Single‑Cell Data: Immune States and Lineages

Jonathan Alles

EVOBYTE Digital Biology

By EVOBYTE Your partner in bioinformatics

Introduction

Immunology is messy by design. The same immune cells that protect us can inflame tissues, shut down tumors, or quietly stand guard for decades. They do this by switching “states,” branching into lineages, and reacting to context that changes minute by minute. If you’ve ever stared at a bulk RNA-seq heatmap hoping to spot why one patient responded to a therapy and another didn’t, you’ve felt the limits of averaging across millions of cells. Important signals disappear in the mean.

Single‑cell technologies flip that perspective. Instead of one composite picture, you get a crowd scene where every cell speaks. With single‑cell RNA‑seq (scRNA‑seq), single‑cell ATAC‑seq (scATAC‑seq), CITE‑seq protein tagging, and TCR/BCR sequencing, you can separate cell types, resolve cell states, infer developmental paths, and even follow clonal families through disease and treatment. The result is not just a prettier plot. It’s a more faithful model of how the immune system works and, crucially, how to intervene.

In this post, we’ll unpack why immune systems are hard to model, how single‑cell profiling clarifies states and lineages, and where these insights translate into clinical trials and target discovery. Along the way, we’ll define core acronyms and show concise code to get started.

Why immune systems are hard to model with averages

Immune biology is dynamic, diversified, and distributed. Dynamic, because a T cell can progress from naïve to activated to effector to memory—or slide into exhaustion—over days, each step rewiring its transcriptome and chromatin. Diversified, because function isn’t purely typological. Two CD8+ T cells can share markers yet differ in cytotoxic potential or exhaustion depth. Distributed, because immune function emerges from interactions across cell neighborhoods and tissues: dendritic cells prime T cells, B cells exchange help with T follicular helpers, myeloid cells shape inflammation, and stromal cells set the stage.

Traditional bulk assays compress this heterogeneity. They measure average expression, which is fine when a homogeneous signal dominates. But in immunology, rare populations often matter most. Think of tumor‑infiltrating lymphocytes that make up a small fraction of a biopsy yet drive response to checkpoint blockade. Or pathogenic Th17 subsets in autoimmunity that are easy to drown out in a whole blood sample. Even flow cytometry, while single‑cell by nature, relies on predefined panels and a limited set of markers. You see what you label.

Single‑cell multiomics broadens that view. scRNA‑seq captures thousands of genes per cell, painting a transcriptomic portrait. CITE‑seq adds antibody‑derived tags to read surface proteins alongside RNA, resolving cases where mRNA and protein diverge. scATAC‑seq maps chromatin accessibility, which helps distinguish poised from active states and highlights transcription factors driving fate decisions. TCR/BCR sequencing links clonotypes—unique receptor sequences defining T- or B‑cell families—to phenotypes, so you can ask not just “what’s here?” but “what is this clone doing?” These modalities complement one another and, when integrated, reveal how immune cells move through state space and lineage space at once.

How single‑cell profiling delineates immune cell states and lineages

When people say “cell type,” they often mean a bucket like “CD4+ T cells.” Single‑cell data invites a different vocabulary: “state” describes what a cell is doing right now, while “lineage” describes where it came from and where it’s headed. In immunology, both matter.

A state is anchored in transcriptional and protein signatures. Activated CD8+ T cells upregulate cytotoxic genes, while exhausted cells express inhibitory receptors and distinct metabolic programs. Regulatory T cells carry FOXP3 and suppressive machinery, but they also exhibit tissue‑adapted states in the gut or tumor microenvironment. CITE‑seq helps confirm state by adding protein markers such as PD‑1, LAG‑3, or CD39 to the picture, which reduces misclassification when RNA and protein decouple.

Lineage connects snapshots into paths. Trajectory inference orders cells along a pseudotime axis, so you can see transitions like naïve to effector to memory, or naïve B cells to germinal center to plasma cells. Methods like RNA velocity estimate a short‑term “arrow of time” by comparing unspliced and spliced transcripts, which helps resolve bifurcations or transient intermediates that are easy to miss. scATAC‑seq adds regulatory context; if a branch toward exhaustion opens accessibility at AP‑1 and NFAT motifs, you can hypothesize which transcription factors steer fate. With TCR sequencing, you can overlay clonotypes onto these trajectories. When one clone appears across early and late states, it’s a living lineage tracer connecting cause and consequence.

These ideas are more than theory. In a tumor biopsy, single‑cell models routinely reveal gradients rather than discrete categories: cytotoxic cells shading into progenitor‑exhausted and terminal‑exhausted states. In infection or vaccination, B‑cell lineages display affinity maturation steps, with scATAC‑seq highlighting regulatory elements that flip on as cells transit the germinal center. In autoimmunity, monocyte subpopulations separate into inflammatory and resolving phenotypes, and their proportions track with flares and remission.

You don’t have to pick a single modality. Multimodal analysis lets you integrate scRNA‑seq with scATAC‑seq and proteins, reconcile batch effects across donors and timepoints, and project new samples into a reference atlas for consistent annotation. That consistency is vital in clinical contexts where you compare treated versus control arms or baseline versus on‑treatment biopsies across sites and timelines.

From cell states to clinical signals: designing and reading trials

Clinical trials in immunology live or die on endpoints that often lag biology. Response assessments in oncology, for example, rely on radiographic criteria weeks to months after dosing. Single‑cell readouts add earlier, mechanism‑anchored signals that can steer decisions mid‑trial and sharpen interpretation at the end.

Imagine a phase 2 study testing an immune checkpoint inhibitor in solid tumors. Baseline and on‑treatment biopsies are profiled by scRNA‑seq with CITE‑seq and TCR‑seq. At baseline, you might quantify the fraction of progenitor‑exhausted CD8+ T cells—those that co‑express TCF7 and intermediate levels of inhibitory receptors—relative to terminally exhausted cells. After dosing, you track clonal expansion and phenotypic shifts. Responders often show expansion of specific clonotypes that migrate toward effector‑like or reinvigorated states, accompanied by antigen presentation and interferon‑response signatures in myeloid and tumor cells. Non‑responders may enrich for myeloid‑derived suppressor cells, Tregs with high suppressive markers, or terminal exhaustion with high TOX and epigenetic commitment. These patterns provide pharmacodynamic evidence aligned to mechanism rather than waiting for tumor shrinkage.

Timing matters. Early on‑treatment samples—say day 8 or day 15—can reveal whether the drug engages the right nodes. A lack of shift in T‑cell states or persistent immunosuppressive myeloid programs is an early warning. Conversely, emerging inflammatory programs can flag immune‑related adverse events before they peak. Because single‑cell data preserves cellular proportions and states, you can distinguish true upregulation from compositional changes. That distinction prevents false positives when one population expands without per‑cell changes.

Trials in autoimmunity benefit similarly. In ulcerative colitis, for example, single‑cell profiling of colon biopsies can quantify pathogenic Th17‑like states, tissue‑resident memory T cells, and activated fibroblasts at baseline, then monitor how each compartment changes under therapy. If a drug reduces inflammatory monocytes and restores epithelial repair programs, that’s stronger evidence of disease modification than a symptom score alone. In vaccines, blood single‑cell data can capture early plasmablast bursts, germinal center‑like B‑cell states, and T follicular helper activation that predict downstream neutralizing titers, enabling faster iteration on dose and schedule.

For teams building trial analytics, the practical challenge is turning high‑dimensional single‑cell data into decision‑quality summaries. That means prespecifying cell states and lineage signatures, harmonizing annotations across batches and sites, and setting thresholds that are clinically interpretable. It also means planning for missingness and variability in tissue sampling. Techniques like reference mapping, batch‑aware integration, and probabilistic labeling help translate thousands of features into a concise panel of state scores and clonal metrics you can trend over time.

Target discovery with confidence: connecting states, regulators, and specificity

Target discovery is, at heart, a filtering problem. You want a lever you can pull—an antigen, receptor, enzyme, or transcription factor—that moves a disease‑relevant cell state toward a healthier one, without unacceptable off‑target effects. Single‑cell data sharpens each step of that filter.

First, it helps localize the problem. If pathogenesis centers on a specific cell state—say, a suppressive myeloid subset expressing ARG1 and high IL‑10, or a terminally exhausted CD8+ subset locked by certain chromatin features—you can prioritize targets that are upregulated or essential in that state. Co‑analyzing scATAC‑seq links those state markers to upstream regulators. If accessibility at ETS or IRF motifs spikes in the pathogenic state, the regulators behind those motifs rise in your shortlist. This creates hypotheses that connect observable states to causal knobs.

Second, it tests specificity. Single‑cell atlases across tissues and donors provide the backdrop to check where else your target appears. If a candidate surface protein is abundant on the pathogenic state but scarce on critical healthy cells in heart or brain, risk falls. If a transcription factor is widely expressed in hematopoietic stem cells, risk rises. Single‑cell resolution turns anecdotal “safe enough” into quantifiable expression profiles across many contexts, which is especially helpful in toxicology and safety pharmacology.

Third, it links clones to function. For T‑cell or B‑cell targets, pairing phenotype with TCR/BCR sequences shows whether disease‑relevant clonotypes depend on the target. If expanded pathogenic clones consistently express a particular receptor or rely on a cytokine axis, modulating that axis is more likely to change outcomes. Conversely, if protective clones also rely on the same axis, you can foresee trade‑offs before you design the molecule.

Finally, it accelerates validation. Because single‑cell data yields state scores and regulatory hypotheses, you can test them in organoids, ex vivo cultures, or mouse models while measuring the exact state signature you aim to fix. If an antibody against a myeloid receptor reduces the suppressive signature and restores antigen presentation in human tumor slices, you have a crisp mechanism readout to pair with efficacy.

A pragmatic example ties these threads together. Suppose a tumor study reveals a progenitor‑exhausted CD8+ pool that predicts response, a terminal‑exhausted pool that resists reinvigoration, and a suppressive myeloid niche that blunts T‑cell function. A target that tips the balance—perhaps a receptor enriched in the myeloid subset that constrains antigen presentation—would move up your list. scATAC‑seq highlights that the receptor’s promoter is accessible only in that myeloid state, de‑risking off‑tumor effects. CITE‑seq confirms protein expression, and spatial transcriptomics shows the myeloid cells cluster near T cells at the invasive margin. Those layers make a compelling, testable story that guides therapeutic design, patient selection, and combination strategy.

Summary / Takeaways

Immune systems are complex because they are meant to be. Cells adapt in real time, branch into alternative fates, and coordinate across tissues. Bulk assays blur those nuances, but single‑cell multiomics restores them, revealing states, lineages, and interactions that directly inform decisions.

If you’re running trials, single‑cell readouts give you early, mechanism‑anchored markers of engagement and response, help explain heterogeneous outcomes, and surface safety signals before they escalate. If you’re discovering targets, single‑cell data localizes disease‑relevant states, exposes upstream regulators, quantifies specificity across tissues, and links clonal behavior to function.

The practical path is straightforward. Define the states you care about, integrate modalities to anchor those states in both expression and regulation, and build analysis plans that turn complex measurements into stable, interpretable scores. Then connect those scores to clinical outcomes and experimental perturbations. When you do, the immune system’s messiness becomes an asset rather than an obstacle, because it gives you more handles to grab and more routes to efficacy.

What state, lineage, or interaction do you most need to see clearly in your program this quarter? Start there, choose the right single‑cell modalities, and design your analyses to make that signal unmissable.