Objectives Central nervous system (CNS) metastasis is a major driver of morbidity in metastatic breast cancer, yet the molecular determinants of CNS tropism remain incompletely defined. LYN, a Src-family kinase integrating receptor tyrosine kinase and integrin signaling, is a biologically plausible mediator of metastatic traits. Design We performed a retrospective, multi-study analysis of publicly available breast cancer cohorts aggregated in cBioPortal. After harmonization and de-duplication, LYN status was determinable in 5,947 invasive carcinoma of no special type (NST) tumors across 29 studies. The primary endpoint was CNS metastasis at any time (Yes/No), harmonized via a prespecified controlled vocabulary (case-insensitive substring mapping). Somatic LYN variants (coding SNVs/indels) were collapsed to patient-level classes (missense-only; truncating if any nonsense/frameshift/splice). Variants with resolvable positions were mapped to Src-family modules (SH4/Unique, SH3, SH2, SH2-kinase linker, kinase). Two-group comparisons used two-sided Fisher’s exact tests with exact 95% CIs; domain screens used omnibus χ² and Benjamini–Hochberg FDR control. A prespecified Firth logistic model evaluated truncating vs. missense within LYN-mutant tumors.  Setting Public cancer genomics repositories (cBioPortal); multi-institutional cohorts. Participants 5,947 tumors across multi-study cohorts with LYN status available. Interventions None. Main outcome measures Primary: ever-CNS metastasis (yes/no). Secondary: distribution of LYN variant classes and domains (SH4/Unique, SH3, SH2, linker, kinase).  Results CNS metastasis occurred in 5/46 (10.9%) LYN-mutated tumors vs. 110/5,901 (1.9%) LYN wild-type tumors (odds ratio (OR) = 6.42; 95% confidence interval (CI), 2.49–16.56; p = 0.0018). The endpoint was captured as ever vs. never CNS involvement (event dates unavailable), precluding time-to-event inference. Within LYN-mutant cases, an exploratory domain analysis indicated that distributions differed by CNS status (omnibus χ² p ≈ 0.014); a one-versus-rest signal at the SH4/Unique N-terminus was nominally significant and borderline after false discovery rate (FDR) (unadjusted p ≈ 0.010; q ≈ 0.052; small in-domain n = 3). By mutation class, truncating vs. missense showed a higher CNS-positive proportion (28.6% vs. 7.9%) but did not reach significance (Fisher p = 0.166; alternatively framed OR = 4.22; exact 95% CI, 0.58–30.75; p = 0.182). Firth estimates were directionally consistent with wide profile CIs under sparse counts. Conclusions Across pooled cohorts, LYN mutation is associated with increased odds of CNS metastasis, and domain context appears informative, with a small-sample, FDR-borderline enrichment at the SH4/Unique N-terminus. The truncating-class signal is exploratory given limited power. Signals by domain (notably SH4/Unique) are exploratory and require independent validation in larger, uniformly annotated datasets. Given small mutant denominators and ever-CNS endpoint capture, findings are hypothesis-generating and not actionable for risk stratification or treatment selection. Results motivate domain-aware annotation in future validation studies and mechanistic work.

LYN mutations in breast cancer: hypothesis-generating evidence for an association with central nervous system metastasis and domain-level insights

期刊论文
出版模式：

开放获取
发表日期：

2025年12月5日

查看全文

Elif Kardelen Çağdaş,
Berkay Çağdaş

摘要

Objectives

Central nervous system (CNS) metastasis is a major driver of morbidity in metastatic breast cancer, yet the molecular determinants of CNS tropism remain incompletely defined. LYN, a Src-family kinase integrating receptor tyrosine kinase and integrin signaling, is a biologically plausible mediator of metastatic traits.

Design

We performed a retrospective, multi-study analysis of publicly available breast cancer cohorts aggregated in cBioPortal. After harmonization and de-duplication, LYN status was determinable in 5,947 invasive carcinoma of no special type (NST) tumors across 29 studies. The primary endpoint was CNS metastasis at any time (Yes/No), harmonized via a prespecified controlled vocabulary (case-insensitive substring mapping). Somatic LYN variants (coding SNVs/indels) were collapsed to patient-level classes (missense-only; truncating if any nonsense/frameshift/splice). Variants with resolvable positions were mapped to Src-family modules (SH4/Unique, SH3, SH2, SH2-kinase linker, kinase). Two-group comparisons used two-sided Fisher’s exact tests with exact 95% CIs; domain screens used omnibus χ² and Benjamini–Hochberg FDR control. A prespecified Firth logistic model evaluated truncating vs. missense within LYN-mutant tumors.

Setting

Public cancer genomics repositories (cBioPortal); multi-institutional cohorts.

Participants

5,947 tumors across multi-study cohorts with LYN status available.

Interventions

None.

Main outcome measures

Primary: ever-CNS metastasis (yes/no). Secondary: distribution of LYN variant classes and domains (SH4/Unique, SH3, SH2, linker, kinase).

Results

CNS metastasis occurred in 5/46 (10.9%) LYN-mutated tumors vs. 110/5,901 (1.9%) LYN wild-type tumors (odds ratio (OR) = 6.42; 95% confidence interval (CI), 2.49–16.56; p = 0.0018). The endpoint was captured as ever vs. never CNS involvement (event dates unavailable), precluding time-to-event inference. Within LYN-mutant cases, an exploratory domain analysis indicated that distributions differed by CNS status (omnibus χ² p ≈ 0.014); a one-versus-rest signal at the SH4/Unique N-terminus was nominally significant and borderline after false discovery rate (FDR) (unadjusted p ≈ 0.010; q ≈ 0.052; small in-domain n = 3). By mutation class, truncating vs. missense showed a higher CNS-positive proportion (28.6% vs. 7.9%) but did not reach significance (Fisher p = 0.166; alternatively framed OR = 4.22; exact 95% CI, 0.58–30.75; p = 0.182). Firth estimates were directionally consistent with wide profile CIs under sparse counts.

Conclusions

Across pooled cohorts, LYN mutation is associated with increased odds of CNS metastasis, and domain context appears informative, with a small-sample, FDR-borderline enrichment at the SH4/Unique N-terminus. The truncating-class signal is exploratory given limited power. Signals by domain (notably SH4/Unique) are exploratory and require independent validation in larger, uniformly annotated datasets. Given small mutant denominators and ever-CNS endpoint capture, findings are hypothesis-generating and not actionable for risk stratification or treatment selection. Results motivate domain-aware annotation in future validation studies and mechanistic work.