By Deborah Borfitz 

June 8, 2022 | The pioneering OneK1K study in Australia has identified an immune “fingerprint” of seven autoimmune disorders using single-cell RNA sequencing (scRNA-seq). The general framework, which combines the scRNA-seq data with genotype data to classify individual cells, can be applied to many different diseases, including other autoimmune disorders, cardiovascular diseases, neuroinflammatory conditions, and cancer where the immune system is thought to play a role, according to Joseph Powell, director of cellular science at the Garvan Institute of Medical Research. 

Selection of the original seven diseases—multiple sclerosis, rheumatoid arthritis, lupus, type 1 diabetes, spondylitis, inflammatory bowel disease, and Crohn’s disease—were based on their prevalence in the world of autoimmune diseases and high genetic component, he says. The study relied on scRNA-seq data from 1.27 million peripheral blood mononuclear cells collected from 982 “healthy” donors, many of whom carry the genetic loci found in people who have these diseases. 

Take Crohn’s disease, which has molecular markers found on roughly 190 positions in the genome, Powell cites as an example. On average, patients collectively have roughly 90 risk alleles at those loci, but individually about 60. “It is the difference between having 60 and 90 that takes you over the threshold and leads to occurrences of disease.” 

This phenomenon holds true for most every disease afflicting humans, which is why a big population group is ideal for gaining mechanistic insights on disease-associated genes, he continues. At the cellular level, genome regulation changes brought on by those genes are the same even if there is no clinical manifestation of disease. 

Using a Mendelian randomization approach, Powell and his colleagues uncovered the causal route by which 305 loci contribute to autoimmune disease through changes in gene expression in specific cell types and subsets—and that these genetic mechanisms are the same for healthy individuals as in autoimmune disease cohorts. Results published recently in Science (DOI: 10.1126/science.abf30). 

Single-cell RNA sequencing was used to look at genetic variants affecting gene expression in 14 different immune cell types, says Powell. It is the largest study to date linking disease-causing genes to specific types of immune cells. 

Researchers developed a classification method based on the transcriptomic signature found in individual cells and aligned it back to what is currently understood about more common immune cell types at the top of the hierarchy (e.g., T cells, CD4, CD8, naïve to memory B cells). Although 68 immune cells have been classified, they focused on those they were sure to find enough copies of across the OneK1K cohort to confidently link the genetic differences between people to the signatures in the cells.  

Tissue-To-Tissue Variability 

Powell says he has been interested in genetic control of gene expression, and its contribution to disease, for more than a decade now. For many years, this involved bulk RNA analysis that produces an “average” signal. 

An important clue emerged when researchers began seeing how vastly different the genetics worked in one tissue versus another, he says. Only a few years ago, the Genotype-Tissue Expression Program (GTEx) of the National Institutes of Health examined RNA sequencing samples from 49 tissues of postmortem donors to characterize genetic associations for gene expression and found regulatory associations for almost all genes. Cell type composition was identified as a key factor in understanding gene regulatory mechanisms. 

That study showed instances where genetic effects were seen in one tissue and not another, or generated completely different effects, Powell notes. It was published in 2020 when scRNA-sequencing was just emerging as a staple technology—and Powell had just started his computational genomics laboratory at the Garvan Institute. 

“We were stuck with this interesting question: If we see these differences between tissues, and know the tissues are comprised of really distinct cell types with really specific functional roles, and the transcriptomic cell signatures are different, can we try to create a system to undo the genetics we saw in tissue inside a cell?” That led Powell and his colleagues to scale up their scRNA-sequencing efforts.  

At the time, the vast amount of generated transcript data generated from 1,000 individuals would have made scRNA-sequence data entirely cost-prohibitive to generate. But Powell helped pioneer a biometric technique to pool cells from multiple samples, as well as a method to analyze the transcripts of individual cells that solved the challenge of determining what portion of them would provide the most useful information in defining a cell type.  

Up until then, a few other groups had published studies using samples from perhaps 50 or 100 individuals suggesting signatures of disease, but they were all under-powered, says Powell. “They could show genetic differences between cells, but they were all too under-powered to link them to disease, or to resolve why there were differences between cells.” 

TenK10K Study 

Statistically speaking, disease fingerprints that capture the genetic heterogeneity of patients will never be fully defined, says Powell. But the OneK1K study has probably moved the needle from the 10th percentile to the 50th percentile on the saturation curve. 

Powell is now aiming for the 95th percentile with a TenK10K study that will be seeking to enroll 10,000 individuals and generate single cell data on about 50 million cells. The multi-year initiative will involve partnerships with multiple hospitals across Australia, and both a healthy population group and patients newly diagnosed with autoimmune disorders, cardiovascular disease, and cancer. 

Autoimmune diseases affect about one in 12 Australians, he notes. They are incurable and require lifelong treatments to minimize the damage. Patients often try many different treatments before finding one that works for them. 

“The genetic mechanisms are actually really generic,” he says. “You can learn a lot by linking what we see in [OneK1K study] data to what already know just about genetic positions in disease. Now, we’re taking those genetic positions from genome-wide association studies and will show mechanisms of action and specifically the cell types they are acting in.” 

Signal-Seeking 

Pure fundamental science is a major driver for the work, says Powell, “knowledge creation for human disease... and making all the data publicly available.” Since the Science paper published in April, he has been fielding multiple data requests daily.  

The more translational outcome is the possibility that the catalogue of genetic mechanisms will be useful in predicting which treatments will work best for individual patients, Powell says. To test that hypothesis, Powell’s lab is in the process of conducting a series of retrospective and prospective signal-seeking studies using currently marketed drugs—starting with immunotherapy treatment of cancer. 

If successful, patients will one day be able to get a “very cheap test costing literally tens of dollars [in Australia anyway]” to guide the clinical decision-making of their treating physician, Powell says. “We think that across a population we will be able to move the efficacy of a drug from, say, 30%—which is pretty common for immunotherapy and a lot of inhibitory drugs in autoimmune disease—to 50% or 60%.” Even a 10-percentile gain for a single drug would be remarkable in terms of patient impact, he adds.  

Along the way, the molecular mechanisms of disease are being unearthed at the cellular RNA level and that data could be shared with pharmaceutical companies to inform their early-stage drug development work, including which targets to take forward to phase 1 clinical trials, says Powell. Already, four pharma companies and one biotech have approached him about just such partnering opportunity, which would additionally aid them in the selection of patient groups for treatment trials.  

Scientific strides have been a team effort by many committed individuals, he says, crediting the 16 co-authors on the latest study hailing from Sydney, Hobart, Melbourne, Brisbane, and San Francisco. They consider themselves part of the larger, decade-long movement toward open science, marked by transparency, open communication, and access to the data and computer code used to reach conclusions.