence of the effects of the microbial environment on the human exposure is supported by the more than 900 nasal, oral, and skin samples obtained from current workers in the facility where this novel occupational lung disease was described. Prior literature has supported that the environmental microbiota contributes to the shape of the human microbiota. For example, horses held in pasture have a different microbiome than they do when moved to a barn (4). However, a major challenge is the design of human studies in which the degree of exposure could be controlled. A prior investigation in workers from a mouse research facility found that the composition of the nasal and skin microbiota is affected by the location of their work among the facility (24). A second challenge for the design of these human studies is to determine the possible clinical implications for the change in the human microbiota induced by the environmental exposure. Our investigation represented a unique opportunity to study a distinct microbial environment among a large set of individuals who have a well-defined degree of exposure to that environmental microbiota, given the defined regionality of workers’ locations in addition to a distinct cluster of a novel occupational disease with plausibility for being microbial triggered.

The subjects who underwent lung biopsy and were found to have BADE carried lung microbiota with distinctive features characterized by a microbial composition with greater similarity to the environmental microbiota present in the MWF in their workplace than the similarity between the microbiota in the lungs of control subjects and the MWF. Differences in β diversity and taxonomy demonstrate that the microbiota within the case tissue were different from those within the control tissue and were notably enriched with Pseudomonas. This specific Pseudomonas OTU was found to be highly dominant in nonpreserved MWF samples. Interestingly, this OTU was not found to be differentially enriched in air samples, suggesting that the main mode of transmission of this microbe is through contact rather than airborne. However, it is also possible that airborne transmission of this microbe occurs, but our sampling and/or sequencing method was not sensitive enough to detect this specific microbial signature. Interestingly, taxonomic identification of the identified Pseudomonas OTU using WGS suggested that it belonged to the species P. pseudoalcaligenes, an opportunistic human pathogen (25). Although the effects of the environment on the lung microbial populations have been suspected, paired analysis of the environment and lung microbiota is still uncommon in the literature, and here, we showed that it has the potential to provide novel insights about the effects of the microbial environment surrounding us.

The in-use (but not neat) MWF promoted enhanced survival, proliferation, and differentiation of murine B cells in vitro. The finding is important because it suggests that the MWF may be able to stimulate B cells in vitro. However, further experiments will be needed to understand how MWF and P. pseudoalcaligenes and/or other microbes may be associated with the pathological features found in this cluster of cases with a novel occupational lung disease that included bronchiolocentric lymphoplasmacytic infiltrates with CD20⁺ B-cell primary lymphoid follicles (notably, without germinal centers) involving both respiratory bronchioles and alveolar ducts (8). The primary function of mucosa-associated B cells in the airways is the production of immunoglobulins in response to antigens. In the normal lung, bronchial-associated lymphoid tissue (BALT) is found rarely (26, 27). Our findings of closer microbial relationships between case tissue and case environmental microbiota compared with control samples and the stimulation of murine B cells by in-use MWF in vitro are consistent with the hypothesis that the microbiota extant in the workplace influenced the lung microbiota and led to expansion of B cells seen in the lung tissue of these workers. This hypothesis is biologically plausible, given the known associations between MWF and other immune-mediated lung diseases, namely, hypersensitivity pneumonitis and asthma (6, 10). In addition, microbes have been found to have a clear role on the proliferation of mucosa-associated lymphoid tissues. For example, Helicobacter pylori has been found to have a well-defined pathogenic role, supported by extensive experimental data, in the development of gastric mucosa-associated lymphoid tissue, which is often responsive to antibiotics (12, 13). In the lung, multiple case reports have suggested associations between lung bacteria and BALT proliferation (28–32). This is relevant given the shared histological features of the lung disease in this facility and pulmonary BALT (8). Nonetheless, we recognize that the data presented do not provide causal evidence. Further assessment of the lung microbiota of exposed but healthy workers from this facility as well as experimental investigations using preclinical animal models would be needed to further support causality.

Our investigation has several limitations. The sample size for the lung tissue observations is small. A larger lower airway sampling of the workforce of this facility among subjects with symptoms and control subjects would be desirable but not feasible under the performed NIOSH investigation. Regarding the environmental samples, it is possible that our air sampling methodology was not able to capture aerosolized microbes of large size that may travel short distances. This NIOSH investigation did not use personal air samplers but rather stationary samples, for which a more granular individual environmental exposure is difficult to assess. However, most of the movement of the participants occurred in well-defined areas (administration, assembly, or machine shop). Thus, we categorized the participants by their work area to determine the similarities between their samples and those from the environment. The low microbial biomass in lung tissue and in some of the environmental samples represents a challenge, and a paraffin sample from the same hospital that could be used as a negative control was not available. However, multiple negative control samples were sequenced, including DNA-free water, and exhibited a microbiome pattern different from true samples. Moreover, neat MWF samples showed significant differences in diversity and composition compared with in-use MWF samples. Furthermore, all samples were processed together, and investigators conducting the DNA sequencing were blinded to the case–control allocation. Under ideal circumstances, lung tissue control samples would have come from a geographically proximate metalworking facility without cases of lung disease as well as from healthy workers at the study plant, although it would be ethically unacceptable for healthy workers to undergo lung biopsy for research purposes. Instead, we attempted to address the potential impact of geographic conditions on microbiota by using lung tissue control samples from the same hospital where cases underwent biopsy. Given that our study is a cross-sectional study, the directionality of any microbial transfer between the environment and humans cannot be firmly established. However, the P. pseudoalcaligenes identified in the MWF, in the lungs of the cases, and in the skin and nasal swabs from those working in zones where MWF is used is a well-defined environmental bacterium that colonizes nonpreserved MWF (7). Finally, murine B cells respond differently to microbial products than human B cells (33), and therefore, it is probable that the findings described here for murine B cells’ responses to used MWF are not generalizable to human B cells. Moreover, although mice and humans share many pattern-recognition receptors, including many homologous TLRs and NLRs, there are many specific differences driven by pathogen–host specificity and evolutionary biology that are not fully captured in the experiments herein. We used MWF to stimulate murine B lymphocytes because this represented a readily available source of cells and enabled a reductionist approach to study host–pathogen interaction. In the future, we look to understand further the in vivo interactions between MWF, P. pseudoalcaligenes, and lower airway immune response in BADE.

In summary, our investigation supports the hypothesis that the environmental microbiota affects the host microbiota, having cross pollination of microbes not just to the superficial upper airway respiratory interface but also further down to the lower airways. The findings that enrichment with distinct microbes occur in association with a cluster of the newly described occupational lung disease BADE is consistent with the hypothesis of a possible pathogenic role for the environmental microbiota. The finding that used, but not neat, MWF can trigger the proliferation of B cells is also potentially supportive. Further longitudinal and experimental investigation is needed to elucidate the potential mechanisms among human B cells in the pathogenesis of this disease. Indeed, considering the growing evidence of associations between respiratory health and the microbial environment (e.g., asthma and farming) (2, 34), the results presented here highlight the potential opportunity for mechanistic discovery through the characterization of the environmental and human microbiota in a well-defined ecological niche.