Divergent Transcription in the BDNF Locus: Two Novel Genes That May Rewrite Our Understanding of BDNF Regulation and Schizophrenia Risk
New findings from the Lieber Institute challenge long-held assumptions about the BDNF locus and introduce a compelling regulatory framework with direct psychiatric relevance
For decades, the BDNF locus has been one of the most intensively studied regions in neuropsychiatric genetics. The gene’s complex promoter architecture, at least nine unique promoters splicing to a single coding exon, has made it both a fascinating model of transcriptional regulation and a persistent source of interpretive challenges in disease association studies. Add in the antisense gene BDNF-AS, the heavily scrutinized Val66Met polymorphism (rs6265), and conflicting GWAS signals, and you have a locus that has generated more questions than answers.
A new preprint from Martinowich, Ursini, and colleagues at the Lieber Institute for Brain Development introduces a finding that adds meaningful complexity, and, importantly, meaningful clarity, to this picture. Using postmortem human brain RNA sequencing data, the team has identified and rigorously characterized two previously unannotated genes in the BDNF locus: BDNF-DT, a divergent transcript gene, and BDNF-AS-DT, a readthrough gene formed by splicing between BDNF-AS and BDNF-DT. Their data suggest these are not transcriptional noise, they are developmentally regulated, activity-dependent, and genetically associated with schizophrenia ris
How the Genes Were Found and Characterized
The discovery began with a careful examination of a large Poly-A+ RNAseq dataset from postmortem dorsolateral prefrontal cortex (DLPFC) tissue (Jaffe et al., 2018), using IGV to identify reads in the 5′ upstream region of BDNF exon I that aligned to the positive DNA strand, antisense to BDNF, which is transcribed from the negative strand. Critically, these reads did not correspond to any annotated transcript in GRCh38/hg38.
The team validated the novel sequences through RT-PCR on independent DLPFC samples, followed by 5′ and 3′ RACE-PCR to define transcript boundaries. End-to-end PCR, nested PCR confirmation, and PacBio long-read sequencing of amplicons produced a high-confidence structural characterization of three BDNF-DT isoforms (BDNF-DT1, 2, and 3) and one BDNF-AS-DT readthrough transcript, the latter formed by splicing of BDNF-AS exons with BDNF-DT exons, notably skipping the BDNF-AS exon that contains rs6265.
Conservation analysis confirmed the existence of homologous divergent transcription in the mouse Bdnf locus, with four murine Bdnf-DT transcripts identified. The authors also make a compelling case that what was previously annotated as mouse Bdnf-AS (Modarresi et al., 2012) is in fact part of the Bdnf-DT gene — a reclassification with meaningful implications for prior in vivo antisense inhibition studies.
Developmental Trajectory, Activity Dependence, and CRISPR Validation
Several findings stand out as particularly compelling. First, both BDNF-DT and BDNF-AS-DT show strong developmental regulation in human DLPFC (N=296), with expression trajectories that closely mirror BDNF itself, rising through prenatal and early postnatal life, dipping at the childhood-to-adolescence transition, peaking again around age 25, and declining through adulthood. This stands in clear contrast to BDNF-AS, which increases during adulthood and senescence, consistent with established antisense RNA aging patterns.
In adult donors, BDNF-DT expression is positively correlated with BDNF (r=0.168, p=0.006) but uncorrelated with BDNF-AS (p=0.265). BDNF-AS-DT shows a negative correlation with BDNF (r=-0.17, p=0.036), pointing toward distinct regulatory relationships among these transcripts that are not captured by existing locus models.
The activity-dependence data are equally informative. In KCl-depolarized mouse primary cortical neurons, Bdnf-DT induction followed BDNF transcript upregulation with a temporal delay — peaking at 6 hours post-washout compared to 3 hours for Bdnf I-IX, suggesting a secondary response consistent with a feedback regulatory role. In human ReNcell VM neural progenitor cells, BDNF-DT showed an 11-fold increase post-washout, with BDNF-AS-DT also elevated. Mechanistically, CRISPRa-mediated upregulation of BDNF exon I and exon IV transcription in VPR-ReNcell VM cultures drove significant increases in BDNF-DT (>2-fold and >3-fold, respectively), providing direct evidence that BDNF transcription induces its own divergent transcript, a potential autoregulatory loop.
Why This Matter for the Field
The implications for BDNF biology are substantial. The authors propose a regulatory model in which BDNF transcription induces BDNF-DT, which in turn facilitates splicing of BDNF-AS with BDNF-DT to generate BDNF-AS-DT, a readthrough transcript that may function as a negative regulator of BDNF. This framework is consistent with the activity-dependent temporal dynamics, the positive BDNF–BDNF-DT correlation, the negative BDNF–BDNF-AS-DT correlation, and the discordant co-expression patterns observed in RRHO2 analyses.
For the schizophrenia field specifically, the genetic findings are striking. Summary-based Mendelian Randomization (SMR) analysis identifies BDNF-AS-DT as having the most significant potential causal association with schizophrenia in the BDNF locus, with lower genetically predicted expression associated with disease risk (p=9.89e-06 whole sample; p=7.64e-05 European ancestry). The eQTL analysis shows that rs6265 is strongly associated with BDNF-AS-DT expression (t=10.365, p=3.63e-22), more strongly than the top GWAS-index SNP rs4923457, consistent in both neurotypical controls and patients.
The mechanistic interpretation is nuanced and worth highlighting: the BDNF-AS-DT transcript skips the exon containing rs6265, which is present in all canonical BDNF-AS isoforms. The C/Val-coding allele is associated with higher CpG methylation at chr11:27658369 (the rs6265 position), lower BDNF-AS-DT expression, and higher BDNF-AS expression, suggesting allele-specific splicing effects mediated at least in part through intragenic methylation. This reframes the clinical correlates of rs6265 considerably, given that most mechanistic studies of Val66Met have focused exclusively on activity-dependent BDNF secretion in genomic contexts that lack BDNF-AS, BDNF-AS-DT, and BDNF-DT.
What This Work Opens Up
Several high-priority research directions follow naturally from these findings.
First, loss-of-function and gain-of-function studies targeting BDNF-DT and BDNF-AS-DT in relevant neural models will be essential to test the proposed autoregulatory feedback hypothesis and to determine whether these transcripts modulate BDNF protein output, secretion, or downstream TrkB signaling.
Second, single-cell resolution analyses, particularly in developing human cortex, would clarify which cell types drive the observed developmental trajectories and whether the expression patterns are layer- or lineage-specific.
Third, the allele-specific splicing model for rs6265 warrants direct experimental interrogation, ideally using isogenic iPSC-derived neurons carrying Val/Val versus Met/Met genotypes in a genomic context that preserves the full locus architecture.
Finally, the observation that BDNF-DT co-expression networks are enriched for REST/NRSF target genes opens an interesting avenue: if REST directly regulates BDNF-DT through the NRSE element in BDNF exon II, this could place these divergent transcripts within a broader activity-dependent transcriptional program relevant to multiple neuropsychiatric conditions.
Conclusion
This study does what the best genomic discoveries do: it takes a locus we thought we understood and reveals that there is significantly more to the story. The identification of BDNF-DT and BDNF-AS-DT as developmentally regulated, activity-dependent, and schizophrenia-associated transcripts adds a previously invisible regulatory layer to one of neuroscience’s most important genes. For researchers working on BDNF biology, psychiatric genetics, lncRNA function, or transcriptional regulation in the brain, this paper is essential reading, and the questions it raises are as exciting as the answers it provides.
We’ve created an accompanying article that focuses on the big picture and real-world impact of this research, without the technical details.
Dr. Kristen Maynard leads the Molecular Neuroanatomy Team at the Lieber Institute for Brain Development. Her work focuses on exploring gene expression, brain structure, and brain function in relation to psychiatric diseases.
We’ve created an accompanying article that focuses on the big picture and real-world impact of this research, without the technical details.
Dr. Kristen Maynard leads the Molecular Neuroanatomy Team at the Lieber Institute for Brain Development. Her work focuses on exploring gene expression, brain structure, and brain function in relation to psychiatric diseases.