Immunogenetic risk and protective factors for the development of L-tryptophan–associated eosinophilia–myalgia syndrome and associated symptoms

Authors


Abstract

Objective

To assess L-tryptophan (LT) dose, age, sex, and immunogenetic markers as possible risk or protective factors for the development of LT-associated eosinophilia–myalgia syndrome (EMS) and related clinical findings.

Methods

HLA–DRB1 and DQA1 allele typing and Gm/Km phenotyping were performed on a cohort of 94 white subjects with documented LT ingestion and standardized evaluations. Multivariate analyses compared LT dose, age, sex, and alleles among groups of subjects who ingested LT and subsequently developed surveillance criteria for EMS, developed EMS or characteristic features of EMS (EMS spectrum disorder), or developed no features of EMS (unaffected).

Results

Considering all sources of LT, higher LT dose (odds ratio [OR] 1.4, 95% confidence interval [95% CI] 1.1–1.8), age >45 years (OR 3.0, 95% CI 1.0–8.8), and HLA–DRB1*03 (OR 3.9, 95% CI 1.2–15.2), DRB1*04 (OR 3.9, 95% CI 1.1–16.4), and DQA1*0601 (OR 13.7, 95% CI 1.3–1.8) were risk factors for the development of EMS, whereas DRB1*07 (OR 0.12, 95% CI 0.02–0.48) and DQA1*0501 (OR 0.23, 95% CI 0.05–0.85) were protective. Similar risk and protective factors were seen for developing EMS following ingestion of implicated LT, except that DRB1*03 was not a risk factor and DQA1*0201 was an additional protective factor. EMS spectrum disorder also showed similar findings, but with DRB1*04 being a risk factor and DRB1*07 and DQA1*0201 being protective. There were no differences in sex distribution, Gm/Km allotypes, or Gm/Km phenotypes among any groups.

Conclusion

In addition to the xenobiotic dose and subject age, polymorphisms in immune response genes may underlie the development of certain xenobiotic-induced immune-mediated disorders, and these findings may have implications for future related epidemics.

INTRODUCTION

An epidemic outbreak of an unusual condition characterized by an acute onset of incapacitating myalgia and peripheral eosinophilia (eosinophil count of ≥109/liter without known causes for eosinophilia) and termed the eosinophilia–myalgia syndrome (EMS) was observed in 1989 (1). Several epidemiologic studies suggested that most patients who developed EMS in late 1989 had consumed L-tryptophan (LT) produced by a single manufacturer (1–3). Soon thereafter, an import alert by the US Food and Drug Administration (FDA) effectively removed LT from the US market and the epidemic resolved.

Previous studies indicated that in addition to the source of LT, an increased dose of LT, increased age, and use of LT as a sleeping aid were risk factors in the development of EMS (3, 4). Methodologic limitations in epidemiologic studies and the inability to identify the exact pathogenic components of the implicated batches of LT have hampered the understanding of disease pathogenesis (5, 6). Moreover, sporadic cases meeting EMS surveillance criteria existed prior to 1989, continued after 1989, and have been reported in persons who did not take LT (7, 8).

Because only a fraction of persons who ingested implicated batches of LT developed the disease, additional factors likely played pathogenetic roles. Few investigations, however, have assessed genetic susceptibility for EMS (9, 10). In hopes of understanding the additional risk and protective factors that may have implications for future similar epidemics, we explored the possible role of LT dose, age, sex, and immunogenetic markers for susceptibility in the development of LT-associated EMS and associated clinical findings in a unique cohort in which LT intake and clinical outcomes were carefully documented.

SUBJECTS AND METHODS

Subjects.

Blood samples were available from 96 white subjects who were previously studied as part of a single physician's practice (3) and were enrolled into protocols approved by the Centers for Disease Control and Prevention and the FDA. All 96 subjects had ingested LT and 60 of them took implicated LT. Two subjects were excluded to avoid genetic confounding because they were blood relatives of other subjects. The remaining 94 unrelated LT users were divided into 3 groups based on outcomes (3): 1) EMS (n = 28), defined as those who developed surveillance criteria for EMS (incapacitating myalgia and a peripheral eosinophil count of ≥109/liter without known causes for eosinophilia); 2) EMS spectrum disorder (n = 57), defined as those with EMS, those who had an eosinophil count of ≥109/liter but without incapacitating myalgias, or those who developed more than 1 symptom characteristic of EMS (myalgias, arthralgias, neuropathy, alopecia, or skin thickening); and 3) unaffected (n = 37), defined as those who took LT and did not develop any of the above. In extension studies, those who developed specific signs and symptoms found in EMS subjects after LT ingestion, including incapacitating myalgia (n = 34), muscle weakness (n = 22), increased skin thickness (n = 11), or numbness (n = 28), were assessed, and healthy white controls (n = 872) enrolled in studies of the pathogenesis of autoimmune diseases at the National Institutes of Health were also studied as another comparison group.

Methods.

Genomic DNA was isolated from peripheral blood and amplified by polymerase chain reactions using primers and sequence-specific oligonucleotide probes as previously described (11) to identify 34 HLA–DRB1 and 8 DQA1 alleles. Gm/Km allotypes and phenotypes, genetic markers of γ and κ light chains, respectively, which regulate immunoglobulin production and are known risk factors for several immune-mediated disorders, were determined as previously described (12).

Statistical analyses.

The primary study assessed variables potentially influencing susceptibility to develop EMS or EMS spectrum disorder following LT ingestion from any source and following ingestion of implicated LT. Each potential predictor variable (sex, LT dose, age, HLA alleles, and Gm/Km allotypes and phenotypes) was analyzed using Fisher's exact test, and odds ratios (ORs) and 95% confidence intervals (95% CIs) were determined. Univariate analyses were performed for each outcome to identify polymorphisms associated with the risk of the specific outcome. Results of the univariate analyses were used to develop a multivariate model of potential predictors of outcome following LT ingestion.

We performed multivariate analyses to minimize possible bias due to confounding effects by the dose of LT, the age of the subjects, and sex (13). We used logistic regression models with a backward stepwise analysis in 3 ways. First, all variables, including HLA alleles, dose of LT, age of the subjects, and sex, were entered simultaneously into a model followed by a stepwise analysis. Second, HLA alleles were forcedly entered into a model and other variables were included for a stepwise analysis. Also, LT dose was forcedly entered into a model, followed by a stepwise analysis for other variables. In each analysis, conditional stepwise methods and stepwise methods based on likelihood ratios were performed. After identifying variables that remained in the model, the penalized maximum likelihood estimate was used for bias correction to determine the OR and 95% CI (14, 15).

Comparisons were also made between EMS or EMS spectrum disorder and normal controls without multivariate analysis. Whether the homozygosity of any associated alleles altered the risk was also studied. SAS (SAS Institute, Cary, NC), SPSS, version 16.0j (SPSS, Tokyo, Japan), and LogXact8 (Cytel, Cambridge, MA) were used for these statistical analyses. Power analyses were performed using StatXact (Cytel) to determine the power to detect a significant difference (2-tailed P values less than 0.05) between the groups being compared (unconditional difference).

RESULTS

As previously described (3), the dose of LT ingested regardless of the LT source was a risk factor for the development of EMS, and it was also found to be a risk factor for EMS spectrum disorder (Table 1). The dose of LT, however, was not significantly different in those who did or did not develop incapacitating myalgia (Table 1) or muscle weakness (data not shown). Although the overall ages were not significantly different among any groups, an analysis of age distributions by decade showed that those who were age >45 years were more likely to develop EMS or EMS spectrum disorder, whereas those who were age >29 years were more likely to have incapacitating myalgia (Table 1). Sex was not found to be a risk factor in any analyses.

Table 1. Differences in the mean dosage of LT and age distributions in LT users with different outcomes*
 EMSEMSSDMyalgia
LTA (n = 28)LTU (n = 37)LTA vs. LTU, OR (95% CI)LTA (n = 57)LTU (n = 37)LTA vs. LTU, OR (95% CI)LTA (n = 34)LTU (n = 32)LTA vs. LTU, OR (95% CI)
  • *

    LT = L-tryptophan; EMS = eosinophilia–myalgia syndrome; LTA = LT-affected group; LTU = LT-unaffected group; OR = odds ratio; 95% CI = 95% confidence interval; EMSSD = EMS spectrum disorder (EMS or characteristic features of EMS); NS = not significant.

  • Significantly different in those who took only implicated LT (total: n = 60, EMS: n = 27, EMSSD: n = 43, myalgia: n = 26, unaffected: n = 18).

Mean LT dosage, mg/day4,160.72,898.71.35 (1.05–1.79)3,965.82,898.71.39 (1.12–1.80)3,648.53,550.0NS
Mean LT dosage, mg/day4,166.72,777.81.43 (1.02–2.22)4,145.42,777.81.58 (1.11–2.46)3,884.64,350.0NS
Age >45 years, %70.344.13.01 (1.03–8.75)66.144.12.47 (1.03–5.91)   
Age >45 years, %69.244.42.81 (0.81–9.80)69.044.42.79 (0.90–8.69)   
Age >29 years, %      10083.314.45 (1.53–1,930)

Significant differences were noted in the frequencies of certain HLA alleles in LT users who developed EMS compared with those who were unaffected by multivariate analyses incorporating LT dose and age (Table 2). There were no significant differences noted in these associations whether conditional stepwise methods or stepwise methods based on likelihood ratios were used. Multivariate analyses showed that HLA–DRB1*03, DRB1*04, and DQA1*0601 were risk factors because those alleles were more frequently seen in subjects who developed EMS than in unaffected subjects. HLA–DRB1*07 and DQA1*0501, however, appeared to be protective factors because they were found more frequently in unaffected subjects compared with those who developed EMS. An analysis of a subset of subjects with EMS and unaffected subjects who took implicated LT showed that DRB1*04 was again a risk factor and DRB1*07 and DQA1*0501 were protective for the development of EMS, but that DQA1*0201 was an additional protective allele (Table 2). Because the most common allele in the DRB1*07 supertype (DRB1*0701) is in linkage disequilibrium with DQA1*0201, we analyzed the frequencies of the presumptive haplotype DRB1*07–DQA1*0201 among the study groups. The frequency of this presumptive haplotype was found to be significantly lower in EMS subjects (13.0%) compared with unaffected subjects (57.1%; OR 0.11, 95% CI 0.02–0.77) among those who took implicated LT.

Table 2. Differences in HLA types in LT users who did and did not develop EMS and healthy controls*
HLA alleleLTA with EMS (n = 28), %LTU without EMS (n = 37), %Healthy controls (n = 872), %LTA vs. LTU, OR (95% CI)LTA vs. LTU, % powerLTA vs. healthy controls, OR (95% CI)LTA vs. healthy controls, % power
  • *

    See Table 1 for abbreviations.

  • OR obtained by multivariate analysis.

  • Power to detect a significant difference (2-tailed P < 0.05) between the groups being compared.

  • §

    Significantly different in those who took only implicated LT (total: n = 60, EMS: n = 27, unaffected: n = 18).

  • Protective factors.

  • #

    Protective factors. Significantly different in those who took only implicated LT (total: n = 60, EMS: n = 27, unaffected: n = 18).

  • **

    This power calculation uses the proportion of DQA1*0601 in the LTU as 0.000001.

HLA–DRB1       
 *0339.324.321.93.89 (1.15–15.20)25.32.31 (1.06–5.02)37.8
 *0432.116.228.13.93 (1.08–16.37)33.6NS2.7
 *04§33.3§5.6§28.17.52 (1.30–86.05)§60.0NS3.2
 *0714.340.525.00.12 (0.02–0.48)64.7NS5.9
 *07#14.3#50.0#20#0.12 (0.02–0.48)#73.7NS4.1
HLA–DQA1       
 *0201#17.4#57.1#24.00.08 (0.003–0.64)#56.9NS0.9
 *050126.150.040.20.23 (0.05–0.85)49.0NS7.6
 *0501#26.1#71.4#40.20.008 (0.01–0.36)#59.4NS7.6
 *060121.700.713.74 (1.32–1,874)76.8**38.61 (9.56–155.96)97.4
 *0601§21.7§00.7§NS1.6**38.61 (9.56–155.96)§97.4

Given concerns that the surveillance criteria for EMS may not have included all persons that had been affected by ingesting LT, we analyzed additional cases that had characteristic features of EMS, which together with EMS were defined as EMS spectrum disorder. As was the case with EMS, DRB1*04 was a risk factor and DRB1*07 and DQA1*0201 were protective for the development of EMS spectrum disorder when implicated sources of LT were considered (Table 3). In contrast to EMS, however, DQA1*0501 was not a risk factor for EMS spectrum disorder in those who ingested implicated LT. Given the small sample sizes for some groups, however, the power to detect risk or protective factors is relatively low in some cases (Table 3), and this might explain some of the differences noted between risk factors for those who ingested any form of LT and those who took implicated LT. We analyzed the frequencies of the presumptive haplotype DRB1*07–DQA1*0201 among the study groups and again found that the frequency of this haplotype was significantly less frequent in subjects with EMS spectrum disorder (13.5%) compared with unaffected subjects (57.1%; OR 0.15, 95% CI 0.02–0.93) among those who took implicated LT.

Table 3. Differences in HLA types in LT users who did or did not develop EMSSD and healthy controls*
HLA alleleLTA with EMSSD (n = 57), %LTU without EMSSD (n = 37), %Healthy controls (n = 872), %LTA vs. LTU, OR (95% CI)LTA vs. LTU, % powerLTA vs. healthy controls, OR (95% CI)LTA vs. healthy controls, % power
  • *

    See Table 1 for abbreviations.

  • OR obtained by multivariate analysis.

  • Power to detect a significant difference (2-tailed P < 0.05) between the groups being compared.

  • §

    Significantly different in those who took only implicated LT (total: n = 60, EMSSD: n = 43, unaffected: n = 18).

  • Protective factors. Significantly different in those who took only implicated LT (total: n = 60, EMSSD: n = 43, unaffected: n = 18).

HLA–DRB1       
 *0338.624.321.9NS28.82.24 (1.29–3.92)64.3
 *03§39.5§22.221.9§NS24.72.33 (1.24–4.39)§61.2
 *04§30.2§5.6§28.16.30 (1.20–67.09)§58.8NS2.8
 *0716.750.025.00.13 (0.03–0.50)76.1NS19.9
HLA–DQA1       
 *020113.557.124.00.07 (0.004–0.56)27.2NS17.7
 *060110.000.7NS23.015.44 (4.01–59.54)93.7
 *0601§13.5§    00.7§NS    021.72 (5.56–84.81)§94.1

Because of the relatively small sample sizes in our primary study cohort, we next asked if there was bias in the frequencies of HLA alleles between white subjects with EMS or EMS spectrum disorder and white healthy controls. DRB1*03 was significantly more frequent in those with EMS who took any source of LT than in healthy controls (Table 2). Regardless of the source of LT, DQA1*0601 was more frequent in EMS subjects than in healthy controls. Moreover, regardless of the source of LT, DRB1*03 and DQA1*0601 were also significantly more frequent in those with EMS spectrum disorder than in healthy controls. Significant differences in allele frequencies were not observed for DRB1*04, DRB1*07, DQA1*0201, and DQA1*0501 in those with EMS in comparison with healthy controls, although these alleles were observed in different frequencies among those with different outcomes following LT ingestion.

Since little is known about risk factors for the development of specific signs and symptoms found in subjects with EMS, we next asked if there was any bias in the frequencies of HLA alleles between subjects with certain characteristic EMS symptoms, including incapacitating myalgia, muscle weakness, increased skin thickness, or numbness. Among the subjects who took LT from any source, DRB1*07 was protective for developing incapacitating myalgia, and among subjects who took implicated LT, DQA1*0501 was protective for developing incapacitating myalgia (Table 4). Furthermore, HLA–DQA1*0101 was a risk factor for developing muscle weakness among subjects who took LT from any source, whereas DQA1*0501 was protective among subjects who took implicated LT (Table 5). No DQA1 or DRB1 alleles, however, were found to be risk or protective factors for the development of increased skin thickness or numbness (data not shown). The frequencies of DQA1*0601 were again significantly higher in those with incapacitating myalgia or muscle weakness who took implicated LT compared with normal controls.

Table 4. Differences in HLA types in LT users who did or did not develop myalgia and healthy controls*
HLA alleleLTA with myalgia (n = 34), %LTU without myalgia (n = 32), %Healthy controls (n = 872), %LTA vs. LTU, OR (95% CI)LTA vs. LTU, % powerLTA vs. healthy controls, OR (95% CI)LTA vs. healthy controls, % power
  • *

    LT = L-tryptophan; LTA = LT-affected group; LTU = LT-unaffected group; OR = odds ratio; 95% CI = 95% confidence interval; NS = not significant.

  • OR obtained by multivariate analysis.

  • Power to detect a significant difference (2-tailed P < 0.05) between the groups being compared.

  • §

    Protective factors.

  • Protective factors. Significantly different in those who took only implicated LT (myalgia: n = 26, no myalgia: n = 10).

  • #

    Significantly different in those who took only implicated LT (myalgia: n = 26, no myalgia: n = 10).

HLA–DRB1       
 *07§25.9§46.7§25.0§0.19 (0.04–0.89)§41.9NS0.8
HLA–DQA1       
 *050126.980.040.20.09 (0.02–0.54)92.9NS7.4
 *06018.86.70.7NS3.813.5 (2.89–62.86)73.4
 *0601#11.5#20.20.7#NS11.#16.59 (3.53–77.97)#82.0
Table 5. Differences in HLA types in LT users who did or did not develop muscle weakness and healthy controls*
HLA alleleLTA with muscle weakness (n = 22), %LTU without muscle weakness (n = 41), %Healthy controls (n = 872), %LTA vs. LTU, OR (95% CI)LTA vs. LTU, % powerLTA vs. healthy controls, OR (95% CI)LTA vs. healthy controls, % power
  • *

    See Table 4 for definitions.

  • OR obtained by multivariate analysis.

  • Power to detect a significant difference (2-tailed P < 0.05) between the groups being compared.

  • ‡, §

    Protective factors. Significantly different in those who took only implicated LT (muscle weakness: n = 14, no muscle weakness: n = 18).

  • Significantly different in those who took only implicated LT (muscle weakness: n = 14, no muscle weakness: n = 18).

HLA–DQA1       
 *010131.87.727.57.22 (1.26–76.66)69.1NS2.6
 *0501§14.3§72.2§40.2§0.06 (0.01–0.40)§93.90.21 (0.05–0.94)§12.5
 *06019.12.40.7NS17.89.55 (1.81–50.29)64.5
 *060114.35.60.7NS13.020.53 (3.47–121.56)73.1

The homozygosity of any of the alleles noted above was not found to play a significant role in developing EMS or EMS spectrum disorder. Finally, there were no significant differences in Gm/Km allotypes or phenotypes among any of the groups.

DISCUSSION

LT-associated EMS is a heterogeneous group of connective tissue disorders that share the common features of eosinophilia and incapacitating myalgia, and that transiently reached epidemic proportions in late 1989. Despite intense research, the specific components of LT responsible for and the mechanisms causing LT-associated EMS remain unknown. Nonetheless, clinical, serologic, and pathologic findings suggest that immune mechanisms play a role (16). Further evidence for the role of the immune system in the pathogenesis of EMS includes the presence of chronic inflammatory infiltrates in the skin, muscle, and nerves, characterized by activated T lymphocytes, macrophages, eosinophils, and fibroblasts (16), and the response of some patients to immunosuppressive therapies (17).

Because of the low attack rates, prior studies assessed LT-associated EMS for possible risk factors. Previous investigations did find that LT dose and age were risk factors, but these could not explain the entire risk (3). Other investigations assessed immunogenetics and compared LT-associated EMS cases with normal controls that likely did not take LT as a specific agent, and found that HLA–DR4 (9) and certain cytochrome P450 polymorphisms (10) were possibly associated with EMS. In this study, we performed analyses including LT dose, age, sex, and immunogenetics in comparisons of subjects who ingested LT and either developed features of EMS or did not. We did not find an influence of sex, but we did find that LT dose, age, and HLA–DRB1*04 were risk factors for the development of LT-associated EMS or EMS spectrum disorder. Nonetheless, we also identified other immunogenetic risk and protective factors. For example, DRB1*03 and DQA1*0601 were risk factors for the development of EMS following the ingestion of LT from any source. This finding was further supported by comparison of the allele distributions in EMS with those in healthy controls. Not all associations of other alleles comparing LT-affected groups with LT-unaffected subjects, however, were replicated when comparing LT-affected groups with healthy controls, and when assessing groups of subjects who took implicated LT and those taking LT from any source. The reasons for these differences, as well as the variable findings between this study and other investigations, remain unclear. Nonetheless, the varying study designs, different populations investigated, variable numbers of subjects in the groups, and the resulting different powers to detect risk or protective factors (Tables 2, 3, 4, and 5) may account for some of the variable results. We also do not know why age appears to be a risk factor for the development of EMS and EMS spectrum disorder, although several lines of evidence suggest that age and aging of the immune system are implicated in the development of a number of autoimmune diseases (18, 19).

The findings in our investigation (Table 6) and in prior studies showing that LT dose, age, and certain metabolizer and immune response genes are associated with the development of EMS support the hypothesis that the dose of LT, the age of the subject, and both metabolic and immune-mediated mechanisms may be important for the pathogenesis of EMS. Although the reasons for the genetic associations remain unknown, some of the associated genes such as DRB1*03 have been linked to many other immune-mediated conditions via multiple mechanisms, including a favored Th2 profile (20). It is interesting that Th2 profiles also appear to predominate during in vitro responses to possible contaminants in implicated LT (21) and that lung specimens from a related disorder called toxic oil syndrome, which developed in Spain from 1981–1983 after ingestion of contaminated rapeseed oil, appear to support a Th2 mechanism in that disease (22). Additionally, genetic studies in toxic oil syndrome implicated certain HLA–A, B, DR4, and DQA alleles as possible risk and protective factors, implying that polymorphisms in immune response genes may underlie the development of multiple xenobiotic-induced immune-mediated disorders and that these findings may have implications for future related epidemics (23).

Table 6. Summary of risk and protective factors for different syndromes following LT ingestion in this study*
SyndromeRisk factorsProtective factors
  • *

    LT = L-tryptophan; EMS = eosinophilia–myalgia syndrome.

EMS developing after ingestion of LT from multiple manufacturersLT doseDRB1*07
 Age >45 yearsDQA1*0501
 DRB1*03 
 DRB1*04 
 DQA1*0601 
EMS developing after ingestion of LT from a single implicated manufacturerLT dose Age >45 years DRB1*04DRB1*07 DQA1*0201 DQA1*0501
EMS spectrum disorder developing after ingestion of LT from multiple manufacturersLT dose Age >45 yearsNone
EMS spectrum disorder developing after ingestion of LT from a single implicated manufacturerLT dose Age > 45 years DRB1*04DRB1*07 DQA1*0201
Myalgia developing after ingestion of LT from multiple manufacturersAge >29 years No HLA allelesDRB1*07
Myalgia developing after ingestion of LT from a single implicated manufacturerNoneDQA1*0501
Muscle weakness developing after ingestion of LT from multiple manufacturersDQA1*0101None
Muscle weakness developing after ingestion of LT from a single implicated manufacturerNoneDQA1*0501

EMS may present in a variety of ways and with different symptoms, including incapacitating myalgia, muscle weakness, skin thickening, and peripheral neuropathy. Our findings suggest that different immune response genes in LT users appear to be associated with the development of myalgia and eosinophilia (EMS) compared with the development of the other manifestations of the LT-associated EMS spectrum disorder. Therefore, an individual's genetic background may explain some of the variability in the development of specific manifestations following LT ingestion.

Why differences in genetic risk and protective factors were found in EMS cases that developed after ingestion of LT from any source compared with cases developing after exposure to implicated batches of LT remains unknown, although the relatively small sample sizes available for this study and the lack of any additional subjects to study limit our capacity to more fully assess these data at this time. Given the limited understanding of the pathogenesis of EMS, and since EMS has developed in persons who did not take LT as well as in persons taking LT from different manufacturers, it is possible that there are many potential mechanisms for the development of this syndrome. It is not clear whether the genetic risk and protective factors described in this investigation also play a role in EMS cases that have developed without LT exposure. Further studies of such cases are needed to understand if a variety of gene–environment interactions may underlie the evolution of eosinophilia and myalgia in different individuals.

AUTHOR CONTRIBUTIONS

All authors were involved in contributions to study conception and design, acquisition of data, or analysis and interpretation of data, and drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Miller had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Acknowledgements

The authors thank Drs. Ejaz Shamim and Terrance O'Hanlon for technical support, Dr. Elizabeth Sullivan for clinical assistance, and Drs. Richard Calvert and Sharon Adams for helpful discussions regarding the manuscript.

Ancillary