CHRFAM7A-assigned transcript detection is enriched in a LAMP3+CCR7+ dendritic-cell state in esophageal squamous cell carcinoma

Technical Summary: CHRFAM7A-Transcript Enrichment in ESCC Dendritic Cells

Problem Statement
The study addresses the challenge of interpreting CHRFAM7A, a human-restricted fusion gene related to CHRNA7. Due to extensive sequence similarity between the two genes and the sparse nature of single-cell RNA sequencing (scRNA-seq) data, distinguishing CHRFAM7A-assigned transcripts from CHRNA7 is difficult. Previous literature has linked CHRFAM7A and the alpha7 nicotinic acetylcholine receptor to inflammatory regulation and macrophage polarization, but these findings are often derived from non-human models or bulk data where the two genes cannot be resolved. The authors sought to answer a narrower, testable question: In a large esophageal squamous cell carcinoma (ESCC) myeloid dataset, is CHRFAM7A-assigned transcript detection associated with a specific myeloid state after accounting for sample structure and technical depth, and can this matrix-level signal be validated by raw sequencing reads?

Methodology
The research employed a multi-layered analytical approach using public datasets (GSE160269 for scRNA-seq and GSE23400 for bulk RNA-seq):

1. Data Re-analysis and Annotation: The authors reanalyzed 19,273 myeloid cells from 64 ESCC samples (60 patients). Cells were manually annotated into specific states, including monocytes, tumor-associated macrophages (TAMs), conventional dendritic cells (cDC), plasmacytoid dendritic cells (pDC), and a specific LAMP3+CCR7+ dendritic cell state (referred to as tDC).
2. Statistical Modeling:
   • Mixed-Effects Models: A binomial generalized linear mixed model evaluated CHRFAM7A detection across myeloid subtypes, adjusting for sequencing depth (total UMI), mitochondrial percentage, and sample-level random intercepts.
   • Conditional Logistic Regression: To control for all sample-level characteristics, sample-stratified conditional logistic regression compared tDC cells against non-tDC myeloid cells within the same tumor.
   • Robustness Checks: Analyses were repeated with higher UMI thresholds (≥2 and ≥3 UMIs) and compared against 100 abundance-matched control genes to determine if the enrichment was unique to CHRFAM7A or a general property of sparse transcripts.
3. Raw-Read Technical Validation: To verify that matrix counts reflected actual biological sequences rather than mapping artifacts, raw reads from two specific tumor samples (P31T and P126T) were audited.
   • Reads were extracted for cells identified as LAMP3+CCR7+.
   • Sequences were classified using a "frozen" Q1-Q3 scheme based on identity to CHRFAM7A and CHRNA7 RefSeq transcripts.
   • Genomic alignments were checked for structural compatibility with annotated CHRFAM7A exons.
   • Concordance was measured between the presence of validated raw-read evidence and the original matrix-assigned counts.
4. Bulk Context: GSE23400 was used to provide paired tumor-normal context, though the authors noted that the available probes could not distinguish between CHRNA7 and CHRFAM7A.

Key Results

• Enrichment in tDC State: CHRFAM7A-assigned counts were detected in 8.19% of all myeloid cells. However, detection was highly concentrated in the LAMP3+CCR7+ dendritic cell state (tDC), where 30.55% (289/946) of cells were positive. This was significantly higher than in other myeloid subtypes (e.g., TAMs, cDCs, pDCs).
• Statistical Significance:
  • The mixed-effects model showed a strong subtype effect (Likelihood-ratio chi-square = 520.92, P < 1.49x10^-105).
  • Within-sample conditional logistic regression yielded an odds ratio (OR) of 5.04 (95% CI 4.39–5.78) for CHRFAM7A detection in tDCs versus other myeloid cells.
  • The enrichment persisted when comparing tDCs specifically to cDCs and pDCs (OR 4.35).
  • The pattern remained robust at higher UMI thresholds (≥2 and ≥3) and ranked at the 98th percentile among 100 abundance-matched control genes, suggesting the enrichment is not merely a generic artifact of low-abundance genes.
• Raw-Read Concordance:
  • In sample P31T, 17 of 20 matrix-positive tDC cells contained high-confidence raw-read sequences structurally compatible with CHRFAM7A, whereas 0 of 72 matrix-negative cells did.
  • In the rule-frozen replication sample P126T, 11 of 15 matrix-positive cells showed support, with 0 of 15 matrix-negative cells showing support.
  • Two identical 100-bp sequences were shared between the two samples, providing additional molecule-level concordance.
• Negative Findings: An exploratory analysis of a four-gene transcriptional program (ALOX15B, BIRC3, GAPDH, LPP) associated with CHRFAM7A detection within tDCs failed to meet replication criteria. Bulk data could not independently validate differential expression due to probe ambiguity.

Significance and Claims
The paper concludes that CHRFAM7A-assigned transcript detection shows a technically supported enrichment in a LAMP3+CCR7+ dendritic-cell state in ESCC. This finding is supported by sample-aware statistical models, robustness analyses, and two-sample raw-read technical concordance.

Crucially, the authors maintain a modest scope regarding biological interpretation:

• No Causality or Function: The study does not establish that CHRNA7 is expressed, that alpha7 receptor activity is present, that vagal signaling is active, or that CHRFAM7A causes the observed cellular state.
• Human-Specific Hypothesis: The results define a focused, human-specific hypothesis for future validation.
• Technical Distinction: The findings highlight the necessity of sequence-aware analysis to distinguish CHRFAM7A from CHRNA7, acknowledging that standard short-read mapping and bulk probes often fail to resolve these loci.

The authors emphasize that while the data robustly map a transcript-detection phenotype to a specific transcriptomic state, further work using long-read sequencing, junction-specific assays, and protein-level functional experiments is required to determine the biological mechanisms and relevance to tumor immunity.