Autoimmune markers for progression of Libby amphibole lamellar pleural thickening

Abstract

Exposure to Libby Asbestiform Amphibole (LAA) is associated with asbestos-related diseases, including mesothelioma, pulmonary carcinoma, pleural fibrosis, and systemic autoimmune diseases. The pleural fibrosis can manifest as a rapidly progressing lamellar pleural thickening (LPT), which causes thoracic pain, dyspnea, and worsening pulmonary function tests (PFT). It is refractory to treatment and frequently fatal.

Objective:

Because of the immune dysfunction that has been described in the LAA-exposed population and the association of pleural manifestations with the presence of autoantibodies, this study tested whether specific immunological factors were associated with progressive LPT and whether they could be used as markers of progressive disease.

Methods:

Subjects were placed into three study groups defined as (1) progressive LPT, (2) stable LPT, (3) no LPT. Serum samples were tested for antinuclear autoantibodies, mesothelial cell autoantibodies, anti-plasminogen antibodies, IL1 beta, and IL17; which have all been shown to be elevated in mice and/or humans exposed to LAA.

Results:

Group 1 had significantly higher mean values for all of the autoantibodies, but not IL1 or IL-17, compared to the control Group 3. All three autoantibody tests had high specificity but low sensitivity, but ROC area-under-the-curve values for all three antibodies were over 0.7, statistically higher than a test with no value. When all LPT subjects were combined (Progressive plus Stable), no marker had predictive value for disease.

Conclusion:

The data support the hypothesis that progressive LPT is associated with immunological findings that may serve as an initial screen for progressive LPT.

Introduction

Through the Libby Epidemiology Research Program (LERP), funded by the Agency for Toxic Substances and Disease Registry (ATSDR, TS000099–01), autoimmune outcomes were found to be associated with exposure to Libby Asbestiform Amphibole (LAA) fibers (Marchand et al. 2012; Black et al. 2014; Pfau et al. 2014; Diegel et al. 2018). While typically asbestos exposure (most often commercial chrysotile asbestos) is known to cause cancer (mesothelioma and pulmonary carcinoma) and pulmonary fibrosis (asbestosis), people exposed to LAA also tend to manifest a highly inflammatory and progressive pleural fibrosis now termed Lamellar Pleural Thickening (LPT) (Whitehouse 2004; Black et al. 2014; Szeinuk et al. 2016). Exposures to LAA occurred in and around Libby, Montana due to the mining and extensive community use of vermiculite that was laden with fibrous amphiboles, which was then shipped to many processing sites throughout the United States and widely used in commercial products such as insulation for homes and other buildings. More recently, similar fibrous amphiboles have been discovered in the arid lands around Las Vegas, Nevada, where dust storms and recreational activities often lift these toxic mineral fibers into the air, leading to concerns regarding widespread exposures (Buck et al. 2013; Metcalf and Buck 2015). Recent studies demonstrate that the health outcomes from even low-level exposures to this material could mimic those seen with LAA, which suggests an emerging public health problem (Pfau et al. 2017).

Progressive LPT is defined as a diffuse pleural scarring seen on CT scan, severe pleural pain, and/or progressive pulmonary function decline (Black et al. 2014; Szeinuk et al. 2016; Miller et al. 2018). These conditions lead to disability and death. Currently, the only treatments are palliative. LPT may be associated with specific autoantibodies that are driving the inflammation and scarring in the pleura as well as the associated systemic manifestations that can resemble systemic autoimmune diseases like rheumatoid arthritis (RA), systemic sclerosis (SSc) and systemic lupus erythematosus (SLE) (Diegel et al. 2018; Pfau 2018). Autoantibodies in both mice and humans exposed to LAA include anti-nuclear autoantibodies (ANA) and mesothelial cell autoantibodies (MCAA) (Marchand et al. 2012; Ferro et al. 2013; Serve et al. 2013; Gilmer et al. 2016), and both have been shown to be associated with radiographic evidence of pleural disease (Marchand et al. 2012). LAA-induced MCAA drive collagen matrix production by mesothelial cells in both cell culture and in naïve mice, suggesting that they could be a mechanism behind LPT (Serve et al. 2013; Gilmer et al. 2016). Anti-plasminogen antibodies were shown to be a component of the MCAA, and inhibition of the MCAA binding interaction reduced collagen matrix formation by MCAA (Hanson et al. 2016). Therefore, these autoantibodies are implicated in LAA-induced LPT. In mouse models, exposure to LAA also elicits cytokines indicative of a T_H17 response (Ferro et al. 2013; Zebedeo et al. 2014). IL-1 beta drives T_H17 responses and is part of the inflammasome pathway (Wilson et al. 2010; Biswas et al. 2011; Song et al. 2014). This cascade may be critical in the convergence of fibrotic and autoimmune characteristics that are seen in Libby.

The objectives of this project were to (1) determine whether or not these specific autoantibodies or cytokines were associated with progressive LPT, and (2) to perform sensitivity and specificity testing to determine the potential value of these markers in screening for progressive disease. The study successfully identified several immunological parameters that were associated with progressive disease, and these data may inspire new searches for therapeutic strategies. In addition, autoantibody testing, including ANA, may be valuable as a first screening step in early identification, or even prediction of progressive disease.

Materials and methods

Subjects

All human subjects protocols were approved by the Montana State University Institutional Review Board (IRB) and the Providence Health Care IRB, which oversees the activities of the Center for Asbestos Related Diseases (CARD). All subjects were adults and were members of the CARD research database, which requires that they had lived or worked in Libby for at least 6 months, at least 10 years ago (due to latency of disease). To be included in the current study, subjects needed to have a CT scan within the last 2 years and to have already provided, or be willing to provide, a serum sample within 2 years of their CT scan. The CARD database contains extensive data regarding LAA exposure, medical history, CT scans, pulmonary function testing (PFT), and symptoms including the St. George’s Respiratory Questionnaire (SGRQ) on subjects who have given informed consent to be included in on-going research regarding LAA exposures.

Subjects were identified who met the following criteria for the study groups, using the following definitions:

A patient has progressive LPT IF:

1. CT scan in the last 2 years is positive for lamellar pleural thickening

2. Significant change from previous imaging based on physician’s review and/or symptoms of chest pain

3. SGRQ total score of >15 and any one score (symptoms, activity, or impacts) of >30

4. Requires any of the following: increasing pain medication for chest pain; reducing daily activities due to dyspnea, increasing use of oxygen

A patient has stable LPT if:

1. CT scan in the last 2 years is positive for LPT

2. No significant change from previous imaging based on physician’s review and/or symptoms of chest pain

3. Does not require any of the following: increasing pain medication for chest pain; reducing daily activities due to dyspnea, increasing use of oxygen

A patient has no LPT if

1. CT scan in the last 2 years is negative for LPT

2. No significant decline in pulmonary function over the last 2 years

3. Does not require any of the following: pain medication for chest pain; reducing daily activities due to dyspnea, use of oxygen

After selection criteria were met, two subjects were selected for both the Stable LPT (n = 38) and the No LPT (n = 38) groups that were age- and sex-matched as closely as possible to the subjects in the Progressor group (n = 19). Table 1 describes the demographics of the three study groups.

Table 1.

Study group demographics and history.

	Progressive LPT	Stable LPT	No LPT	p value
N	19	38	38
Age (SD)	61.5 (8.3)	61.9 (8.5)	60.7 (7.1)	0.79c
Male: Female	17:2	34:4	34:4
Years in Libby (SD)	21.0 (16.0)	31.4 (14.1)	27.2 (17.5)	0.07c
Exposure Scorea (SD)	42 (29.9)	56.6 (39.7)	48.1 (35.8)	0.31c
Exposure Typeb				0.08d
Environmental n (%)	3 (15.8%)	12 (31.6%)	12 (31.6%)
Household n (%)	10 (52.6%)	21 (55.3%)	22 (57.9%)
Occupational n (%)	6 (31.6%)	6 (15.8%)	2 (5.3%)
Smoking
Pack years (SD)	6.2 (9.1)	26.2 (21.1)	18.2 (19.5)	0.0008c
Current n (%)	2 (10.5%)	10 (26.3%)	8 (21.1%)	0.02d
Previous n (%)	6 (31.6%)	21 (55.3%)	21 (55.3%)
Never n (%)	11 (57.9%)	8 (21.1%)	9 (23.7%)

^aModified from Noonan (2006), based on primary exposure pathway and duration.

^bPrimary exposure route reported by patient.

^cOne-Way ANOVA, with Bonferroni post hoc test.

^dChi squared test, 3 × 3 table for all exposure or smoking history.

The presence/absence of autoimmune diseases was not a criterion for the study. The hypothesis is that progressive LPT is an autoimmune disease, and autoimmune diseases can occur in combination. If ANA-positive individuals or people with SAID were excluded, those would be the very people hypothesized to be at highest risk. The study bins were filled solely based on their pleural disease status.

CT scans

All CT scans were performed prior to this study as part of regular patient care at the CARD clinic. All subjects used in this study had signed consent forms for research as part of an approved IRB protocol. Tests were either low-dose lung cancer screening CT scans, or high-resolution images for evaluation of asbestos-related disease. Scans were performed at Cabinet Peaks Medical Center in Libby, Montana. Subjects were scanned in a prone position using a 16 slice GE Lightspeed CT scanner. Scans were read by a radiologist contracted with Cabinet Peaks Medical Center within 24 hours of the scan so that urgent findings could be immediately identified and addressed by medical staff. All images were also read by Dr. Brad Black at the CARD clinic for the presence of asbestos-related disease.

LAA exposure

All LAA exposure data are based on exposure pathways and an exposure matrix developed specifically for this exposure cohort (Noonan 2006; Noonan et al. 2015). Pathways include occupational (worked at the mine or processing facilities), household (home insulated with Libby vermiculite or worker brought fibers home on clothes), and environmental (used Libby vermiculite for gardening, recreated in areas containing Libby vermiculite). For this study, LAA exposure was ranked as 3 = Occupational, 2 = Household, and 1 = Environmental. These were multiplied by the number of years when such exposures occurred to calculate a rough exposure score (Table 1) (Noonan 2006). Table 1 also provides the average number of years spent in the Libby area for each subject group and the number of people reporting primarily Environmental, Household, or Occupational exposure.

Serum collection and storage

All serum was collected according to established clinical protocols at the CARD clinic, and stored at −80 °C. The samples were sent to Montana State University on dry ice for serology testing. Repeat freeze/thaw cycles were avoided, and samples were stored at −80 °C until use, and at 4 °C during the testing phase.

ANA testing

Briefly, samples were diluted at 1:40 in phosphate-buffered saline and then placed on wells of commercial Hep2 indirect immunofluorescence ANA slides from ImmunoConcepts (Sacramento, CA). Following incubation, slides were washed and then the secondary antibody (anti-human IgG FITC conjugate) was added to each well. After incubation, washing and cover-slipping, slides were viewed by two blinded readers using a Leica upright fluorescence microscope, and compared against positive controls provided with the slides. Patterns and brightness were evaluated as negative or positive, with positives given a score of 1–4 based on staining intensity. If positive, the samples were titered. For sensitivity/specificity testing, a test was positive for any score from 1–4 at the 1:40 dilution. For ROC analysis, titers of 1:40, 1:80, 1:160 and 1:320 were considered as the different positive/negative cutoff points.

MCAA testing

A cell-based ELISA was performed as previously described (Marchand et al. 2012) to determine the presence of MCAA. Briefly, MeT-5A mesothelial cells (ATCC, Manassas VA) were seeded at confluency on 96-well plates, attached overnight and fixed in 1% paraformaldehyde. Following washing with PBS-Tween (0.05%), cells were blocked with 5% dried milk/PBS and then exposed to serum samples diluted in 3% BSA/PBS (1:100). Following a 2-h incubation with primary antibody, cells were washed and blocked a second time. The secondary antibody HRP-conjugated goat anti-human IgG (Jackson ImmunoResearch, West Grove, PA) was applied at a dilution of 1:1000 in 3% BSA/PBS and incubated for 1 h. Excess antibody was washed off and plates developed using TMB reagent (Thermo Scientific) followed by 1 M HCl. Plates were analyzed at 450 nm on a VersaMax microtiter plate reader with Softmax Pro software (Molecular Devices, San Jose CA). Nonspecific secondary antibody binding was corrected for on a plate-to-plate basis by subtracting the mean optical density (OD) for the secondary antibody-only control wells from the mean OD of each sample. Samples were determined to be positive if the OD was 3 standard deviations above the mean for a set of normal human serum samples (PrecisionMed, Solana CA). Similarly, for each sample, the number of standard deviations above the mean of the normal sera was considered the MCAA score. For ROC testing, cutoff points were 2, 3, 5, and 8 standard deviations above the mean for the normal samples.

Antiplasminogen testing

Purified human PLG (R&D Systems, Minneapolis, MN) was coated to high binding ELISA plates in carbonate-coating buffer (pH 8) overnight. After blocking with 5% nonfat dry milk for 2 h, samples from LA-exposed MCAA positive and negative serum samples were added to wells in duplicate. Before being added to the wells, the serum was diluted 1:100 in 3% BSA in PBS. Serum samples from healthy controls that had been shown to be MCAA negative by cell-based ELISA as previously described (Marchand et al. 2012) were used to provide a control group. After 2 h at room temperature, the wells were washed in PBS-Tween and then stained with anti-human IgG HRP antibodies for 1 h, followed by development, using TMB substrate. The optical density of the wells was analyzed at 450 nm using the BioTek plate reader. Samples were determined to be anti-PLG positive if the OD value for that sample was at least three standard deviations above the mean OD for wells with known normal/negative controls (Hanson et al. 2016). For ROC testing, cutoff values at 2, 3, 5, and 8 standard deviations above the mean were also used.

Cytokine testing

The presence of IL-1beta or IL-17 was detected using commercial human ELISA kits from eBioscience, with testing performed according to the manufacturer’s protocol. Each serum sample was diluted in the kit’s dilution buffer and placed in duplicate wells coated with antibodies to the cytokine. After incubation and washing, anti-human IgG with HRP conjugate was added to the wells. After incubation and washing, the wells were developed using TMB substrate, followed by 1 M HCl. Absorbance/optical density (OD) was read on the plate reader. The concentration of cytokine in each sample was determined using the standard curve provided with the kits. This value was used for both graphing the averages and testing for sensitivity/specificity. For ROC testing, four cutoff points were determined by dividing the entire set of OD values into quartiles.

Collagen matrix testing

Human Met5a cells were plated into a 96-well plate at 70,000 cells per well. The plates were incubated at 37 °C for 3 h. The cells were then treated with diluted serum (1:100) from each of the subjects. The cells were then incubated for 72 h at 37 °C and rinsed once with PBS. All blocking and incubation steps were performed under gentle agitation at room temperature for 1 h with washing steps occurring immediately following the incubation steps. The wells were then blocked with 200 μl 5% nonfat dry milk in PBS. The wells were incubated with 2 μl of mouse IgG targeted to human collagen type 1 (ab6308; Abcam, Cambridge, MA) in 100 μl of 3% BSA. The plates were then washed 39 with 200 μl of 0.05% Tween in PBS for 4 min each wash. The plates were blocked again with 5% milk in PBS. The plates were then incubated with goat anti-mouse IgG conjugated to HRP (Life Technologies). The plates were washed 39 with 200 μl of 0.05% Tween in PBS. The plates were developed with 100 μl of Pierce One Step TMB ELISA reagent. The reactions were stopped with 50 μl of 1 M HCl. The plates were then read for absorbance at 450 nm on the plate reader. The OD value was used for graphing the averages. For testing for sensitivity/specificity and for ROC testing, four cutoff points were determined by dividing the entire set of values into quartiles.

Sensitivity/specificity testing and ROC analysis

Sensitivity and specificity test was performed using Group 1 (Progressor) only and then combining Groups 1 and 2 to include any LPT, whether stable or progressive. The equations used for each of the assays are as follows:

$$\text{Sensitivity = True positives/( True positives}\text{+ False negatives)}$$

$$\text{Specificity = True negatives/( True negatives}\text{+ False positives)}$$

Positive and Negative Predictive Values were determined using Open Source OpenEpi, Version 3 for diagnostic tests (Sullivan et al. 2009).

Combination sensitivity and specificity testing were performed for ANA plus MCAA and for ANA plus anti-PLG, using the following equations (Kanchanaraksa 2008). In each case, ANA testing was test A and the other test was test B. For 2-stage (sequential) testing (Test B follows Test A if A is positive), the equations were:

$$\text{ Sensitivity of AB}=(\text{ Sensitivity of A})\times (\text{ Sensitivity of B})$$

$$\text{Specificity of AB}=(\text{Spec}\text{A}+\text{Spec}\text{B})\times (\text{Spec}\text{A}\times \text{Spec}\text{B})$$

For simultaneous (‘net’) testing (both A and B test done at the same time, and result must be positive in either or both tests), the equations were:

$$\begin{gathered} \text{ Sensitivity of AB}=(\text{ Sens A}+\text{ Sens B})-(\text{ Sens A}\times \text{ Sens B}) \\ \text{ Specificity of AB}=(\text{ Spec A})\times (\text{ Spec B}) \end{gathered}$$

Receiver/operator curve (ROC) testing

ROC testing was also performed using OpenEpi, providing the True Positive Rate (Sensitivity) for the Y-axis and the False Positive Rate (1-Specificity) for the X-axis. OpenEpi also provided the 95% confidence intervals for sensitivity, specificity, positive and negative predictive values and for the ROC area under the curve (AUC).

Data presentation and statistics

All data are presented in the graphs either as bars representing the mean value for the study group with an error bar representing the standard error of the mean (SEM), or as bars representing frequency (percentage). For the former, data were analyzed by one-way ANOVA, with post hoc testing including Bonferroni’s test and the Tukey–Kramer test, using StatPlus software. The frequency data were analyzed using Chi square testing (GraphPad QuickCalcs on-line). Statistical significance for both types of analyses was defined as p < 0.05.

ROC statistics were analyzed using OpenEpi, version 3 (Sullivan et al. 2009), followed by calculation of the Standard Error, Z value, and p value using the method by Delong et al. (1988), and recommended by Hajian-Tilaki and Hanley (2002) (Hanley and McNeil 1982; DeLong et al. 1988; Hajian-Tilaki and Hanley 2002).

Results

LAA exposure

Due to the limited exposure data available for all of the pathways of exposure to LAA (Noonan 2006), exposure scores were estimated based on primary exposure pathways reported by each subject multiplied by the length of time exposure may have occurred via the primary pathways. Table 1 shows that the primary pathways were similar in all three groups, and there was no significant difference between the average estimated exposures in the three subject groups. There was a slightly higher frequency of occupational exposures in the Progressive LPT group compared to the other two groups, and while not statistically significant, it might be valuable in future studies to evaluate differences in disease progression or severity that might be related to confounders for occupational exposures such as other contaminants. The Progressive LPT group reported significantly fewer smoking pack-years that the other two groups. There is no evidence in the literature to suggest a protective effect of smoking on pleural fibrosis or autoantibody production. It is possible that this group stopped smoking earlier in life due to respiratory problems.

Antinuclear autoantibodies

The frequency of positive ANA tests was significantly higher in the Progressor group compared to either the Stable or No LPT groups (Figure 1, top). This was true in both males and females (Figure 1, bottom). There was no association between LPT and the titer of the ANA, and no difference in median ANA titer (only performed on positive tests) among the groups (data not shown). The ANA titers were only used for the cutoff points required for ROC testing, below, and ranged from 1:40 to 1:320.

Figure 1.

ANA testing was performed as described in Materials and Methods. Top: Frequency of positive ANA tests in each subject group at 1:40 serum dilution. Bottom: Frequency of positive ANA tests in each subject group, separated by sex. *=p < 0.05 by Chi square analysis, compared to the Progressor group. Note: n = 2 females in the Progressor group.

Mesothelial cell autoantibodies

Positive MCAA tests were most frequent in the Progressor group, though statistically only greater than the No LPT group (Figure 2, top). When comparing the MCAA scores (semi-quantitative value for amount of MCAA), the Progressor group had the highest score compared to both the Stable and No LPT group (Figure 2, middle), suggesting a higher titer of the antibodies in this group. When broken down by sex of the subjects, this relationship only existed within the males (Figure 2, bottom). It is important to note that there were only 2 females in the Progressor group.

Figure 2.

MCAA testing was performed as described in Materials and Methods. Top: Frequency of positive MCAA tests in each subject group. *=p < 0.05 by Chi square analysis. Middle: MCAA scores (3 of standard deviations above the mean of a normal serum set) in each subject group. Bottom: MCAA scores, separated by sex. *=p < 0.05, compared to the Progressor group. Note: n = 2 females in the Progressor group.

Antiplasminogen autoantibodies2

Anti-PLG antibodies belong to the overall MCAA family of autoantibodies (Hanson et al. 2016). We hypothesized that they might be more specific to LAA exposure, and therefore more predictive. Similar to the MCAA results, anti-PLG antibodies were most frequent in the Progressor group, though statistically only greater than the No LPT group (Figure 3, top). The Progressors had significantly higher anti-PLG scores than both the Stable LPT and No LPT groups (Figure 3, middle). When separated by sex, the Progressors had significantly higher scores than either the Stable LPT or No LPT group, but this was only statistically significant in the males (Figure 3, bottom). Again, it is important to note that there were only two females in the Progressor group.

Figure 3.

Anti-PLG testing was performed as described in Materials and Methods. Top: Frequency of positive anti-PLG tests in each subject group. * = p < 0.05 by Chi Square analysis. Middle: Anti-PLG scores (3 of standard deviations above the mean of a normal serum set) in each subject group. Bottom: Anti-PLG scores, separated by sex. *=p < 0.05, compared to the Progressor group. Note: n = 2 females in the Progressor group.

Collagen production by mesothelial cells

Testing was performed to determine the ability of antibodies in the patients’ serum to drive collagen production by cultured human mesothelial cells. Because this is a functional test and not a biochemical marker, this is not proposed as a clinical test, but rather as a demonstration of the meaningfulness of MCAA testing as a determinant of pathological potential of these antibodies. The results with this test were very similar to the results with MCAA testing. Figure 4 illustrates that collagen production is stimulated to a higher degree by sera from the Progressor group compared to the No Disease group. The Stable disease group was intermediate in terms of collagen production, still statistically higher than the No LPT group.

Figure 4.

Collagen matrix formation in vitro by Met5A mesothelial cells was stimulated by addition of serum (diluted 1:100) from each of the subjects in each group to the cultured cells in wells. Data are shown as the average optical density (OD) for each subject group. a=p<0.05 compared to Progressor group; b=p<0.05 compared to Stable group.

IL-1 beta and IL-17

Figures 5 and 6 show the data for IL-1beta and IL-17 ELISA optical density (absorbance) values in serum samples from each of the groups, as average and standard error of the mean. There were no statistically significant differences between the groups for either cytokine. The averages ranges for these cytokines were low to nondetectable, as would be expected from normal human serum according to the information provided with the ELISA kit. Therefore, in this population, these cytokines do not appear to be elevated due to the LAA exposure, nor disease.

Figure 5.

IL-1beta was measured in serum samples using an ELISA kit. Data shown are concentrations in the serum as calculated using a standard curve included with the kit. Error bars are SEM.

Figure 6.

IL-17 was measured in serum samples using an ELISA kit. Data shown are concentrations in the serum as calculated using a standard curve included with the kit. Error bars are SEM.

Sensitivity/specificity and ROC testing

Table 2 summarizes the results of testing done to evaluate the efficacy of these tests as potential screening tools for progressive LPT. All three antibody tests (ANA, MCAA and PLG) had higher specificity than sensitivity, such that a negative test has value in stating that progressive LPT is unlikely, but that false positives would occur. The most specific test, for anti-PLG antibodies had the strongest predictive values and had an AUC p value of 0.0001.

Table 2.

Sensitivity/specificity data for progressive disease^a.

	Sensitivity % (95% CI)	Specificity % (95% CI)	Predictive Value (%) (95% CI)	ROC Curve AUC (95% CI)	AUC p valueb
ANA	52.6 (31.7, 72.7)	84.2 (69.6, 92.6)	Pos: 62 (38.6, 81.5)	0.6796	0.0012
			Neg: 78 (63.3, 88.0)	(0.5244–0.8348)
MCAA	36.8 (19.2, 59.0)	92.1 (79.2, 97.3)	Pos: 70 (39.7, 89.2)	0.6852	0.0009
			Neg: 74 (60.5, 84.8)	(0.5242–0.8182)
Plasminogen	42.1 (23.1, 63.7)	94.7 (82.7, 98.5)	Pos: 80 (49.0, 94.3)	0.7444	0.0001
			Neg: 77 (62.8, 86.4)	(0.5976–0.8912)
IL1 alpha	31.6 (15.4, 54.0)	81.6 (66.6, 90.8)	Pos: 46 (23.2, 70.9)	0.5582	0.2028
			Neg: 70 (55.8, 81.8)	(0.3958–0.7207)
IL17	84.2 (62.4, 94.5)	32.1 (17.9, 50.7)	Pos: 46 (30.5, 61.8)	0.6275	0.0129
			Neg: 75 (46.8, 91.1)	(0.4794–0.7757)
Collagen	10.5 (2.94, 31.4)	97.4 (86.5, 99.5)	Pos: 67 (20.8, 93.8)	0.6939	0.0006
			Neg: 69 (55.3, 79.3)	(0.5482–0.8396)

^astatistics by Open Source OpenEpi, Version 3: Diagnostic Test.

^bcompared to an AUC of 0.5, which is the null hypothesis (a test with no clinical value).

The data for both cytokines support the lack of association with progressive disease as shown in Figures 5 and 6, in that they have low predictive value and low ROC AUC values which were not statistically different from the ‘useless’ test AUC of 0.5.

Table 3 summarizes the results of testing done to evaluate the efficacy of these tests as potential screening tools for any LPT, whether progressive or stable. With the Stable disease group included, none of the tests had predictive value except for the collagen test, which while significantly different from the ‘useless’ test AUC, its positive and negative predictive tests were low, and its ROC AUC was less than 0.7.

Table 3.

Sensitivity/specificity data for any LPT (Progressive + Stable Groups).

	Sensitivity % (95% CI)	Specificity % (95% CI)	Predictive Value (%) (95% CI)	ROC Curve AUC (95% CI)	AUC p valueb
ANA	24.1 (14.9, 36.5)	89.7 (76.4, 95.9)	Pos: 78 (54.8, 91)	0.542	0.198
			Neg: 44 (33.9, 55.3)	(0.426–0.659)
MCAA	32.8 (22.1, 45.6)	76.9 (61.7, 87.4)	Pos: 68 (49.3, 82.1)	0.579	0.024
			Neg: 44 (32.4, 55.2)	(0.467–0.691)
Plasminogen	63.8 (50.9, 74.9)	64.1 (48.4, 77.3)	Pos: 73 (59.1, 82.9)	0.591	0.011
			Neg: 54 (40.2, 67.9)	(0.476–0.705)
IL1 alpha	22.8 (13.8, 35.2)	79.5 (64.5, 89.2)	Pos: 62 (40.9, 79.3)	0.515	0.642
			Neg: 41 (30.9, 52.6)	(0.397–0.632)
IL17	65.5 (52.7, 76.4)	48.7 (33.9, 63.8)	Pos: 66 (52.7, 76.4)	0.557	0.092
			Neg: 49 (33.9, 63.8)	(0.438–0.676)
Collagen	14.0 (7.3, 25.3)	84.6 (70.3, 92.8)	Pos: 57 (32.6, 78.6)	0.674	0.00002
			Neg: 40 (30.3, 51.1)	(0.557–0.790)

^astatistics by Open Source OpenEpi, Version 3: Diagnostic Test.

^bcompared to an AUC of 0.5, which is the null hypothesis (a test with no clinical value).

Since none of the tests alone were strongly predictive of progressive LPT with high sensitivity and specificity, calculations were performed to determine the sensitivity and specificity of 2-stage or simultaneous combination testing (Kanchanaraksa 2008). In two-stage combinations, the patient would be screened for ANA, followed by another test if the ANA test was positive. In simultaneous testing, both tests would be given at the same time, without knowing results of either test. Table 4 shows that for two-stage testing, addition of the second test reduces sensitivity but increases specificity to 99%, for both MCAA and anti-PLG as the second test. For simultaneous testing, addition of the second test improves sensitivity and somewhat reduces specificity, for both MCAA and anti-PLG as the second test with ANA testing.

Table 4.

Combination screening tests for sensitivity and specificity.

	Single test Sensitivity %	Specificity %	2-Stage Sensitivity %	Specificity %	Simultaneous Sensitivity %	Specificity %
ANA	52	84	19	99	70	78
MCAA	37	92
ANA	52	84	22	99	72	80
PLG	41	95

Discussion

Several studies have now linked LAA exposure with a significant risk for autoimmune outcomes, including a high frequency of positive ANA tests and a range of systemic autoimmune diseases (Marchand et al. 2012; Black et al. 2014; Pfau et al. 2014; Diegel et al. 2018). In addition, mesothelial cell autoantibodies (MCAA), including anti-plasminogen antibodies, have been demonstrated in mice and humans exposed to LAA (Serve et al. 2013; Hanson et al. 2016). In cell culture and in mice, these autoantibodies drive collagen production by mesothelial cells, leading to the hypothesis that they play a role in fibrosis (Serve et al. 2013; Gilmer et al. 2016; Hanson et al. 2016).

Although ANA and MCAA are associated with radiographic changes indicative of pleural disease (Marchand et al. 2012), the association was not strong, and there were many individuals with LPT on CT scans who did not have these autoantibodies. Through the years of working with the LAA-exposed population, the Center for Asbestos Related Diseases (CARD) in Libby, Montana, anecdotally noted that people with the most severe and progressive LPT also tended to have indices of inflammatory disorders, including pleuritis and arthritis with significant levels of pain. This led to our hypothesis that the autoantibodies might be more closely associated with a more progressive form of LPT, whereas LPT in the absence of autoantibodies might be more stable and less symptomatic. If that were the case, screening for autoantibodies might be a marker for progressive disease, leading eventually to a) early detection of the most severe and fatal outcome, and b) potential therapeutic strategies.

The CARD maintains an extensive database of the LAA-exposed population through their research program, now containing almost 7000 patient records. From this database, subjects were selected who met criteria to identify individuals with no evidence of LPT, patients with stable LPT and those with progressive LPT. All subjects in the database have had exposure to LAA, documented by exposure routes including occupational, residential and environmental exposures. Occupational exposures tend to be highest due to directly working with the ore, but household and environmental exposures can also be high due to the extensive use of the vermiculite ore in home insulation, gardening, and surfacing of ball fields, playgrounds, and roads (Noonan 2006; Noonan et al. 2015). Dust with high levels of LAA fibers was also brought home on worker clothing. In addition, studies have demonstrated health effects of LAA even at very low (environmental) exposure levels in both mice and humans (Black et al. 2014; Szeinuk et al. 2016; Pfau et al. 2017).

The data presented here confirm the association between autoantibodies (ANA, MCAA and anti-PLG) and progressive LPT, but not consistently with stable LPT, supporting our hypothesis. There was also an association between progressive LPT and the ability of serum antibodies to drive collagen matrix formation in vitro, which supports a mechanistic role for the antibodies as previously demonstrated (Serve et al. 2013; Gilmer et al. 2016). There was no association between the cytokines and either stable or progressive LPT. They may be markers of LAA-induced inflammation, which would require comparison with an unexposed cohort, not included in this study.

There were differences when the data were analyzed for males and females separately. The initial plan was to have equal numbers of males and females in the study, but during the chart reviews to select the Progressive LPT group, very few females met the criteria for this group. This suggests that females may be less susceptible to progression, more often having a stable form of disease. This could be related to their exposure or to some susceptibility factor of females. However, further studies would be needed to support this hypothesis.

Sensitivity and specificity testing revealed that while all three antibody tests (ANA, MCAA and anti-PLG) had moderate predictive value for progressive LPT, sensitivity was consistently low while specificity was fairly high. This suggests that negative tests for autoantibodies could be used to suggest a low risk for progressive LPT, but there would be frequent false-positive tests. This is expected since ANA testing is generally used to screen for systemic autoimmune diseases such as SLE and SSc. Positive ANA tests occur commonly even in healthy people, although usually at low titers, so they actually have limited predictive values for systemic autoimmune disease as well (Slater et al. 1996; Soto et al. 2015). However, it is well known that ANA can occur prior to clinical symptoms even in systemic autoimmune diseases (Arbuckle et al. 2003; Heinlen et al. 2007), such that ANA testing can be used for early diagnosis and prediction of progression to clinical disease (Choi et al. 2016). Therefore, because ANA testing is well standardized and had the highest sensitivity of the antibody tests, ANA testing with titer may be valuable in predicting progression of LPT in patients exposed to asbestos and experiencing evidence of pleural disease. Combination testing with two different autoantibody tests might have some predictive value, especially simultaneous testing for ANA and anti-PLG.

There are several weaknesses and challenges with this study. First, there were significant differences between the study groups with regard to smoking history (Table 1). There is no evidence in the literature suggesting a protective effect of smoking in terms of autoantibody production or autoimmune diseases, so that there is not a clear explanation as to why pack year history was so much lower in the Progressive LPT group. It might be possible that this group stopped smoking earlier in life due to respiratory problems. Smoking has not been shown to be a risk factor for pleural thickening (Yano et al. 1993), yet has been linked to a variety of autoimmune diseases (Costenbader and Karlson 2006). Therefore, it can at least be said that the smoking history in our groups did not contribute to the pleural disease.

Second, data for both males and females were included here, but it was noted that there were only limited numbers of females, making it difficult to interpret any differences seen between sexes. The reason for this was that only two females met all the criteria for the Progressive LPT group, so the other groups had to be matched to that. Physicians caring for these patients over the years have noted anecdotally that LPT appears rarely in women, suggesting that there may be a difference in susceptibility to this disease. The data for the women are therefore reported here, but with the caveat that the statistical analysis becomes questionable with such small numbers.

Third, the exposure data for LAA in Libby are based on matrices developed for this population (Noonan et al. 2015), as described above. The data have limitations due to estimations based on self-reported years involved in different exposure pathways, little data available on the actual exposures that occurred in the different exposure pathways, and little data available on exposures from raw ore (usually occupational) compared to processed ore (usually environmental). It is known, however, that some of the occupational exposures reported were very low (people working in offices at the plant or delivering supplies to the mine were considered occupational) and some of the environmental exposures were extremely high (playing in ore piles, insulating with the vermiculite, etc). Therefore, within the different exposure types (occupational, household, environmental), there are potentially very large ranges of exposure, and it is possible that differences in exposure among the groups were larger than our calculations indicate. Nevertheless, it was not the purpose of this study to evaluate exposure, but rather to evaluate markers of autoimmunity in people with pleural disease.

Finally, a control group with no LAA exposure is not available in Libby, and to our knowledge, Libby is the only place where Progressive LPT is being evaluated. However, it is notable that the frequency of positive ANA tests in the No Disease and Stable groups is similar to that reported for the United States general population for this age group, at around 15–18% (Satoh et al. 2012).

At this time, there are no data regarding when patients develop LPT in relation to being seropositive for antibodies. Therefore, it cannot yet be determined if these tests could predict progressive LPT before it begins to develop. That would require serial antibody testing, including specific antibody profiles that are important in the prediction of SAID such as dsDNA and extractible nuclear antigens (ENA) and that occur in LAA-exposed individuals (Pfau et al. 2018), in the exposed population while also screening for LPT. This is now being done at the CARD clinic. Nevertheless, based on these results, which support an inflammatory and autoimmune component to progressive LPT, several potential therapeutic strategies could be considered in the future if these data are further supported and expanded.

In the future, ANA screening will continue for the LAA-exposed population, and the clinical course of disease will be tracked, including both LPT and systemic autoimmune diseases. As mentioned above, a diagnosis of SAID was not an exclusion criterion for this study, since a risk for autoimmune disease is part of our hypothesis. However, moving forward, autoimmune diagnoses must be considered and should be evaluated using a validated screening questionnaire and detailed tracking of autoimmune symptomology. It will be important to determine whether positive ANA tests are associated with future development of autoimmune disease, whether that disease is a systemic or localized autoimmune disease, and whether the LPT changes are part of that. Testing will be performed for all individuals in the screening program, but noting that a positive test only has a predictive value for progressive LPT if the patient also has evidence of LPT. This is because positive ANA tests are also predictive for SAID. One of the purposes of this paper is to encourage other health care centers and screening agencies to consider adding ANA testing for any people with suspected exposure to asbestos or asbestiform fiber-containing materials, since it might be of value for predicting both SAID and progressive LPT, since asbestos/asbestiform fiber exposure is associated with both (Pfau 2018). Long-term, this approach, of broadly screening at-risk populations, will provide valuable information about its predictive value for both diseases (SAID and LPT), the sequence of antibody seroconversion versus development of disease, and allow early treatment with anti-inflammatory modalities to try to slow progression. Continued study of LPT and its fatal progressive clinical presentation will be essential moving forward due to a) on-going exposures in the Libby area Superfund sites of Montana, b) millions of homes throughout the United States that remain insulated with LAA through the use of Zonolite vermiculite insulation, and c) environmental release of similar amphibole fibers in areas of Arizona and Nevada where millions of people live and recreate in the warm and arid Southwest U.S. where amphibole-containing dusts are causing serious public health concerns (Ewing et al. 2010; Pfau et al. 2017; Wolfe et al. 2017).

Summary.

In summary, the results demonstrate that:

• Positive ANA tests were associated with Progressive LPT. ANA testing, even in this small subject group, may have value as a screening tool for risk of severe clinical disease in people with LPT.

• MCAA and PLG testing also predicted Progressive LPT. This was true only in males, but females represented a very small subset of the total subjects. Combining MCAA or PLG testing with ANA screening may be of value due to high specificity, but sensitivity remains low with either 2-stage or simultaneous testing.

• Serum cytokine testing had no value in predicting Progressive LPT.

• No marker had any value in predicting Stable LPT or Stable/Progressive LPT.

Abbreviations:

ANA

Antinuclear Autoantibodies

ARD

Asbestos-related Diseases

AUC

Area Under the Curve (used for ROC testing)

CARD

Center for Asbestos Related Diseases, Libby MT

CT

Computed Tomography

LAA

Libby Asbestiform Amphibole, aka Libby Amphibole, LA

LERP

Libby Epidemiology Research Program (funded by the ATSDR)

LPT

Lamellar Pleural Thickening

MCAA

Mesothelial Cell Autoantibodies

PFT

Pulmonary Function Test

PLG

Plasminogen

ROC

Receiver Operator Curves (testing for sensitivity and specificity)

SGRQ

St. George’s Respiratory Questionnaire

SLE

Systemic Lupus Erythematosus

SSc

Systemic Sclerosis, Scleroderma