SLAP keeps EPHB2-Src signaling caged in the colon, and its loss writes a drug-selection rule in organoids

What the work claims

This is a primary mouse-genetics and cell-biology study, published in eLife on 2026-09-08. It claims that the Src-like adaptor protein (SLAP) is an epithelial-intrinsic tumor suppressor in the colon, that it acts by binding and post-translationally restraining the receptor tyrosine kinase EPHB2, and that loss of SLAP creates a specific, druggable dependence: Slap-deficient colonic organoids and tumoroids grow fast, and that growth advantage is abolished by inhibiting Src-family kinases or EPHB2 itself.¹

The evidence ladder runs from whole animal to molecular interaction. Constitutive and tamoxifen-inducible, gut-specific Slap deletion both produced colonic hyperplasia, more stem-cell marker expression at crypt bases, and elevated proliferation. In the azoxymethane/DSS inflammation-driven cancer model, induced Slap loss significantly increased tumor numbers. At the bench, SLAP co-immunoprecipitated with EPHB2 in a manner requiring SLAP's SH2 and SH3 domains and the phosphorylated juxtamembrane tyrosines 596 and 602 of EPHB2; a binding-defective SLAP mutant lost all tumor-suppressive activity in human colorectal cancer cells. That mutant rescue experiment is what elevates this from association to mechanism.

How it works

SLAP is a small membrane-anchored adaptor, homologous to the Src N-terminus, long known as a negative regulator of receptor tyrosine kinase signaling in lymphocytes. The group's earlier work had found SLAP abundantly expressed in intestinal epithelium, reduced in about half of human colorectal cancer samples, and capable of promoting degradation of the RTK EPHA2. What was missing was genetic proof in the colon, and the identity of the operative RTK.¹

Here, loss of Slap raised EPHB2 protein without changing EPHB2 mRNA, the signature of post-translational control. EPHB2 phosphorylation rose, its association with active Src rose, and global protein tyrosine phosphorylation in crypts increased two- to threefold. The functional chain was verified pharmacologically in both directions. The Src inhibitor eCF506 abolished the enhanced growth of Slap-deficient organoids and tumoroids. The selective EPHB2 kinase inhibitor ALW-II-49-7 markedly suppressed growth of Slap-deficient organoids while leaving wild-type organoids untouched; a pan-EPH inhibitor gave the same pattern, and EPHB2 inhibition in turn cut the elevated phosphotyrosine and Src activation in the organoids, arguing that EPHB2 sits upstream of Src, not in parallel.

The human-cell experiments complete the axis. In SLAP-low HT29 and SW620 colorectal cancer cells, enforced SLAP expression suppressed tumoroid growth and the stem-cell marker ALDH, and simultaneously abolished sensitivity to EPHB2 inhibition. Re-expressing a SLAP mutant unable to bind EPHB2 failed to suppress growth and restored EPHB2-inhibitor sensitivity. In patient data, SLAP expression correlated weakly but significantly inversely with a cancer stem-cell activity score in TCGA tumors, and co-expression of SLAP or its relative SLA2 together with EPHB2 tracked with better prognosis, specifically in microsatellite-stable tumors, not in the highly immune-infiltrated microsatellite-unstable group.

Where a skeptic should push

The load-bearing assumption is that the SLAP-EPHB2-Src wiring demonstrated in engineered mice and two cell lines stratifies real human colorectal cancer. Three cautions. First, EPHB biology is context-dependent: EPHB2 helps position cells along the crypt axis and can behave as a tumor suppressor in some settings, so a kinase inhibitor that helps in a SLAP-null background could hurt, or do nothing, in other genetic contexts; the paper itself notes that other RTKs, including EGFR-family members, may compensate when SLAP is lost. Second, the mouse cancer model is inflammation-driven and characteristically RAS-mutated, and therein lies an instructive internal control: Src inhibition did nothing to tumoroids from control AOM/DSS mice, because those tumors are RAS-driven. That null result is honest science and simultaneously a warning that the dependence exists only in a defined subset, exactly the kind of subsetting clinical trials failed to do when Src inhibitors were given to unselected patients. Third, the prognosis associations are correlative co-expression signals in a specific molecular subtype (MSS); they are hypothesis-supporting, not proof that SLAP status predicts drug response in patients.

Mechanistically, how SLAP lowers EPHB2 abundance is still not nailed down. Degradation through CBL-family or related E3 ligases is plausible and consistent with prior SLAP biology, but the paper presents the interaction and the mutant epistasis, not the ubiquitination event. And the inhibitors used (eCF506, ALW-II-49-7, ALW-II-41-27) are tool compounds; nothing here tests whether clinically developable versions recapitulate the selectivity.

Organoids as gatekeepers for kinase inhibitors

The non-obvious contribution to organoid-based drug discovery is methodological, and it cuts against how tumor-organoid panels are usually run. The paper's logic is a rescue-with-inhibitor design: a genetic lesion creates a growth advantage in culture, and the inhibitor erases that advantage only in the lesioned background, with the wild-type or RAS-driven controls confirming the dependence is specific. That is the correct architecture for using organoids to decide which kinase inhibitor a patient should get, and this study shows the architecture working end to end, from mouse allele to human cell line to patient-expression data. The field has been burned once by the opposite design: Src is aberrantly active in about 50 percent of colorectal cancers, so it was drugged in everyone, and the inhibitors failed because activity without addiction is not a target. SLAP status is precisely the kind of upstream circuit-state marker that separates addiction from mere activity, and it is measurable in a tumoroid within days of a biopsy.

The opportunity: a prospective co-clinical design where patient-derived colorectal tumoroids are tested against Src-family and EPHB inhibitors with SLAP expression scored in parallel, using the paper's own stratification logic. The MSS-specific prognosis signal suggests the enriched population may be the immune-cold, microsatellite-stable tumors, which also happen to be the subset with the least benefit from current immunotherapy, so a biomarker-guided kinase strategy there would fill a real gap.

The threat is a familiar one at a smaller scale. If PDO panels run these drugs without the control conditions (is the pathway wired on in this tumor? is the tumor RAS-driven and therefore Src-independent?), the assay will produce the same false-negative and false-positive mixture that sank the clinical program, and it will look scientific because it is quantitative. Growth-only readouts compound the risk: EPHB2 has non-canonical, kinase-independent roles in cell positioning, so an expansion assay may miss biology that matters in vivo. Organoids here are gatekeepers, not oracles: they can reject a drug for a patient whose tumor will not respond, but only if the panel is built to ask the dependency question and not just the sensitivity question.

The bottom line

Established: in two independent mouse models and in human colorectal cancer cells, SLAP restrains EPHB2-Src signaling through a direct, domain-specific interaction, and loss of SLAP creates a Src-dependent and EPHB2-dependent growth advantage that inhibitors erase selectively. Established with lower confidence: SLAP-EPHB2 co-expression marks better prognosis in MSS colorectal cancer, hinting at the relevant patient subset. Hypothesis: SLAP-low tumors respond to EPHB or Src inhibition while SLAP-intact tumors do not. What would confirm it: a patient-derived organoid panel stratified by SLAP status with matched clinical response; what would break it: SLAP-low patient organoids that ignore EPHB2 inhibition, or evidence that in vivo EPHB2's context-dependent roles override the clean organoid readout.