Negative Selection

Summary

Negative selection (also called purifying selection) is the evolutionary process by which deleterious mutations — those that reduce fitness — are removed from a population. In cancer, negative selection operates in two distinct regimes: (1) intrinsic negative selection against mutations that impair essential cellular functions, which is weak and difficult to detect because most such mutations are lost at the single-cell stage before a clone can form; and (2) immune-mediated negative selection against clones bearing immunogenic neo-antigens, which is the dominant and clinically significant form — CD8+ T cells recognize neoantigen-MHC-I complexes and eliminate the presenting cells (Turajlic et al., 2019). Immune evasion — through HLA LOH, neoantigen depletion, or immunosuppressive signaling — is fundamentally escape from negative selection (McGranahan & Swanton, 2017). Negative selection is the least-studied of the three evolutionary forces in cancer, partly because its primary arena (the immune system) was not appreciated as an evolutionary selective pressure until the immunotherapy era, and partly because its signal is subtle: it manifests as an absence of expected mutations rather than an excess, making it harder to detect than positive-selection.

Definition

In population genetics, negative selection is the removal of deleterious alleles from a population. Mutations that reduce fitness are eliminated because carriers leave fewer descendants than non-carriers. The strength of negative selection depends on the fitness cost (|s|) and the dominance coefficient: recessive deleterious alleles can persist at low frequency because they are “hidden” from selection in heterozygotes; dominant deleterious alleles are eliminated more rapidly.

In cancer, the concept requires adaptation because the “population” is somatic — there is no sexual reproduction, no dominance, and the unit of selection is the cell, not the organism. A somatic mutation that kills the cell in which it arises is invisible — the cell dies, and the mutation is lost without trace. For negative selection to be observable, the mutation must persist long enough to generate a detectable signal in sequencing data, which requires that the fitness cost be small enough that the lineage survives for multiple cell divisions before elimination.

Mechanisms

Intrinsic Negative Selection

Mutations that impair essential cellular functions — DNA replication, transcription, translation, proteostasis, energy metabolism — reduce cell fitness and are eliminated. The dN/dS ratio (ratio of non-synonymous to synonymous substitution rates) is the standard tool for detecting this: a dN/dS < 1 indicates negative selection (non-synonymous changes are depleted relative to the neutral expectation), while dN/dS > 1 indicates positive selection.

In the cancer genome, dN/dS is close to 1 for most genes — most non-synonymous mutations are passengers, neither selected for nor against, because the fitness cost of a single amino acid change in most proteins is below the detection threshold in a somatic population. Only a small set of essential genes show dN/dS < 1 across large cohorts (PCAWG, 2020). This means intrinsic negative selection is real but weak — most mutations that would be strongly deleterious at the organismal level are either (a) heterozygous and recessive at the cellular level, or (b) compensated by the cell’s regulatory networks.

The practical consequence: passenger mutations are effectively neutral (s ≈ 0) for the purposes of clonal dynamics, even if the mutated protein has slightly reduced function. The fitness landscape of cancer is dominated by a vast neutral plateau punctuated by rare fitness peaks (drivers under positive selection) and occasional fitness valleys (strongly deleterious mutations that are eliminated before detection). See passenger-mutation.

Immune-Mediated Negative Selection

The dominant arena for negative selection in cancer is immunological, not cell-intrinsic. The adaptive immune system — specifically CD8+ cytotoxic T cells — recognizes neoantigen-MHC-I complexes on the tumor cell surface and eliminates the presenting cells. This is negative selection operating through an external agent: the T cell is the selective force, the neoantigen is the marker of “deleterious” (from the tumor’s perspective), and elimination is the outcome.

The process follows the three E’s of cancer immunoediting (Schreiber et al., 2011, as described in Turajlic et al., 2019):

Elimination → Equilibrium → Escape

1. Elimination. Immunogenic clones bearing strongly antigenic neo-antigens are recognized and destroyed by CD8+ T cells. This is classical negative selection — the immune system prunes the most visible clones. The neoantigens eliminated at this stage are predominantly clonal (present in all tumor cells) and conserved (not lost through copy-number changes), suggesting they are functionally constrained.

2. Equilibrium. Surviving clones — bearing fewer, less immunogenic, or subclonal neoantigens — persist but are held in check by residual immune pressure. Net tumor growth approaches zero. This is negative selection in balance: immune killing offsets cell division, and the population size is stable. This phase can last years and is clinically invisible (no detectable tumor growth).

3. Escape. Clones acquiring immune-evasion mechanisms are released from negative selection and expand. immune-evasion mechanisms — HLA LOH, B2M loss, neoantigen-coding segment deletion, PD-L1 upregulation — are adaptations that nullify the selective pressure rather than increasing intrinsic fitness. The escape clone may have identical proliferation and death rates to its competitors; its advantage is that it is no longer being killed by T cells.

Copy-Number-Mediated Escape

CNA provides a particularly efficient route to escaping negative selection: deletion of a chromosomal segment containing a neoantigen-coding mutation eliminates the antigen without requiring a second SNV at the epitope. A single CNA event can delete multiple neoantigens simultaneously — each residing on the same chromosomal segment — making CNA-based immune escape vastly more efficient than SNV-based antigen loss (which would require independent inactivating mutations at each epitope) (Turajlic et al., 2019).

This creates a dual role for CIN in immune selection: CIN generates neoantigens (through frameshifts and aberrant protein products, increasing immune visibility), and CIN also provides the genetic substrate for immune escape (through CNA-mediated HLA LOH and neoantigen deletion). The net effect depends on which process dominates — a question that varies by tumor type, CIN burden, and immune context. See chromosomal-instability and neo-antigen.

flowchart TD
    M["Somatic mutation<br>generates neoantigen"] --> Present["Neoantigen presented<br>on MHC-I"]

    Present --> Recog{"T-cell recognition?"}

    Recog -->|"Strongly immunogenic"| Elim["ELIMINATION<br>CD8+ T cells kill<br>neoantigen-bearing clone"]
    Recog -->|"Weakly/non-immunogenic"| Survive["Clone survives<br>No immune response"]

    Elim --> Result1["Negative selection:<br>immunogenic clones<br>removed from population"]

    Survive --> Accum["Clone accumulates<br>further mutations"]

    Accum --> Escape["ESCAPE MECHANISMS"]
    Escape --> E1["HLA LOH<br>Loss of one MHC-I allele"]
    Escape --> E2["Neoantigen depletion<br>CNA deletes antigen-coding<br>chromosomal segment"]
    Escape --> E3["B2M loss<br>Complete MHC-I loss<br>(NK-cell visible)"]
    Escape --> E4["PD-L1 upregulation<br>T-cell exhaustion"]

    E1 --> Expand["IMMUNE ESCAPE<br>Clone expands free from<br>T-cell-mediated killing"]
    E2 --> Expand
    E3 --> Expand
    E4 --> Expand

    Expand -->|"Was under negative selection"| Note["Escape = relief from<br>negative selection,<br>not intrinsic fitness gain"]

Figure: Immune-mediated negative selection and escape. Neoantigen-bearing clones are eliminated by CD8+ T cells (negative selection). Clones that acquire immune-evasion mechanisms — HLA LOH, neoantigen depletion, B2M loss, PD-L1 upregulation — escape this negative selection and expand. The escape clone’s fitness advantage is relief from immune predation, not increased intrinsic proliferation. Synthesized from Turajlic et al. (2019) and McGranahan & Swanton (2017).

Detection

Detecting negative selection in cancer genomes is harder than detecting positive selection for two reasons:

1. Negative selection is an absence. Positive selection produces an excess of mutations (in specific genes, at specific hotspots, at high VAF). Negative selection produces a deficit — fewer mutations than expected in a given region or gene. Statistical power to detect a deficit is lower than power to detect an excess for the same effect size.

2. The neutral null is uncertain. The expected mutation rate varies across the genome due to replication timing, chromatin state, and sequence context. A region with fewer mutations than average may be under negative selection, or it may simply have a lower background mutation rate. Disentangling these requires careful calibration of the regional mutation rate (PCAWG, 2020).

The primary tool is dN/dS (ratio of non-synonymous to synonymous mutations, normalized by the expected ratio under neutrality). A gene with dN/dS < 1 across a large cohort is under negative selection — non-synonymous mutations are depleted because cells carrying them are eliminated. In practice, only a handful of genes show statistically significant dN/dS < 1 in cancer (PCAWG, 2020), reflecting the weakness of intrinsic negative selection and the dominance of immune-mediated negative selection (which operates on neoantigens, not on specific genes).

For immune-mediated negative selection, the signal is neoantigen depletion: clones with high neoantigen burden are underrepresented in later-stage tumors compared to early-stage tumors, suggesting they were eliminated. The depletion is most pronounced for clonal neoantigens (present in all cells, consistently presented) and weakest for subclonal neoantigens (heterogeneous presentation, easier to escape). This pattern is consistent with negative selection acting most strongly on the most visible, most consistently presented antigens (McGranahan & Swanton, 2017).

Relationship to Other Evolutionary Forces

Force	Fitness effect	Signature	In cancer
Negative selection	s < 0	dN/dS < 1, neoantigen depletion	Weak intrinsically; strong immunologically
Positive selection	s > 0	dN/dS > 1, VAF enrichment	The primary driver of clonal expansion
Genetic drift	s ≈ 0	1/f² VAF distribution	The dominant force for most mutations

Negative selection and positive-selection are not symmetric in cancer. Positive selection is concentrated — a small number of driver genes are recurrently mutated, and the signal is strong. Negative selection is diffuse — it operates across the genome on any mutation that reduces fitness or generates an immunogenic peptide, and the signal is spread thin. This asymmetry reflects the biology: gaining a new function (oncogenic activation) requires specific mutations at specific sites, while losing fitness can happen through mutations at thousands of sites across hundreds of genes.

Clinical Significance

Immunotherapy Relieves Negative Selection

Immune checkpoint inhibitors (anti-CTLA4, anti-PD1/PD-L1) block the inhibitory signals that exhausted T cells receive, reactivating immune-mediated negative selection against neoantigen-bearing clones. The clinical response — tumor regression — is negative selection in action: T cells that were held in check now eliminate the clones they recognize.

This explains why mutational burden predicts immunotherapy response: more mutations → more neoantigens → more clones under negative selection pressure → more clones eliminated when that pressure is restored. It also explains why tumors with HLA LOH or B2M loss respond poorly — the machinery of negative selection (antigen presentation) is broken, so releasing the brakes on T cells has no effect.

Immune Evasion as an Evolutionary Biomarker

The presence of immune-evasion mechanisms — HLA LOH, neoantigen depletion, B2M loss — is evidence that negative selection was operating. A tumor that has undergone HLA LOH was, at some point in its history, under immune-mediated negative selection strong enough to favor the HLA-loss clone. Detecting these escape events retrospectively reveals the history of immune pressure even when the immune response is no longer active (McGranahan & Swanton, 2017).

Limitations

1. Detection asymmetry. Negative selection is harder to detect than positive selection because it produces a deficit, not an excess. Most genes under weak negative selection will appear neutral in cohort sizes typical of current studies.

2. Single-cell invisibility. Strongly deleterious mutations kill the cell in which they arise and leave no trace in bulk sequencing data. The negative selection we detect is, by definition, weak negative selection — strong enough to produce a statistical deficit in large cohorts, weak enough that the affected lineages survive long enough to be sampled.

3. Immune context dependence. Immune-mediated negative selection depends on the patient’s HLA type, T-cell repertoire, and tumor microenvironment. The same neoantigen may be strongly immunogenic in one patient (HLA-matched, T cells present) and invisible in another (HLA-mismatched, immunosuppressed). This makes negative selection harder to study systematically than positive selection, which operates through largely cell-intrinsic mechanisms.

4. Timing. Most immune-mediated negative selection occurs early — clonal neoantigens are more conserved and less immunogenic than subclonal neoantigens, consistent with early elimination of the most visible clones. Late-stage tumors have already passed through the elimination and equilibrium phases; what we observe is the escape phase. The negative selection that shaped the tumor’s early history must be inferred from its absence in the present.