Search

Abstract Distinct Natural Killer (NK)-like CD57+ and PD-1+ CD8+ exhausted-like T cell populations (Tex) have both been linked to beneficial immunotherapy response in autoimmune type 1 diabetes (T1D) patients. The origins and relationships between these cell types are poorly understood. Here we show that while PD-1+ and CD57+ Tex populations are epigenetically similar, CD57+ Tex cells display unique increased chromatin accessibility of inhibitory Killer Cell Immunoglobulin-like Receptor (iKIR) and other NK cell genes. PD-1+ and CD57+ Tex also show reciprocal expression of Inhibitory Receptors (IRs) and iKIRs accompanied by chromatin accessibility of Tcf1 and Tbet transcription factor target sites, respectively. CD57+ Tex show unappreciated gene expression heterogeneity and share clonal relationships with PD-1+ Tex, with these cells differentiating along four interconnected lineage trajectories: Tex-PD-1+, Tex-CD57+, Tex-Branching, and Tex-Fluid. Our findings demonstrate new relationships between Tex-like populations in human autoimmune disease and suggest that modulating common precursor populations may enhance response to autoimmune disease treatment.

Abstract Human type 1 diabetes (T1D) is caused by autoimmune attack on the insulin-producing pancreatic beta cells by islet antigen-reactive T cells. How human islet antigen-reactive (IAR) CD4+ memory T cells from peripheral blood affect T1D progression in the pancreas is poorly understood. Here, we aim to determine if IAR T cells in blood could be detected in pancreas. We identify paired αβ (TRA/TRB) T cell receptors (TCRs) in IAR T cells from the blood of healthy, at-risk, new-onset, and established T1D donors, and measured sequence overlap with TCRs in pancreata from healthy, at risk and T1D organ donors. We report extensive TRA junction sharing between IAR T cells and pancreas-infiltrating T cells (PIT), with perfect-match or single-mismatch TRA junction amino acid sequences comprising ~29% total unique IAR TRA junctions (942/3,264). PIT-matched TRA junctions were largely public and enriched for TRAV41 usage, showing significant nucleotide sequence convergence, increased use of germline-encoded versus non-templated residues in epitope engagement, and a potential for cross-reactivity. Our findings thus link T cells with distinctive germline-like TRA chains in the peripheral blood with T cells in the pancreas.

Exhausted CD8 T cells (TEX) are associated with worse outcome in cancer yet better outcome in autoimmunity. Building on our past findings of increased TIGIT+KLRG1+ TEX with teplizumab therapy in type 1 diabetes (T1D), in the absence of treatment we found that the frequency of TIGIT+KLRG1+ TEX is stable within an individual but differs across individuals in both T1D and healthy control (HC) cohorts. This TIGIT+KLRG1+ CD8 TEX population shares an exhaustion-associated EOMES gene signature in HC, T1D, rheumatoid arthritis (RA), and cancer subjects, expresses multiple inhibitory receptors, and is hyporesponsive in vitro, together suggesting co-expression of TIGIT and KLRG1 may broadly define human peripheral exhausted cells. In HC and RA subjects, lower levels of EOMES transcriptional modules and frequency of TIGIT+KLRG1+ TEX were associated with RA HLA risk alleles (DR0401, 0404, 0405, 0408, 1001) even when considering disease status and cytomegalovirus (CMV) seropositivity. Moreover, the frequency of TIGIT+KLRG1+ TEX was significantly increased in RA HLA risk but not non-risk subjects treated with abatacept (CTLA4Ig). The DR4 association and selective modulation with abatacept suggests that therapeutic modulation of TEX may be more effective in DR4 subjects and TEX may be indirectly influenced by cellular interactions that are blocked by abatacept.

Variation in the preservation of β cell function in clinical trials in type 1 diabetes (T1D) has emphasized the need to define biomarkers to predict treatment response. The T1DAL trial targeted T cells with alefacept (LFA-3–Ig) and demonstrated C-peptide preservation in approximately 30% of new-onset T1D individuals. We analyzed islet antigen–reactive (IAR) CD4+ T cells in PBMC samples collected prior to treatment from alefacept- and placebo-treated individuals using flow cytometry and single-cell RNA sequencing. IAR CD4+ T cells at baseline had heterogeneous phenotypes. Transcript profiles formed phenotypic clusters of cells along a trajectory based on increasing maturation and activation, and T cell receptor (TCR) chains showed clonal expansion. Notably, the frequency of IAR CD4+ T cells with a memory phenotype and a unique transcript profile (cluster 3) were inversely correlated with C-peptide preservation in alefacept-treated, but not placebo-treated, individuals. Cluster 3 cells had a proinflammatory phenotype characterized by expression of the transcription factor BHLHE40 and the cytokines GM-CSF and TNF-α, and shared TCR chains with effector memory–like clusters. Our results suggest IAR CD4+ T cells as a potential baseline biomarker of response to therapies targeting the CD2 pathway and warrant investigation for other T cell–related therapies.

BACKGROUND Low-dose anti–thymocyte globulin (ATG) transiently preserves C-peptide and lowers HbA1c in individuals with recent-onset type 1 diabetes (T1D); however, the mechanisms of action and features of the response remain unclear. Here, we characterized the post hoc immunological outcomes of ATG administration and their potential use as biomarkers of metabolic response to therapy (i.e., improved preservation of endogenous insulin production).METHODS We assessed gene and protein expression, targeted gene methylation, and cytokine concentrations in peripheral blood following treatment with ATG (n = 29), ATG plus granulocyte colony–stimulating factor (ATG/G-CSF, n = 28), or placebo (n = 31).RESULTS Treatment with low-dose ATG preserved regulatory T cells (Tregs), as measured by stable methylation of FOXP3 Treg-specific demethylation region (TSDR) and increased proportions of CD4+FOXP3+ Tregs (P < 0.001) identified by flow cytometry. While treatment effects were consistent across participants, not all maintained C-peptide. Responders exhibited a transient rise in IL-6, IP-10, and TNF-α (P < 0.05 for all) 2 weeks after treatment and a durable CD4+ exhaustion phenotype (increased PD-1+KLRG1+CD57– on CD4+ T cells [P = 0.011] and PD1+CD4+ Temra MFI [P < 0.001] at 12 weeks, following ATG and ATG/G-CSF, respectively). ATG nonresponders displayed higher proportions of senescent T cells (at baseline and after treatment) and increased methylation of EOMES (i.e., less expression of this exhaustion marker).CONCLUSION Altogether in these exploratory analyses, Th1 inflammation-associated serum and CD4+ exhaustion transcript and cellular phenotyping profiles may be useful for identifying signatures of clinical response to ATG in T1D.TRIAL REGISTRATION ClinicalTrials.gov NCT02215200.FUNDING The Leona M. and Harry B. Helmsley Charitable Trust (2019PG-T1D011), the NIH (R01 DK106191 Supplement, K08 DK128628), NIH TrialNet (U01 DK085461), and the NIH NIAID (P01 AI042288).

A genetic variant in the gene PTPN22 (R620W, rs2476601) is strongly associated with increased risk for multiple autoimmune diseases and linked to altered TCR regulation and T cell activation. Here, we utilize Crispr/Cas9 gene editing with donor DNA repair templates in human cord blood-derived, naive T cells to generate PTPN22 risk edited (620W), non-risk edited (620R), or knockout T cells from the same donor. PTPN22 risk edited cells exhibited increased activation marker expression following non-specific TCR engagement, findings that mimicked PTPN22 KO cells. Next, using lentiviral delivery of T1D patient-derived TCRs against the pancreatic autoantigen, islet-specific glucose-6 phosphatase catalytic subunit-related protein (IGRP), we demonstrate that loss of PTPN22 function led to enhanced signaling in T cells expressing a lower avidity self-reactive TCR, but not a high-avidity TCR. In this setting, loss of PTPN22 mediated enhanced proliferation and Th1 skewing. Importantly, expression of the risk variant in association with a lower avidity TCR also increased proliferation relative to PTPN22 non-risk T cells. Together, these findings suggest that, in primary human T cells, PTPN22 rs2476601 contributes to autoimmunity risk by permitting increased TCR signaling and activation in mildly self-reactive T cells, thereby potentially expanding the self-reactive T cell pool and skewing this population toward an inflammatory phenotype.

Proprotein convertase subtilisin/kexin type 9 (Pcsk9) binds to hepatic low‐density lipoprotein receptor (LDLR) and induces its internalization and degradation. Pcsk9 inhibition increases LDLR expression by hepatocytes, which causes increased uptake of circulating LDL, thereby reducing plasma LDL‐cholesterol. However, by increasing the uptake of LDL by the liver, Pcsk9 inhibition increases the exposure of the liver to cholesterol, which may result in higher risk of steatohepatitis and ever carcinogenesis. We compared Pcsk9‐/‐ knockout (KO) mice and appropriate wild‐type (WT) controls of the same strain assigned to a high‐fat (15%, wt/wt) diet for 9 months supplemented with 0.25%, 0.5%, or 0.75% dietary cholesterol. Pcsk9 KO mice on a high‐fat, high‐cholesterol diet exhibited higher levels of hepatic free cholesterol loading and hepatic cholesterol crystallization than their WT counterparts. Pcsk9 KO mice developed crown‐like structures of macrophages surrounding cholesterol crystal‐containing lipid droplets and hepatocytes, exhibited higher levels of apoptosis, and developed significantly more hepatic inflammation and fibrosis consistent with fibrosing steatohepatitis, including 5‐fold and 11‐fold more fibrosis at 0.5% and 0.75% dietary cholesterol, respectively. When injected with diethylnitrosamine, a hepatic carcinogen, early‐in‐life Pcsk9 KO mice were more likely to develop liver cancer than WT mice. Conclusion: Pcsk9 KO mice on high‐cholesterol diets developed increased hepatic free cholesterol and cholesterol crystals and fibrosing steatohepatitis with a higher predisposition to liver cancer compared with WT mice. Future studies should evaluate whether patients on long‐term treatment with anti‐PSCK9 monoclonal antibodies are at increased risk of hepatic steatosis, steatohepatitis or liver cancer, while accounting for concurrent use of statins.

Elevated levels and enhanced sensing of the pro-inflammatory cytokine interleukin-6 (IL-6) are key features of many autoimmune and inflammatory diseases. To better understand how IL-6 signaling may influence human T cell fate, we investigated the relationships between levels of components of the IL-6R complex, pSTAT responses, and transcriptomic and translational changes in CD4+ and CD8+ T cell subsets from healthy individuals after exposure to IL-6. Our findings highlight the striking heterogeneity in mbIL-6R and gp130 expression and IL-6-driven pSTAT1/3 responses across T cell subsets. Increased mbIL-6R expression correlated with enhanced signaling via pSTAT1 with less impact on pSTAT3, most strikingly in CD4+ naïve T cells. Additionally, IL-6 rapidly induced expression of transcription factors and surface receptors expressed by T follicular helper cells and altered expression of markers of apoptosis. Importantly, many of the features associated with the level of mbIL-6R expression on T cells were recapitulated both in the setting of tocilizumab therapy and when comparing donor CD4+ T cells harboring the genetic variant, IL6R Asp358Ala (rs2228145), known to alter mbIL-6R expression on T cells. Collectively, these findings should be taken into account as we consider the role of IL-6 in disease pathogenesis and translating IL-6 biology into effective therapies for T cell-mediated autoimmune disease.

Understanding factors that impact CAR T cell expansion in the clinic is crucial to improving its therapeutic success. Here the authors document heterogeneity in the clonal dynamics of CAR-T cells by tracking individual clones using the endogenous TCR and integration sites and provide further insights into the role of transcriptional states in clonal kinetics.

Human islet antigen reactive CD4+ memory T cells (IAR T cells) play a key role in the pathogenesis of autoimmune type 1 diabetes (T1D). Using single-cell RNA sequencing (scRNA-Seq) to identify T cell receptors (TCRs) in IAR T cells, we have identified a class of TCRs that share TCRα chains between individuals (“public” chains). We isolated IAR T cells from blood of healthy, new-onset T1D and established T1D donors using multiplexed CD154 enrichment and identified paired TCRαβ sequences from 2767 individual cells. More than a quarter of cells shared TCR junctions between 2 or more cells (“expanded”), and 29/47 (~62%) of expanded TCRs tested showed specificity for islet antigen epitopes. Public TCRs sharing TCRα junctions were most prominent in new-onset T1D. Public TCR sequences were more germline like than expanded unique, or “private,” TCRs, and had shorter junction sequences, suggestive of fewer random nucleotide insertions. Public TCRα junctions were often paired with mismatched TCRβ junctions in TCRs; remarkably, a subset of these TCRs exhibited cross-reactivity toward distinct islet antigen peptides. Our findings demonstrate a prevalent population of IAR T cells with diverse specificities determined by TCRs with restricted TCRα junctions and germline-constrained antigen recognition properties. Since these “innate-like” TCRs differ from previously described immunodominant TCRβ chains in autoimmunity, they have implications for fundamental studies of disease mechanisms. Self-reactive restricted TCRα chains and their associated epitopes should be considered in fundamental and translational investigations of TCRs in T1D.

Clinical trials of biologic therapies in type 1 diabetes (T1D) aim to mitigate autoimmune destruction of pancreatic β cells through immune perturbation and serve as resources to elucidate immunological mechanisms in health and disease. In the T1DAL trial of alefacept (LFA3-Ig) in recent-onset T1D, endogenous insulin production was preserved in 30% of subjects for 2 years after therapy. Given our previous findings linking exhausted-like CD8+ T cells to beneficial response in T1D trials, we applied unbiased analyses to sorted CD8+ T cells to evaluate their potential role in T1DAL. Using RNA sequencing, we found that greater insulin C-peptide preservation was associated with a module of activation- and exhaustion-associated genes. This signature was dissected into 2 CD8 memory phenotypes through correlation with cytometry data. These cells were hypoproliferative, shared expanded rearranged TCR junctions, and expressed exhaustion-associated markers including TIGIT and KLRG1. The 2 phenotypes could be distinguished by reciprocal expression of CD8+ T and NK cell markers (GZMB, CD57, and inhibitory killer cell immunoglobulin-like receptor [iKIR] genes), versus T cell activation and differentiation markers (PD-1 and CD28). These findings support previous evidence linking exhausted-like CD8+ T cells to successful immune interventions for T1D, while suggesting that multiple inhibitory mechanisms can promote this beneficial cell state.

There are a limited number of stimuli-responsive biomaterials that are capable of delivering customizable dosages of a therapeutic at a specific location and time. This is especially true in tissue engineering and regenerative medicine applications, where it may be desirable for the stimuli-responsive biomaterial to also serve as a scaffolding material. Therefore, the purpose of this study was to engineer a traditionally non-stimuli responsive scaffold biomaterial to be thermally responsive so it could be used for on-demand drug delivery applications. Fibrin hydrogels are frequently used for tissue engineering and regenerative medicine applications, and they were functionalized with thermally labile oligonucleotide tethers using peptides from substrates for factor XIII (FXIII). The alpha 2-plasmin inhibitor peptide had the greatest incorporation efficiency out of the FXIII substrate peptides studied, and conjugates of the peptide and oligonucleotide tethers were successfully incorporated into fibrin hydrogels via enzymatic activity. Single-strand complement oligo with either a fluorophore model drug or platelet-derived growth factor-BB (PDGF-BB) could be released on demand via temperature increases. These results demonstrate a strategy that can be used to functionalize traditionally non-stimuli responsive biomaterials suitable for on-demand drug delivery systems (DDS).

Checkpoint inhibitors target the inhibitory receptors expressed by tumor-infiltrating T cells in order to reinvigorate an anti-tumor immune response. Therefore, understanding T cell composition and phenotype in human tumors is crucial. We analyzed by flow cytometry tumor-infiltrating lymphocytes (TILs) from two independent cohorts of patients with different cancer types, including RCC, lung, and colon cancer. In healthy donors, peripheral T cells are usually either CD4+ or CD8+ with a small percentage of CD4+ CD8+ DP cells (5%, reaching 30–50% of T cells in some patients). These DP T cells have an effector memory phenotype and express CD38, 4-1BB, and HLA-DR, suggesting antigen-driven expansion. In fact, TCR sequencing analysis revealed a high degree of clonality in DP T cells. Additionally, there were high levels of PD-1 and TIM-3 expression on DP T cells, which correlated with higher expression of PD-1 and TIM-3 in conventional single positive CD8 T cells from the same patients. These results suggest that DP T cells could be dysfunctional tumor-specific T cells with the potential to be reactivated by checkpoint inhibitors.

Globally, diarrheal diseases are a leading cause of death in children under five and disproportionately affect children in developing countries. Children who contract diarrheal diseases are rarely screened to identify the etiologic agent due to time and cost considerations associated with pathogen-specific screening and hence pathogen-directed therapy is uncommon. The development of biomarkers to rapidly identify underlying pathogens could improve treatment options and clinical outcomes in childhood diarrheal diseases. Here, we perform RNA sequencing on blood samples collected from children evaluated in an emergency room setting with diarrheal disease where the pathogen(s) present are known. We determine host response gene signatures specific to Salmonella, Shigella and rotavirus, but not E. coli, infections that distinguish them from each other and from healthy controls. Specifically, we observed differential expression of genes related to chemokine receptors or inflammasome signaling in Shigella cases, such as CCR3, CXCR8, and NLRC4, and interferon response genes, such as IFI44 and OASL, in rotavirus cases. Our findings add insight into the host peripheral immune response to these pathogens, and suggest strategies and limitations for the use host response transcript signatures for diagnosing the etiologic agent of childhood diarrheal diseases.

Combination therapies mediate drug synergy to improve treatment efficacy and convenience, leading to higher levels of compliance. However, there are challenges with their manufacturing as well as reduced flexibility in dosing options. This study reports on the design and characterization of a polypill fabricated through the combination of material jetting and binder jetting for the treatment of hypertension. The drugs lisinopril and spironolactone were loaded into hydrophilic hyaluronic acid and hydrophobic poly(ethylene glycol) (PEG) photocurable bioinks, respectively, and dispensed through a piezoelectric nozzle onto a blank preform tablet composed of two attachable compartments fabricated via binder jetting 3D printing. The bioinks were photopolymerized and their mechanical properties were assessed via Instron testing. Scanning electron microscopy (SEM) was performed to indicate morphological analysis. The polypill was ensembled and drug release analysis was performed. Droplet formation of bioinks loaded with hydrophilic and hydrophobic active pharmaceutical ingredients (APIs) was achieved and subsequently polymerized after a controlled dosage was dispensed onto preform tablet compartments. High-performance liquid chromatography (HPLC) analysis showed sustained release profiles for each of the loaded compounds. This study confirms the potential of material jetting in conjunction with binder jetting techniques (powder-bed 3D printing), for the production of combination therapy oral dosage forms involving both hydrophilic and hydrophobic drugs.

We identified gene expression signatures predicting responsiveness to a Kinesin-5 (KIF11) inhibitor (Kinesin-5i) in cultured colon tumor cell lines. Genes predicting resistance to Kinesin-5i were enriched for those from chromosome 20q, a region of frequent amplifi cation in a number of tumor types. siRNAs targeting genes in this chromosomal region identified AURKA, TPX2 and MYBL2 as genes whose disruption enhances response to Kinesin-5i. Taken together, our results show functional interaction between these genes, and suggest that their overexpression is involved in resistance to Kinesin-5i. Furthermore, our results suggest that patients whose tumors overexpress AURKA due to amplification of 20q will more likely resist treatment with Kinesin-5 inhibitor, and that inactivation of AURKA may sensitize these patients to treatment.

Enhancing pre-existing anti-tumor immunity leads to therapeutic benefit for some patients, but why some tumors are more immunogenic than others remains unresolved. We took a unique systems approach to relate patient survival to immune gene expression in >3,500 tumor RNAseq profiles from a dozen tumor types. We found significant links between immune gene expression and patient survival in 8/12 tumor types, with tumors partitioned by gene expression comprising distinct molecular subtypes. T/NK cell genes were most clearly survival-related for melanoma, head and neck, and bladder tumors, whereas myeloid cell genes were most clearly survival-related with kidney and breast tumors. T/NK or myeloid cell gene expression was linked to poor prognosis in bladder and kidney tumors, respectively, suggesting tumor-specific immunosuppressive checkpoints. Our results suggest new biomarkers for existing cancer immunotherapies and identify targets for new immunotherapies.

While immunotherapies are rapidly becoming mainstays of cancer treatment, significant gaps remain in our understanding of how to optimally target them, alone or in combination. Here we describe a novel method to monitor levels of immune cells and pathways in expression data from solid tumors using pre-defined groups or modules of co-regulated immune genes. We show that expression of an interconnected sub-network of type I interferon-stimulated genes (ISGs) in melanomas at the time of diagnosis significantly predicted patient survival, as did, to a lesser extent, sub-networks of T helper/T regulatory and NK/T Cytotoxic cell genes. As a group, poor prognosis tumors with reduced ISG and immune gene levels exhibited significant copy number loss of the interferon gene cluster located at chromosome 9p21.3. Our studies demonstrate a link between type I interferon action and immune cell levels in melanomas, and suggest that therapeutic approaches augmenting both activities may be most beneficial.

Mice lacking the p27(Kip1) Cdk inhibitor (Cdkn1b) exhibit increased susceptibility to lymphomas from the Maloney murine leukemia virus (M-MuLV), and exhibit a high frequency of viral integrations at Xpcl1 (Kis2), a locus on the X-chromosome. Xpcl1 encodes miR-106a~363, a cluster of microRNAs that are expressed in response to adjacent retroviral integrations. We report the first large-scale profile of microRNA expression in MuLV-induced lymphomas, in combination with microarray gene expression analysis. The source material was T-cell lymphomas induced by M-MuLV in p27(Kip1) knockout mice and normal thymus. Surprisingly, the overall levels of miRNA expression were equivalent in lymphomas and normal thymus. Nonetheless, the expression of specific microRNAs was altered in tumors. The miR-106a~363 miRNA were over-expressed in lymphomas, particularly those with viral integrations at the Xpcl1 locus. In contrast, p27(Kip1) deletion itself was associated with a different pattern of microRNA expression. Gene expression was dramatically altered in lymphomas, yet paralleled data from T-cell lymphomas induced by other mechanisms. Genes with altered expression in association with the p27(Kip1) null genotype were of similar functional classes to those associated with Xpcl1 integration, but with the opposite pattern of expression. Thus, the effect of p27(Kip1) deletion may be to oppose an anti-oncogenic effect of Xpcl1 rather than enhancing its oncogenic functions. A subset of miR-106a~363 target genes was consistently reduced in lymphomas with Xpcl1 integrations, particularly genes with cell cycle and immune functions. We identify four predicted target genes of miR-106a~363 miRNA, including N-Myc (Mycn), and the TGF-beta receptor (Tgfbr2) using 3'UTR reporter assays. Still, bioinformatic miRNA target predictions were poor predictors of altered gene expression in lymphomas with Xpcl1 integration. Confirmation of miR-106a~363 gene targeting relevant to the tumor phenotype requires in vivo validation, because only a subset of predicted targets are consistently reduced in tumors that overexpress miR-106a~363.

We identified gene expression signatures predicting responsiveness to a Kinesin-5 (KIF11) inhibitor (Kinesin-5i) in cultured colon tumor cell lines. Genes predicting resistance to Kinesin-5i were enriched for those from chromosome 20q, a region of frequent amplification in a number of tumor types. siRNAs targeting genes in this chromosomal region identified AURKA, TPX2 and MYBL2 as genes whose disruption enhances response to Kinesin-5i. Taken together, our results show functional interaction between these genes, and suggest that their overexpression is involved in resistance to Kinesin-5i. Furthermore, our results suggest that patients whose tumors overexpress AURKA due to amplification of 20q will more likely resist treatment with Kinesin-5 inhibitor, and that inactivation of AURKA may sensitize these patients to treatment.

Uncovering pathways underlying drug-induced toxicity is a fundamental objective in the field of toxicogenomics. Developing mechanism-based toxicity biomarkers requires the identification of such novel pathways and the order of their sufficiency in causing a phenotypic response. Genome-wide RNA interference (RNAi) phenotypic screening has emerged as an effective tool in unveiling the genes essential for specific cellular functions and biological activities. However, eliciting the relative contribution of and sufficiency relationships among the genes identified remains challenging. In the rodent, the most widely used animal model in preclinical studies, it is unrealistic to exhaustively examine all potential interactions by RNAi screening. Application of existing computational approaches to infer regulatory networks with biological outcomes in the rodent is limited by the requirements for a large number of targeted permutations. Therefore, we developed a two-step relay method that requires only one targeted perturbation for genome-wide de novo pathway discovery. Using expression profiles in response to small interfering RNAs (siRNAs) against the gene for peroxisome proliferator-activated receptor alpha (Ppara), our method unveiled the potential causal sufficiency order network for liver hypertrophy in the rodent. The validity of the inferred 16 causal transcripts or 15 known genes for PPARalpha-induced liver hypertrophy is supported by their ability to predict non-PPARalpha-induced liver hypertrophy with 84% sensitivity and 76% specificity. Simulation shows that the probability of achieving such predictive accuracy without the inferred causal relationship is exceedingly small (p < 0.005). Five of the most sufficient causal genes have been previously disrupted in mouse models; the resulting phenotypic changes in the liver support the inferred causal roles in liver hypertrophy. Our results demonstrate the feasibility of defining pathways mediating drug-induced toxicity from siRNA-treated expression profiles. When combined with phenotypic evaluation, our approach should help to unleash the full potential of siRNAs in systematically unveiling the molecular mechanism of biological events.

Small interfering RNAs (siRNAs), the guides that direct RNA interference (RNAi), provide a powerful tool to reduce the expression of a single gene in human cells. Ideally, dominant, gain-of-function human diseases could be treated using siRNAs that specifically silence the mutant disease allele, while leaving expression of the wild-type allele unperturbed. Previous reports suggest that siRNAs can be designed with single nucleotide specificity, but no rational basis for the design of siRNAs with single nucleotide discrimination has been proposed. We systematically identified siRNAs that discriminate between the wild-type and mutant alleles of two disease genes: the human Cu, Zn superoxide dismutase (SOD1) gene, which contributes to the progression of hereditary amyotrophic lateral sclerosis through the gain of a toxic property, and the huntingtin (HTT) gene, which causes Huntington disease when its CAG-repeat region expands beyond approximately 35 repeats. Using cell-free RNAi reactions in Drosophila embryo lysate and reporter assays and microarray analysis of off-target effects in cultured human cells, we identified positions within an siRNA that are most sensitive to mismatches. We also show that purine:purine mismatches imbue an siRNA with greater discriminatory power than other types of base mismatches. siRNAs in which either a G:U wobble or a mismatch is located in the "seed" sequence, the specialized siRNA guide region responsible for target binding, displayed lower levels of selectivity than those in which the mismatch was located 3' to the seed; this region of an siRNA is critical for target cleavage but not siRNA binding. Our data suggest that siRNAs can be designed to discriminate between the wild-type and mutant alleles of many genes that differ by just a single nucleotide.