Erythroid lineage chromatin accessibility maps facilitate identification and validation of NFIX as a fetal hemoglobin repressor.

Human genetics has validated de-repression of fetal gamma globin (HBG) in adult erythroblasts as a powerful therapeutic paradigm in diseases involving defective adult beta globin (HBB)1. To identify factors involved in the switch from HBG to HBB expression, we performed Assay for Transposase Accessible Chromatin with high-throughput sequencing (ATAC-seq)2 on sorted erythroid lineage cells derived from bone marrow (BM) or cord blood (CB), representing adult and fetal states, respectively. BM to CB cell ATAC-seq profile comparisons revealed genome-wide enrichment of NFI DNA binding motifs and increased NFIX promoter chromatin accessibility, suggesting that NFIX may repress HBG. NFIX knockdown in BM cells increased HBG mRNA and fetal hemoglobin (HbF) protein levels, coincident with increased chromatin accessibility and decreased DNA methylation at the HBG promoter. Conversely, overexpression of NFIX in CB cells reduced HbF levels. Identification and validation of NFIX as a new target for HbF activation has implications in the development of therapeutics for hemoglobinopathies.

Targeting RARA overexpression with tamibarotene, a potent and selective RARα agonist, is a novel approach in AML.

A superenhancer at the retinoic acid receptor alpha (RARA) gene is associated with RARA mRNA overexpression in ∼30% of non-acute promyelocytic leukemia acute myeloid leukemia (AML) and in ∼50% of myelodysplastic syndromes (MDS). RARA overexpression is an actionable target for treatment with tamibarotene, an oral potent and selective RARα agonist. Sensitivity to the RARα agonist tamibarotene was demonstrated in RARA-high but not RARA-low preclinical AML models. The combination of oral tamibarotene plus azacitidine was evaluated in a phase 2 clinical study in 51 newly diagnosed unfit patients with AML identified as RARA-positive (n = 22) or RARA-negative (n = 29) for RARA mRNA overexpression in peripheral blasts using a blood-based biomarker test. In 18 response-evaluable RARA-positive patients, complete remission (CR)/CR with incomplete hematologic recovery rate was 61%, CR rate was 50%, and time to initial composite CR was rapid at 1.2 months. Transfusion independence was attained by 72% of RARA-positive patients. In contrast, 28 response-evaluable RARA-negative patients had response rates that were consistent with azacitidine monotherapy. Tamibarotene in combination with azacitidine was well tolerated. The majority of nonhematologic adverse events were low grade and hematologic adverse events were comparable to single-agent azacitidine, demonstrating that there was no additional myelosuppression when tamibarotene was combined with azacitidine. These results support further evaluation of tamibarotene-based treatment strategies in patients with AML or MDS with RARA overexpression to provide a targeted approach with the goal of improving patient outcomes. This trial was registered at www.clinicaltrials.gov as #NCT02807558.

ΔNp63/p73 drive metastatic colonization by controlling a regenerative epithelial stem cell program in quasi-mesenchymal cancer stem cells.

Cancer stem cells (CSCs) may serve as the cellular seeds of tumor recurrence and metastasis, and they can be generated via epithelial-mesenchymal transitions (EMTs). Isolating pure populations of CSCs is difficult because EMT programs generate multiple alternative cell states, and phenotypic plasticity permits frequent interconversions between these states. Here, we used cell-surface expression of integrin ß4 (ITGB4) to isolate highly enriched populations of human breast CSCs, and we identified the gene regulatory network operating in ITGB4+ CSCs. Specifically, we identified ΔNp63 and p73, the latter of which transactivates ΔNp63, as centrally important transcriptional regulators of quasi-mesenchymal CSCs that reside in an intermediate EMT state. We found that the transcriptional program controlled by ΔNp63 in CSCs is largely distinct from the one that it orchestrates in normal basal mammary stem cells and, instead, it more closely resembles a regenerative epithelial stem cell response to wounding. Moreover, quasi-mesenchymal CSCs repurpose this program to drive metastatic colonization via autocrine EGFR signaling.

Discovery and Characterization of SY-1365, a Selective, Covalent Inhibitor of CDK7.

Recent studies suggest that targeting transcriptional machinery can lead to potent and selective anticancer effects in cancers dependent on high and constant expression of certain transcription factors for growth and survival. Cyclin-dependent kinase 7 (CDK7) is the catalytic subunit of the CDK-activating kinase complex. Its function is required for both cell-cycle regulation and transcriptional control of gene expression. CDK7 has recently emerged as an attractive cancer target because its inhibition leads to decreased transcript levels of oncogenic transcription factors, especially those associated with super-enhancers. Here, we describe a selective CDK7 inhibitor SY-1365, which is currently in clinical trials in populations of patients with ovarian and breast cancer (NCT03134638). In vitro, SY-1365 inhibited cell growth of many different cancer types at nanomolar concentrations. SY-1365 treatment decreased MCL1 protein levels, and cancer cells with low BCL2L1 (BCL-XL) expression were found to be more sensitive to SY-1365. Transcriptional changes in acute myeloid leukemia (AML) cell lines were distinct from those following treatment with other transcriptional inhibitors. SY-1365 demonstrated substantial antitumor effects in multiple AML xenograft models as a single agent; SY-1365-induced growth inhibition was enhanced in combination with the BCL2 inhibitor venetoclax. Antitumor activity was also observed in xenograft models of ovarian cancer, suggesting the potential for exploring SY-1365 in the clinic in both hematologic and solid tumors. Our findings support targeting CDK7 as a new approach for treating transcriptionally addicted cancers. SIGNIFICANCE: These findings demonstrate the molecular mechanism of action and potent antitumor activity of SY-1365, the first selective CDK7 inhibitor to enter clinical investigation.

A whole-genome sequence study identifies genetic risk factors for neuromyelitis optica.

Neuromyelitis optica (NMO) is a rare autoimmune disease that affects the optic nerve and spinal cord. Most NMO patients ( > 70%) are seropositive for circulating autoantibodies against aquaporin 4 (NMO-IgG+). Here, we meta-analyze whole-genome sequences from 86 NMO cases and 460 controls with genome-wide SNP array from 129 NMO cases and 784 controls to test for association with SNPs and copy number variation (total N = 215 NMO cases, 1244 controls). We identify two independent signals in the major histocompatibility complex (MHC) region associated with NMO-IgG+, one of which may be explained by structural variation in the complement component 4 genes. Mendelian Randomization analysis reveals a significant causal effect of known systemic lupus erythematosus (SLE), but not multiple sclerosis (MS), risk variants in NMO-IgG+. Our results suggest that genetic variants in the MHC region contribute to the etiology of NMO-IgG+ and that NMO-IgG+ is genetically more similar to SLE than MS.

SMN deficiency in severe models of spinal muscular atrophy causes widespread intron retention and DNA damage.

Spinal muscular atrophy (SMA), an autosomal recessive neuromuscular disease, is the leading monogenic cause of infant mortality. Homozygous loss of the gene survival of motor neuron 1 (SMN1) causes the selective degeneration of lower motor neurons and subsequent atrophy of proximal skeletal muscles. The SMN1 protein product, survival of motor neuron (SMN), is ubiquitously expressed and is a key factor in the assembly of the core splicing machinery. The molecular mechanisms by which disruption of the broad functions of SMN leads to neurodegeneration remain unclear. We used an antisense oligonucleotide (ASO)-based inducible mouse model of SMA to investigate the SMN-specific transcriptome changes associated with neurodegeneration. We found evidence of widespread intron retention, particularly of minor U12 introns, in the spinal cord of mice 30 d after SMA induction, which was then rescued by a therapeutic ASO. Intron retention was concomitant with a strong induction of the p53 pathway and DNA damage response, manifesting as γ-H2A.X positivity in neurons of the spinal cord and brain. Widespread intron retention and markers of the DNA damage response were also observed with SMN depletion in human SH-SY5Y neuroblastoma cells and human induced pluripotent stem cell-derived motor neurons. We also found that retained introns, high in GC content, served as substrates for the formation of transcriptional R-loops. We propose that defects in intron removal in SMA promote DNA damage in part through the formation of RNA:DNA hybrid structures, leading to motor neuron death.

Patient-Centered Therapy Development for Myotonic Dystrophy: Report of the Myotonic Dystrophy Foundation-Sponsored Workshop.

Myotonic dystrophy (DM) is an autosomal dominant, repeat expansion, progressive disorder with no drug therapies. Consequently, to better define a regulatory pathway in anticipation of new treatment strategies under investigation, the Myotonic Dystrophy Foundation convened a workshop entitled "Patient-Centered Therapy Development for Myotonic Dystrophy" in September 2015. Participants included representatives from academia, industry, the patient community, the National Institutes of Health (NIH) and the Food and Drug Administration (FDA). Presenters described the symptom burden of the disease, and existing data on DM biomarkers, endpoints, natural history, and benefit-risk considerations. FDA participants helped clarify the regulatory requirements for new drug treatment approvals and DM-specific issues such as variability, slow progression, and low prevalence. Workshop attendees gained a better understanding of DM and the current status of existing data and tools to support therapeutic drug research and development.

JC Polyomavirus Abundance and Distribution in Progressive Multifocal Leukoencephalopathy (PML) Brain Tissue Implicates Myelin Sheath in Intracerebral Dissemination of Infection.

Over half of adults are seropositive for JC polyomavirus (JCV), but rare individuals develop progressive multifocal leukoencephalopathy (PML), a demyelinating JCV infection of the central nervous system. Previously, PML was primarily seen in immunosuppressed patients with AIDS or certain cancers, but it has recently emerged as a drug safety issue through its association with diverse immunomodulatory therapies. To better understand the relationship between the JCV life cycle and PML pathology, we studied autopsy brain tissue from a 70-year-old psoriasis patient on the integrin alpha-L inhibitor efalizumab following a ~2 month clinical course of PML. Sequence analysis of lesional brain tissue identified PML-associated viral mutations in regulatory (non-coding control region) DNA, capsid protein VP1, and the regulatory agnoprotein, as well as 9 novel mutations in capsid protein VP2, indicating rampant viral evolution. Nine samples, including three gross PML lesions and normal-appearing adjacent tissues, were characterized by histopathology and subject to quantitative genomic, proteomic, and molecular localization analyses. We observed a striking correlation between the spatial extent of demyelination, axonal destruction, and dispersion of JCV along white matter myelin sheath. Our observations in this case, as well as in a case of PML-like disease in an immunocompromised rhesus macaque, suggest that long-range spread of polyomavirus and axonal destruction in PML might involve extracellular association between virus and the white matter myelin sheath.

Rescue of gene-expression changes in an induced mouse model of spinal muscular atrophy by an antisense oligonucleotide that promotes inclusion of SMN2 exon 7.

Spinal muscular atrophy (SMA) is a neuromuscular disease caused by disruption of the survival motor neuron 1 (SMN1) gene, partly compensated for by the paralogous gene SMN2. Exon 7 inclusion is critical for full-length SMN protein production and occurs at a much lower frequency for SMN2 than for SMN1. Antisense oligonucleotide (ASO)-mediated blockade of an intron 7 splicing silencer was previously shown to promote inclusion of SMN2 exon 7 in SMA mouse models and mediate phenotypic rescue. However, downstream molecular consequences of this ASO therapy have not been defined. Here we characterize the gene-expression changes that occur in an induced model of SMA and show substantial rescue of those changes in central nervous system tissue upon intracerebroventricular administration of an ASO that promotes inclusion of exon 7, with earlier administration promoting greater rescue. This study offers a robust reference set of preclinical pharmacodynamic gene expression effects for comparison of other investigational therapies for SMA.

Exome sequencing in amyotrophic lateral sclerosis identifies risk genes and pathways.

Amyotrophic lateral sclerosis (ALS) is a devastating neurological disease with no effective treatment. We report the results of a moderate-scale sequencing study aimed at increasing the number of genes known to contribute to predisposition for ALS. We performed whole-exome sequencing of 2869 ALS patients and 6405 controls. Several known ALS genes were found to be associated, and TBK1 (the gene encoding TANK-binding kinase 1) was identified as an ALS gene. TBK1 is known to bind to and phosphorylate a number of proteins involved in innate immunity and autophagy, including optineurin (OPTN) and p62 (SQSTM1/sequestosome), both of which have also been implicated in ALS. These observations reveal a key role of the autophagic pathway in ALS and suggest specific targets for therapeutic intervention.

Modular analysis of peripheral blood gene expression in rheumatoid arthritis captures reproducible gene expression changes in tumor necrosis factor responders.

OBJECTIVE: To establish whether the analysis of whole-blood gene expression is useful in predicting or monitoring response to anti-tumor necrosis factor (anti-TNF) therapy in patients with rheumatoid arthritis (RA). METHODS: Whole-blood RNA (using a PAXgene system to stabilize whole-blood RNA in the collection tube) was obtained at baseline and at 14 weeks from 3 independent cohorts, consisting of a combined total of 240 RA patients who were beginning therapy with anti-TNF. We used an approach to gene expression analysis that is based on modular patterns of gene expression, or modules. RESULTS: Good and moderate responders according to the European League Against Rheumatism criteria exhibited highly significant and consistent changes in multiple gene expression modules after 14 weeks of therapy, as demonstrated by hypergeometric analysis. Strikingly, nonresponders exhibited very little change in any modules, despite exposure to TNF blockade. These patterns of change were highly consistent across all 3 cohorts, indicating that immunologic changes after TNF treatment are specific to the combination of both drug exposure and responder status. In contrast, modular patterns of gene expression did not exhibit consistent differences between responders and nonresponders at baseline in the 3 study cohorts. CONCLUSION: These data provide evidence that using gene expression modules related to inflammatory disease may provide a valuable method for objective monitoring of the response of RA patients who are treated with TNF inhibitors.

Whole Genome Sequencing Reveals a Chromosome 9p Deletion Causing DOCK8 Deficiency in an Adult Diagnosed with Hyper IgE Syndrome Who Developed Progressive Multifocal Leukoencephalopathy.

PURPOSE: A 30 year-old man with a history of recurrent skin infections as well as elevated serum IgE and eosinophils developed neurological symptoms and had T2-hyperintense lesions observed in cerebral MRI. The immune symptoms were attributed to Hyper IgE syndrome (HIES) and the neurological symptoms with presence of JC virus in cerebrospinal fluid were diagnosed as Progressive Multifocal Leukoencephalopathy (PML). The patient was negative for STAT3 mutations. To determine if other mutations explain HIES and/or PML in this subject, his DNA was analyzed by whole genome sequencing. METHODS: Whole genome sequencing was completed to 30X coverage, and whole genome SNP typing was used to complement these data. The methods revealed single nucleotide variants, structural variants, and copy number variants across the genome. Genome-wide data were analyzed for homozygous or compound heterozygous null mutations for all protein coding genes. Mutations were confirmed by PCR and/or Sanger sequencing. RESULTS: Whole genome analysis revealed deletions near the telomere of both copies of chromosome 9p. Several genes, including DOCK8, were impacted by the deletions but it was unclear whether each chromosome had identical or distinct deletions. PCR across the impacted region combined with Sanger sequencing of selected fragments confirmed a homozygous deletion from position 10,211 to 586,751. CONCLUSION: While several genes are impacted by the deletion, DOCK8 deficiency is the most probable cause of HIES in this patient. DOCK8 deficiency may have also predisposed the patient to develop PML.

Newborn blood spot screening test using multiplexed real-time PCR to simultaneously screen for spinal muscular atrophy and severe combined immunodeficiency.

BACKGROUND: Spinal muscular atrophy (SMA) is a motor neuron disorder caused by the absence of a functional survival of motor neuron 1, telomeric (SMN1) gene. Type I SMA, a lethal disease of infancy, accounts for the majority of cases. Newborn blood spot screening (NBS) to detect severe combined immunodeficiency (SCID) has been implemented in public health laboratories in the last 5 years. SCID detection is based on real-time PCR assays to measure T-cell receptor excision circles (TREC), a byproduct of T-cell development. We modified a multiplexed real-time PCR TREC assay to simultaneously determine the presence or absence of the SMN1 gene from a dried blood spot (DBS) punch in a single reaction well. METHOD: An SMN1 assay using a locked nucleic acid probe was initially developed with cell culture and umbilical cord blood (UCB) DNA extracts, and then integrated into the TREC assay. DBS punches were placed in 96-well arrays, washed, and amplified directly using reagents specific for TREC, a reference gene [ribonuclease P/MRP 30kDa subunit (RPP30)], and the SMN1 gene. The assay was tested on DBS made from UCB units and from peripheral blood samples of SMA-affected individuals and their family members. RESULTS: DBS made from SMA-affected individuals showed no SMN1-specific amplification, whereas DBS made from all unaffected carriers and UCB showed SMN1 amplification above a well-defined threshold. TREC and RPP30 content in all DBS were within the age-adjusted expected range. CONCLUSIONS: SMA caused by the absence of SMN1 can be detected from the same DBS punch used to screen newborns for SCID.

Development of a multi-biomarker disease activity test for rheumatoid arthritis.

BACKGROUND: Disease activity measurement is a key component of rheumatoid arthritis (RA) management. Biomarkers that capture the complex and heterogeneous biology of RA have the potential to complement clinical disease activity assessment. OBJECTIVES: To develop a multi-biomarker disease activity (MBDA) test for rheumatoid arthritis. METHODS: Candidate serum protein biomarkers were selected from extensive literature screens, bioinformatics databases, mRNA expression and protein microarray data. Quantitative assays were identified and optimized for measuring candidate biomarkers in RA patient sera. Biomarkers with qualifying assays were prioritized in a series of studies based on their correlations to RA clinical disease activity (e.g. the Disease Activity Score 28-C-Reactive Protein [DAS28-CRP], a validated metric commonly used in clinical trials) and their contributions to multivariate models. Prioritized biomarkers were used to train an algorithm to measure disease activity, assessed by correlation to DAS and area under the receiver operating characteristic curve for classification of low vs. moderate/high disease activity. The effect of comorbidities on the MBDA score was evaluated using linear models with adjustment for multiple hypothesis testing. RESULTS: 130 candidate biomarkers were tested in feasibility studies and 25 were selected for algorithm training. Multi-biomarker statistical models outperformed individual biomarkers at estimating disease activity. Biomarker-based scores were significantly correlated with DAS28-CRP and could discriminate patients with low vs. moderate/high clinical disease activity. Such scores were also able to track changes in DAS28-CRP and were significantly associated with both joint inflammation measured by ultrasound and damage progression measured by radiography. The final MBDA algorithm uses 12 biomarkers to generate an MBDA score between 1 and 100. No significant effects on the MBDA score were found for common comorbidities. CONCLUSION: We followed a stepwise approach to develop a quantitative serum-based measure of RA disease activity, based on 12-biomarkers, which was consistently associated with clinical disease activity levels.

Optimization of a high-throughput whole blood expression profiling methodology and its application to assess the pharmacodynamics of interferon (IFN) beta-1a or polyethylene glycol-conjugated IFN beta-1a in healthy clinical trial subjects.

BACKGROUND: Clinical trials offer a unique opportunity to study human disease and response to therapy in a highly controlled setting. The application of high-throughput expression profiling to peripheral blood from clinical trial subjects could facilitate the identification of transcripts that function as prognostic or diagnostic markers of disease or treatment. The paramount issue for these methods is the ability to produce robust, reproducible, and timely mRNA expression profiles from peripheral blood. Single-stranded complementary DNA (sscDNA) targets derived from whole blood exhibit improved detection of transcripts and reduced variance as compared to their complementary RNA counterparts and therefore provide a better option for interrogation of peripheral blood on oligonucleotide arrays. High-throughput microarray technologies such as the high-throughput plate array platform offer several advantages compared with slide- or cartridge-based arrays; however, manufacturer's protocols do not support the use of sscDNA targets. RESULTS: We have developed a highly reproducible, high-through put, whole blood expression profiling methodology based on sscDNA and used it to analyze human brain reference RNA and universal human reference RNA samples to identify experimental conditions that most highly correlated with a gold standard quantitative polymerase chain reaction reference dataset. We then utilized the optimized method to analyze whole blood samples from healthy clinical trial subjects treated with different versions of interferon (IFN) beta-1a. Analysis of whole blood samples before and after treatment with intramuscular [IM] IFN beta-1a or polyethylene glycol-conjugated IFN (PEG-IFN) beta-1a under optimized experimental conditions demonstrated that PEG-IFN beta-1a induced a more sustained and prolonged pharmacodynamic response than unmodified IM IFN beta-1a. These results provide validation of the utility of this new methodology and suggest the potential therapeutic benefit of a sustained pharmacodynamic response to PEG-IFN beta-1a. CONCLUSIONS: This novel microarray methodology is ideally suited for utilization in large clinical studies to identify expressed transcripts for the elucidation of disease mechanisms of action and as prognostic, diagnostic, or toxicity markers.

Progressive multifocal leukoencephalopathy (PML) development is associated with mutations in JC virus capsid protein VP1 that change its receptor specificity.

Progressive multifocal leukoencephalopathy (PML), a fatal demyelinating disease caused by JC virus (JCV) infection of oligodendrocytes, may develop in patients with immune disorders following reactivation of chronic benign infection. Mutations of JCV capsid viral protein 1 (VP1), the capsid protein involved in binding to sialic acid cell receptors, might favor PML onset. Cerebrospinal fluid sequences from 37/40 PML patients contained one of several JCV VP1 amino acid mutations, which were also present in paired plasma but not urine sequences despite the same viral genetic background. VP1-derived virus-like particles (VLPs) carrying these mutations lost hemagglutination ability, showed different ganglioside specificity, and abolished binding to different peripheral cell types compared with wild-type VLPs. However, mutants still bound brain-derived cells, and binding was not affected by sialic acid removal by neuraminidase. JCV VP1 substitutions are acquired intrapatient and might favor JCV brain invasion through abrogation of sialic acid binding with peripheral cells, while maintaining sialic acid-independent binding with brain cells.

Sequencing and analysis of JC virus DNA from natalizumab-treated PML patients.

BACKGROUND: Progressive multifocal leukoencephalopathy (PML) in natalizumab-treated MS patients is linked to JC virus (JCV) infection. JCV sequence variation and rearrangements influence viral pathogenicity and tropism. To better understand PML development, we analyzed viral DNA sequences in blood, CSF and/or urine of natalizumab-treated PML patients. METHODS: Using biofluid samples from 17 natalizumab-treated PML patients, we sequenced multiple isolates of the JCV noncoding control region (NCCR), VP1 capsid coding region, and the entire 5 kb viral genome. RESULTS: Analysis of JCV from multiple biofluids revealed that individuals were infected with a single genotype. Across our patient cohort, multiple PML-associated NCCR rearrangements and VP1 mutations were present in CSF and blood, but absent from urine-derived virus. NCCR rearrangements occurred in CSF of 100% of our cohort. VP1 mutations were observed in blood or CSF in 81% of patients. Sequencing of complete JCV genomes demonstrated that NCCR rearrangements could occur without VP1 mutations, but VP1 mutations were not observed without NCCR rearrangement. CONCLUSIONS: These data confirm that JCV in natalizumab-PML patients is similar to that observed in other PML patient groups, multiple genotypes are associated with PML, individual patients appear to be infected with a single genotype, and PML-associated mutations arise in patients during PML development.

Causal modeling using network ensemble simulations of genetic and gene expression data predicts genes involved in rheumatoid arthritis.

Tumor necrosis factor α (TNF-α) is a key regulator of inflammation and rheumatoid arthritis (RA). TNF-α blocker therapies can be very effective for a substantial number of patients, but fail to work in one third of patients who show no or minimal response. It is therefore necessary to discover new molecular intervention points involved in TNF-α blocker treatment of rheumatoid arthritis patients. We describe a data analysis strategy for predicting gene expression measures that are critical for rheumatoid arthritis using a combination of comprehensive genotyping, whole blood gene expression profiles and the component clinical measures of the arthritis Disease Activity Score 28 (DAS28) score. Two separate network ensembles, each comprised of 1024 networks, were built from molecular measures from subjects before and 14 weeks after treatment with TNF-α blocker. The network ensemble built from pre-treated data captures TNF-α dependent mechanistic information, while the ensemble built from data collected under TNF-α blocker treatment captures TNF-α independent mechanisms. In silico simulations of targeted, personalized perturbations of gene expression measures from both network ensembles identify transcripts in three broad categories. Firstly, 22 transcripts are identified to have new roles in modulating the DAS28 score; secondly, there are 6 transcripts that could be alternative targets to TNF-α blocker therapies, including CD86--a component of the signaling axis targeted by Abatacept (CTLA4-Ig), and finally, 59 transcripts that are predicted to modulate the count of tender or swollen joints but not sufficiently enough to have a significant impact on DAS28.

A genome-wide screen of gene-gene interactions for rheumatoid arthritis susceptibility.

The objective of the study was to identify interacting genes contributing to rheumatoid arthritis (RA) susceptibility and identify SNPs that discriminate between RA patients who were anti-cyclic citrullinated protein positive and healthy controls. We analyzed two independent cohorts from the North American Rheumatoid Arthritis Consortium. A cohort of 908 RA cases and 1,260 controls was used to discover pairwise interactions among SNPs and to identify a set of single nucleotide polymorphisms (SNPs) that predict RA status, and a second cohort of 952 cases and 1,760 controls was used to validate the findings. After adjusting for HLA-shared epitope alleles, we identified and replicated seven SNP pairs within the HLA class II locus with significant interaction effects. We failed to replicate significant pairwise interactions among non-HLA SNPs. The machine learning approach "random forest" applied to a set of SNPs selected from single-SNP and pairwise interaction tests identified 93 SNPs that distinguish RA cases from controls with 70% accuracy. HLA SNPs provide the most classification information, and inclusion of non-HLA SNPs improved classification. While specific gene-gene interactions are difficult to validate using genome-wide SNP data, a stepwise approach combining association and classification methods identifies candidate interacting SNPs that distinguish RA cases from healthy controls.