TGF-β1 Variants in Chronic Beryllium Disease and Sarcoidosis¹

Chronic beryllium disease (CBD)⁴ is a granulomatous lung disease caused by a known exposure to inhaled beryllium dust. It is characterized by the formation of an Ag-specific, oligoclonal CD4⁺ T cell-mediated immune response to beryllium which is MHC class II restricted (1). Individuals with CBD display granulomatous inflammation evidenced by noncaseating pulmonary granulomas and demonstrate sensitivity to beryllium based on the beryllium lymphocyte proliferation test. The sensitivity to beryllium is genetically determined, associated with an HLA class II epitope, primarily a glutamic acid at amino acid position 69 (Glu⁶⁹) in HLA-DPB1, and to a lesser extent a glutamic acid at amino acid position 71 in HLA-DRB1 (2). Genetic factors important in disease susceptibility and severity are not well-understood.

Sarcoidosis is a systemic granulomatous disorder of unknown etiology that primarily affects the lungs and lymph nodes. As in CBD, the immune response is T cell driven but in response to an unknown trigger or triggers (3). There are data indicating that distinct disease subsets are genetically determined, in large part associated with a specific MHC allele (4, 5, 6), and evidence exists of oligoclonality of T cell responses (7, 8, 9). Taken together, the MHC associations and oligoclonality would support the concept of an Ag-driven response, as with beryllium, even though the inciting agent or agents have yet to be identified. As with CBD, sarcoidosis is likely a multigenetic disease with prime targets including factors affecting the immune response.

TGF-β1 is a multifunctional cytokine with important effects on cell growth and differentiation, fibrosis, and a recently recognized role in immunomodulation. Several functional single nucleotide polymorphisms (SNPs) within the TGF-β1 gene have been identified including C−509T, codon 10 (T + 29C), and codon 25 (G + 74C) (10). The variant alleles of C−509T and codon 10 are associated with higher TGF-β1 protein levels in serum. Also, the codon 10 variant is associated with increased mRNA levels in PBMC (11, 12) while the wild-type allele of the G + 74C SNP leads to higher TGF-β1 levels in vitro (13).

High TGF-β1 protein production has been associated with several diseases including pulmonary sarcoidosis (14). Numerous studies have demonstrated that both CBD and sarcoidosis have genetic susceptibility factors (2, 4, 5, 15). Given the important roles that TGF-β1 plays in extracellular matrix production and immunomodulation and the clinical, pathologic, and immunologic similarities between CBD and sarcoidosis, we have hypothesized that similar genetic factors within the TGF-β1 gene will be important in the pathogenesis of these two noninfectious granulomatous diseases. Furthermore, we hypothesize that TGF-β1 polymorphisms associated with a more robust immune response will be associated with CBD and sarcoidosis cases compared with controls and will be associated with markers of disease severity.

A case-control study was conducted using two populations: sarcoidosis cases and their matched controls and CBD cases and their frequency matched controls. Case-comparison studies were performed to investigate similarities and differences of genotype and haplotype frequencies between the two diseases and to define associations between the polymorphisms under study and severity of each disease, defined using pulmonary function and chest radiographic indices.

Sarcoidosis cases (n = 198) were taken from the ACCESS (A Case Control Etiologic Study of Sarcoidosis) cohort; all met the American Thoracic Society (ATS)/European Respiratory Society (ERS) criteria for the diagnosis of sarcoidosis (16) with the additional criteria of biopsy confirmation of disease, exclusion of other granulomatous diseases, and disease onset within the preceding 6 mo (17). Sarcoidosis cases with previous exposure to beryllium were excluded unless they had a negative blood beryllium lymphocyte proliferation test (blood BeLPT). These 198 cases were chosen from the larger ACCESS cohort to provide similar gender, race, and geographic location as the CBD cases. Controls (n = 198) were matched to sarcoidosis cases based on age, gender, race, and geographic area and obtained by random digit dialing in each recruiting center area (17). Collected data relevant to our study were gathered from both cases and controls and included demographic information, clinical features chest roentgenography, spirometry, and medication.

CBD cases (n = 130) and controls (n = 135) were enrolled from cases evaluated in the outpatient Occupational and Environmental Medicine Clinic at National Jewish Medical and Research Center for abnormal respiratory symptoms or referral from a workplace surveillance program. CBD cases were enrolled if they demonstrated sensitization by two abnormal blood BeLPTs or one abnormal bronchoalveolar (BAL) BeLPT and noncaseating granulomas on biopsy (2). CBD control subjects had documented exposure to beryllium and at least one normal blood BeLPT. Controls were frequency matched to CBD cases based on age, gender, race, industry, and geographic location. CBD cases and controls completed questionnaires outlining demographics, exposure, and medical history. Cases also completed clinical evaluations as outlined below.

Clinical evaluations were completed on initial evaluation (baseline) and during follow-up over time on the cases and included pulmonary function testing, exercise testing, and chest radiography. The forced expiratory volume in 1 s (FEV₁₎ and forced expiratory vital capacity (FVC) were measured with a pneumotachograph. Total lung capacity (TLC) was measured in a constant pressure body plethysmograph. The single-breath method of Ogilvie et al. (18) was used to evaluate the diffusion capacity of carbon monoxide (DL_CO). Gas exchange and maximal exercise capacity were determined with a 380 B cycle ergometer (Siemens-Elema) with continuous cardiac rhythm and arterial oxygen content monitoring (19). An indwelling arterial line measured arterial blood gases at rest and after each minute of exercise. Results are reported as the arterial partial pressure of oxygen, PaO₂, at rest (PaO₂r) and during maximal exercise (PaO₂m). There were 96 CBD cases that had on average 4 visits (range 1–14); others had not had additional follow up after diagnosis, as they were lost to follow-up or recently diagnosed. Chest radiography is less sensitive than physiology in assessing disease severity and is often normal (20). Chest radiography results were available on n = 74 subjects at baseline. We evaluated the association between genotype and the presence of an abnormal chest radiograph using the International Labor Organization classification scheme (21).

Data available from the ACCESS study included spirometry (FEV₁ and FVC), chest radiography, evaluated using the Scadding chest radiographic staging system (22). Scadding stage is a method of describing lung involvement in sarcoidosis: stage 0 represents no abnormality detected in the lung (least severe), stage I is bilateral hilar or right paratracheal lymph node enlargement with clear lung fields, stage II is lymph node enlargement accompanied by pulmonary infiltration, stage III is pulmonary infiltration with no lymph node enlargement, and stage IV is evidence of pulmonary fibrosis with retraction of the hilar areas cephalad, cystic changes, and bullae (most severe).

Genomic DNA from the sarcoidosis and CBD cases and controls was isolated from PBMC as previously described (2, 4). PCR conditions previously described (23) were used to determine genotypes and haplotypes for two promoter polymorphisms, −509 C/T (rs1800469) and −800 G/A (rs1800468), and for three codons, 10 T/C (rs1982073), 25 G/C (rs1800471), and 263 C/T (rs1800472). The primer sequences were previously published (24). Primers were run at a final concentration of 20 ng/μl with an expected PCR product ranging from 222 to 514 bp.

Genotype frequencies in the control populations at each polymorphism were evaluated for departures from Hardy Weinberg equilibrium, using a χ² goodness-of-fit test (p < 0.001). For frequency matched subjects, tests for allele and genotype differences were conducted in the context of logistic regression while adjusting for gender and age at diagnosis or blood draw stratified by race. Tests using the matched subjects were conducted using matched logistic regression. Linear regression was used to assess the influence of allele and genotype on cross-sectional continuous severity outcomes, while mixed effects models were used to make assessments for longitudinal outcomes. Normalizing transformations were performed on continuous outcome variables when necessary to better approximate model assumptions. Age at test, gender, race, smoking history, and height (for lung function test variables), along with time from first exposure for CBD were included in the model along with the allele or genotype of interest. Clinical data were included in the longitudinal analysis until a subject was begun on steroid treatment. Measurements included in the model for CBD were: FEV₁, FVC, TLC, DL_CO, PaO₂, and PaO₂m. All analyses were performed using SAS version 8.2 (SAS Institute).

D′ and r², measures of pairwise linkage disequilibrium, were calculated using Haploview (Whitehead Institute for Biomedical Research). High values of D′ indicate evidence of recombination, while low values suggest that the alleles are passed independently. R², another measure of linkage disequilibrium, measures statistical association and is a more reliable measure for low allele frequencies. In addition, Haploview was also used to determine haplotype blocks using the confidence interval (CI) method (25). Haplotype frequencies were estimated using Haplo.Score (26). This software uses an expectation maximization-based algorithm to calculate the posterior probability of each possible haplotype combination for each individual when haplotype phase is unknown. To adjust for the uncertainty in haplotype assignments, we used a weighted logistic regression model. Each person could appear in the data set more than once, with each entry weighted by the probability of that haplotype combination for that individual, so that the total contribution of each individual was one observation. We tested for both haplotype combination and single haplotype effects: presence of a specific haplotype pair vs not and carrying at least one specified haplotype vs not carrying the specified haplotype. These comparisons were made when there were sufficient numbers of observed haplotypes to make a valid comparison.

Demographics for CBD and sarcoidosis cases and controls are shown in Table I. CBD controls were older on average than CBD cases or sarcoidosis cases and controls. More females were represented in the sarcoidosis case and control groups than in CBD cases and controls along with a higher percentage of non-Caucasians, which is to be expected given the demographics of the disease that have been previously reported. As shown in Table II, the sarcoidosis cases had mild radiographic abnormalities, with few treated with steroids; this is not surprising given that the ACCESS study enrolled incident cases (22). A greater percentage of the CBD subjects were ever or currently being treated with steroids or other immunosuppressive compared with ACCESS cases.

The genotype frequencies for all polymorphisms were in the Hardy-Weinberg equilibrium for the control populations. Genotype frequencies of TGF-β1 variants are listed in Tables III and IV. No significant differences were found for any of the polymorphisms comparing CBD cases to their controls, or sarcoidosis cases and their controls, or cases of CBD to sarcoidosis.

Haplotype analysis was performed. The highest linkage was found between the −509 C and the codon 10 T alleles (D′ 0.95 CI = 0.91–0.98, r² = 0.67). Additionally, linkage disequilibrium was found between C−509T and G-800A (D′ = 1.0, CI = 0.84–1.0, r² = 0.054 overall). As a result, the first three SNPs (−800/−509/codon 10) form one block of linkage disequilibrium based on D′ and the CI method (Fig. 1). There was lower linkage between these polymorphisms and the codon 25 and 263 polymorphisms (r² = 0.003–0.08). Haplotype frequencies for haplotypes containing 1) all SNPs and 2) only the three SNPs included in the block were estimated as described in Materials and Methods, and association tests were performed. No significant differences were found between cases and controls within each disease nor between case groups for either set of haplotypes (Tables V and VI).

Differences in rates of CBD disease progression were evaluated for associations with TGF-β1 polymorphisms using mixed effects models adjusted for age at test, gender, race, smoking history, and height (for lung function test variables), along with time from first exposure for CBD. This analysis was limited to the Caucasian cases, due to the small number of cases of other race. The coefficients resulting from fitting the mixed effects models to the longitudinal severity data are shown in Table VII and can be interpreted as coefficients from a simple linear regression model. For example, CBD cases homozygous for T at codon 10 had on average a 0.03 L/year greater rate in decline in FVC than those cases without a T at codon 10 (p < 0.0001). Similarly, the CBD cases homozygous for a T at codon 10 also had a significantly greater rate of decline in FEV₁, TLC, and PaO₂ at rest and on maximal exertion than cases with at least one C at this locus. Cases homozygous for C at position −509 showed significantly lower TLC, and PaO₂ at rest and maximum exertion over time than those cases with at least one T at the −509 position (Table VII). No associations were found among the other polymorphisms (data not shown). No associations were noted between abnormal chest radiograph at baseline (n = 11 of 73) and these polymorphisms (data not shown), although numbers were small.

Differences in rates of CBD disease progression were also examined for association with haplotypes when the counts of the haplotypes were sufficient to make the comparison meaningful (i.e., n > 20) and if clinical data was available for the subjects in question to ensure the model converged (Table VIII). The CBD cases homozygous for GCT, and thereby including two alleles that we have already shown to increase susceptibility to more severe disease, had significantly greater FEV₁, FVC (Fig. 2), TLC, and PaO₂ at rest and maximum exertion declines over time than those CBD cases who did not have this haplotype pair. In contrast, CBD individuals who carried the GTC haplotype, had no decline in FEV₁, FVC, TLC, and PaO₂ at rest and maximum exertion over time compared with those who did not carry this haplotype. Unlike the homozygotes for the GCT haplotype, the carriage of one GCT haplotype along with one GTC haplotype did not result in a decline in lung function over time. No association was noted between the GTC haplotype and chest radiographic abnormality or treatment with steroids over time (data not shown).

Markers of disease severity, including spirometry and chest radiographic (Scadding stage) stage, were evaluated for the Caucasian sarcoidosis cases and their association with TGF-β1 alleles after adjusting for gender, smoking, pack years, and age at examination. In concordance with the CBD data, those individuals homozygous for CC at position −509 were more likely to have more advanced disease, as evidenced by a Scadding radiograph stage of II, III, IV than those with a CT or TT genotype (odds ratio (OR) = 2.5, CI (1.3, 4.5), Table IX). Likewise, those homozygous for TT at codon 10 were more likely to have a higher Scadding radiograph stage score than those individuals with a TC or CC genotype (OR = 2.3, CI (1.1, 4.5). A trend was observed for CC at codon 263 being associated with a more severe Scadding stage, but this was not significant after adjusting for covariates.

Consistent with these allele results, associations between Scadding radiograph stage and TGF β haplotypes were observed (Table IX). In this regard, those individuals with two copies of the GCT haplotype tended to have the more severe Scadding stages II, III, IV (OR = 2.2 (0.96–0.4.9, p = 0.06)). In contrast, individuals homozygous for the −800/−509/codon 10 GTC haplotype were significantly less likely to have a Scadding stage of II, III, or IV (OR = 0.39 (0.20–0.76) p = 0.005). If an individual had at least one GTC haplotype, regardless of whether they had a GCT haplotype or not, they were less likely to have Scadding radiograph stage II, III, or IV (OR = 0.41 (0.22–0.77), p = 0.005).

A small subset of the sarcoidosis population studied had erythema nodosum, a skin manifestation of sarcoidosis that is associated with a less severe form of sarcoidosis (27). Eleven of the 12 (91.7%) sarcoidosis cases who had erythema nodosum had at least one T at the −509 position in contrast to 53.1% in the nonerythema nodosum group, consistent with the concept that this −509 allele is associated with less severe disease. Similarly, 91.7% of those with erythema nodosum had at least one C at the codon 10 position compared with 67.5% in the nonerythema nodosum group. There were no significant associations found between FEV₁, FVC, or FEV₁/FVC and the TGF-β1 variants of interest. Other lung function indices were not available for analysis.

This study has for the first time identified associations between CBD disease progression over time and variations in genotype and haplotype in the TGFβ-1 gene. Strikingly, the same patterns of genotype and haplotype were linked in this study to more severe disease in another granulomatous disorder, sarcoidosis, while the opposite genotypes tended to be associated with a milder form of sarcoidosis, erythema nodosum. This concordance between the same genetic associations and these two distinct granulomatous processes has not been reported previously but supports our hypothesis that these granulomatous diseases share similar genetic factors and pathways. There were no associations between gene variants and the likelihood of developing either CBD or sarcoidosis, emphasizing the importance of assessing genetic contributions to disease progression as well as disease presence. Importantly, the susceptibility haplotype for progression of both diseases includes two loci (−509 and codon 10) that are known to be functional. Specifically, the −509T and the codon 10C or proline allele have been associated with higher TGF-β1 mRNA levels in PBMC or protein levels in serum, compared with the −509C and codon 10T or leucine (11, 12). One other site (codon 25) is also a polymorphic site where the wild-type allele leads to higher levels of TGF-β1 (13).

The TGF-β1 locus, and specifically the −509 and codon 10 polymorphisms, has been associated with other forms of more severe lung disease, including cystic fibrosis (28) and idiopathic pulmonary fibrosis (IPF; Ref. 29), but not with the susceptibility of these diseases, as in the case of CBD and sarcoidosis. Interestingly, both of these studies used physiologic markers of disease severity or progression over time; FEV₁ in the case of the cystic fibrosis study and alveolar arterial oxygen gradient, at rest in the case of IPF. This supports our findings that the TGFβ-1 gene is a genetic modifier of lung disease, but not necessarily lung disease susceptibility. To study disease severity in CBD, we assessed the change in lung function parameters over time, adjusting for potential confounders of disease severity, including smoking status, age, time because first exposure to beryllium, with the possibility that exposure may impact disease progression and severity (30). In sarcoidosis, we used chest radiographic stage as a marker of disease severity and found similar associations. No associations were noted between sarcoidosis spirometric values at time of diagnosis; we did not have information for DL_CO or for any of these variables over time for sarcoidosis which may have accounted for the lack of association. Of note, in cystic fibrosis and IPF, the disease severity markers were associated with the higher producing TGFβ-1 polymorphisms from previous studies. However, other studies have shown that these same variants are protective for smoking-induced chronic obstructive pulmonary disease (31, 32), and associated with lower FEV₁ in chronic obstructive pulmonary disease (32). This suggests that the same genetic polymorphisms may function in an opposite manner depending on the disease, the environment, and potentially other genetic factors. It is also possible that these changes in lung function associated with genotypes could be independent of CBD and a function of the genotype.

The striking finding that homozygosity in −509C and codon 10T alleles and their haplotype is associated with progressive decline over time—in spirometry (in the case of the codon 10 and the haplotype), TLC, and gas exchange at rest and exercise in the case of CBD, and chest radiographic Scadding stage at one time point in sarcoidosis—suggests that the TGF-β1 gene is functional in these two granulomatous disease and may have a similar mechanistic role. TGF-β1 has multiple functions, including growth promotion and cell differentiation, extracellular matrix production and fibrosis, and proinflammatory and anti-inflammatory effects.

Knowing that the more severe forms of these two diseases may be associated with a fibrotic response might lead one to expect that higher TGF-β levels would be found in the more severe forms of disease and thus with more progressive fibrosis (14). However, the genotypes associated with more severe CBD and sarcoidosis are the same ones which have been associated with lower TGF-β levels in previous studies (11, 33). This might suggest that TGF-β1 is not playing a role in fibrogenesis, but instead may be playing a role in immune regulation of these granulomatous diseases. A recently published study demonstrates that TGF-β plays a role in negatively regulating both innate and acquitted immunity (34, 35, 36); specifically, the article by Oida et al. (34) indicates that the immunosuppressive properties of TGF-β1 are mediated by CD4⁺CD25⁺ T regulatory cells. The lower TGF-β production expected in those with the −509 CC and codon 10TT in CBD and sarcoidosis could lead to reduced inhibition of the immune response by TGF-β1 and, as a result, a proinflammatory microenvironment more conducive to disease progression. Recent preliminary data from our group supports this hypothesis, as the frequency of FoxP3-expressing CD4⁺CD25⁺ cells in BAL of CBD are significantly decreased compared with beryllium-sensitized subjects without disease (A. Fontenot, personal communication). Although numbers of CD4⁺CD25⁺ BAL cells are reported to be increased in sarcoidosis, their function is dysregulated, possibly supporting a defect in the immunosuppressive properties of TGF-β1 in sarcoidosis (37).

In additional support of this hypothesis is our finding that the presence of even one copy of the alternate haplotype containing the −509T and codon 10C results in a lack of progression of FEV₁, FVC, TLC, and PaO₂ at rest or maximum exercise over time in CBD and in a lower risk of having Scadding stage II, III, or IV chest radiograph in sarcoidosis. Of note, these genotypes were present in the majority of individuals with a benign, resolving form of sarcoidosis, erythema nodosum. In this regard, one study that examined the levels of TGF-β1 in cultured supernatants of BAL cells from sarcoidosis cases (38) demonstrated that cases whose disease underwent spontaneous remission within 6 mo of lavage had significantly higher levels of TGF-β production than cases with more persistent or progressive disease. This study is consistent with our findings in that genotypes that are associated with disease severity are also associated with lower TGF production. These data make two important points: 1) that the gene(s) responsible for disease is/are not necessarily those that are involved in progression or severity, and 2) that there is likely a continuum of effects that is determined by a number of genes each of which contribute variably, and possibly interactively, to disease initiation and severity and ultimate outcome. We would have ideally addressed this issue further by analyzing the subset of Glu⁶⁹-positive individuals to establish whether the severity associations were reproducible in this subset. Unfortunately, numbers were too small in this cohort to perform subset analysis but this is clearly an important aspect for future study.

The function of TGF-β in CBD has not been studied to date, but as in the case of sarcoidosis, these data and the studies above would suggest that in more advanced CBD and sarcoidosis that lower TGF-β1 production is associated with decreased expansion and immunosuppressive capabilities of CD4⁺CD25 T cells, which are present in CBD (A. Fontenot, personal communication), and thus increased inflammation and Th1 and macrophage functions (35, 36). Previous studies indicate that TGF-β1 immunomodulation may affect proliferation, modulation of chemotaxis, and the regulation of cytokine production, including TNF and other Th1 cytokines, or even affect MHC class I and II Ag expression (13, 39), which in turn would affect inflammation in these two diseases. A recent study by Meng et al. (40) supports this concept, as they found higher T cell proliferation to allergens associated with the 509CC polymorphism. This will require future mechanistic study and clarification in CBD and sarcoidosis.

Other studies have explored the relationship between TGF-β1 polymorphisms, especially those in codons 10 and codon 25, and granulomatous lung diseases. The codon 25 locus was examined by Gaede et al. (41) in two populations of cases with CBD, one Israeli/European and the other American. The Israeli/European population showed a significant association between CBD and carrying the low-producing non-GG genotype (codon 25) although no association was found for the population from the United States. The investigators did not assess other polymorphisms assessed in this study or markers of disease severity. The Israeli/European population is known to have significant disease as it is a clinically detected cohort, not one determined due to medical surveillance as in the U.S. population. This difference may account for the investigators different findings in these two populations. Certainly, the findings in the U.S. population are similar to ours, with a lack of disease association apparent. The studies of sarcoidosis to date have supported our lack of an association with disease susceptibility (42, 43, 44).

In conclusion, this study has shown that specific TGF-β genotypes and haplotypes are associated with a more severe pulmonary phenotype in both sarcoidosis and chronic beryllium disease, although not affecting disease susceptibility per se. Two of the loci in the three-locus haplotype include variant alleles that have been associated with lower production of TGF, implicating lesser TGF-mediated immune regulation as at least part of the mechanism whereby disease is allowed to become more progressive. Future studies will be needed to assess the functional implications of these associations on TGF-β production and its role in theses granulomatous diseases.

We acknowledge the ACCESS investigators for their development of the ACCESS cohort; Margaret Mroz, MSPH, for thoughtful suggestions and discussion; Gina Mondello for technical assistance; Mary MacNaugton for administrative support; and most importantly, the workers and patients who participate in these studies and make beryllium- and sarcoidosis-related research possible. A Case Control Etiologic Study of Sarcoidosis (ACCESS) Research Group: Clinical Centers: Beth Israel Deaconess Medical Center: Steven E. Weinberger, Patricia Finn, Erik Garpestad, Allison Moran; Georgetown University Medical Center: Henry Yeager, Jr., David L. Rabin, Susan Stein; Case Western Reserve University—Henry Ford Health Sciences Center: Michael C. Iannuzzi, Benjamin A. Rybicki, Marcie Major, Mary Maliarik, John Popovich, Jr.; Johns Hopkins University School of Medicine: David R. Moller, Carol J. Johns (deceased), Cynthia Rand, Joanne Steimel; Medical University of South Carolina: Marc A. Judson, Susan D’Alessandro, Nancy Heister, Theresa Johnson, Daniel T. Lackland, Janardan Pandey, Steven Sahn, Charlie Strange; Mount Sinai Medical Center: Alvin S. Teirstein, Louis DePalo, Sheldon Brown, Marvin Lesser, Maria L. Padilla, Marilyn Marshall; National Jewish Medical and Research Center: Lee S. Newman, Cecile Rose, Juliana Barnard, John Martyny, Charles McCammon; University of Cincinnati Medical Center: Robert P. Baughman, Elyse E. Lower, Donna B. Winget; University of Iowa College of Medicine: Geoffrey McLennan, Gary Hunninghake, Chuck Dayton, Linda Powers; University of Pennsylvania and Medical College of Pennsylvania—Hahnemann University Medical Centers: Milton D. Rossman, Eddy A. Bresnitz, Ronald Daniele, Jackie Regovich, William Sexauer; National Heart, Lung, and Blood Institute: Robert Musson, Joanne Deshler, Paul Sorlie, Margaret Wu; Study Chairman: Reuben Cherniack; Study Co-Chairman: Lee Newman; Clinical Trials & Surveys Corp.: Genell L. Knatterud, Michael L. Terrin, Bruce W. Thompson, Kathleen Brown, Margaret Frederick, Frances LoPresti, Patricia Wilkins, Martha Canner, Judy Dotson; McKesson Bioservices: Steve Lindenfelser; BBI-Biotech Research Laboratories: Mark Cosentino.

The authors have no financial conflict of interest.