top of page


Pumilio1: targets vs. interactors

PUMILIO1 (PUM1) is an RNA-binding protein, first discovered for its function in spermatogenesis and germ cell lines. We found that PUM1 is also crucial in the brain (Cell, 2015). We then found that PUM1 mutations cause disease in humans. Interestingly enough, loss-of-function mutations of PUM1 underlie very distinct phenotypes: a late-onset, mild ataxia and a severe neurodevelopmental disorder with cognitive deficits and seizures (Cell, 2018).  These are both considered forms of SCA47, but for convenience we call the first PRCA (PUM1-related cerebellar ataxia, or SCA47 OMIM #617931) and the second PADDAS (PUM1-associated developmental delay and seizures, OMIM #620719). Although the severity of the PRCA and PADDAS phenotypes tracks with the levels of functional PUM1 (~75% vs. 50% of wild-type levels), precisely what is happening at the molecular level was unclear (Cell, 2018). Especially puzzling was the fact that the mutation that causes the mild disease prevents PUM1 from binding to RNA (therefore it can't bind to its targets), but the severe mutation doesn't disrupt RNA binding.  Yet known PUM1 targets seemed to be upregulated to the same degree in both cases. We hypothesized that PRCA is caused by deregulation of PUM1 targets, but that the more severe and wide-ranging symptoms of PADDAS result from disruption of PUM1’s native interactions with other proteins, along with de-repression of the targets of these complexes. We therefore set out to understand PUM1’s activities in neurons by identifying its native partners using in vivo proteomics (EMBO J, 2023). Our findings not only lend support to our hypothesis but demonstrate that we need to look at RBP interactions as well as targets.

Gennarino et al., Cell 2018

Botta et al., EMBO J 2023

Alternative Polyadenylation in Neurological Diseases

As a postdoc I was interested in the regulation of MECP2, the gene that is mutated in Rett syndrome. MeCP2 is peculiar for having a very long (~8.5 kb) 3’UTR that contains two prominent poly-adenylation (p(A)) sites: a proximal p(A) and a distal p(A), resulting, respectively, in short and long messenger mRNA isoforms. We found that MeCP2 protein levels are regulated by the CF1m protein complex through these APA sites. We then identified eleven individuals with neuropsychiatric disease who have copy number variations (CNVs) spanning NUDT21, which encodes CFIm25. These individuals suffer from autism spectrum disorder and notable developmental regression, very similar to Rett syndrome. We thus discovered that NUDT21 duplication causes a new autism spectrum disorder (eLife, 2015).

Subsequently, my independent lab began studying CPSF6, another protein in the CF1m complex. We identified human subjects with mutations in the gene, and studied patient cell lines in parallel with a zebrafish deletion model, with the help of CUIMC pediatric cardiologist Kimara Targoff, MD. Along with collaborators Marko Jovanovic (Columbia University), Eric Wagner (Univ. of Rochester), and Hari Yalamanchili (Neurological Research Institute and Baylor College of Medicine), we analyzed the pattern of APA site choice throughout the zebrafish larvae and discovered that loss of CPSF6 function has quite different effects in brain, skeletal, and cardiovascular tissues (de Prisco et al., Science Advances, 2023).

One of the surprising results of this project was the finding that APA governs protein dosage. This observation was made possible because we looked beyond the 3' UTR to include the whole transcript (pA sites actually exist throughout the transcript, though intronic and exonic PAS have not been much studied). We found that the APA machinery chooses binding sites in the gene body or in the 3'UTR according to how much protein the cell needs at that particular moment. During an organ's growth, cells need to produce a large number of proteins as foundational building blocks; once growth is achieved, the cells no longer need to expend so much energy making new proteins. Given that the APA process is sensitive to input from hormones and other extracellular molecular cues, however, this work raises the possibility that anything altering APA site selection could cause disease or dysfunction, such as abnormal heart or skeletal growth or subtle forms of intellectual disability.

Gennarino et al., eLife 2015

de Prisco et al., Science Advances 2023

RNA Regulatory Networks in the Brain

MicroRNAs (miRNAs) are a class of single-stranded RNAs (ssRNAs), 19-25 nucleotides (nt) in length. They control the expression levels of their target genes through an imperfect pairing with target messenger RNAs (mRNAs), mostly in their 3' untranslated regions (3' UTRs). miRNAs are implicated in a wide range of basic biological processes and human diseases. Given that the brain, in both mice and humans, expresses a large spectrum of distinct miRNAs that other organs, it is likely that dysregulation of miRNA networks has an outsized effect on neurological function. The challenge is identifying each miRNA's set of target genes. Existing methods each operate under certain assumptions, and each arrives at a different set of targets, with little overlap among different tools. We demonstrated that it is possible to identify miRNA target genes by looking at their patterns of expression, and we developed two tools that outperform other current tools in the identification of miRNA targets: HOCTAR (Genome Research, 2009; Gene, 2011) and CoMeTa (Genome Research, 2012). Both tools have been already used to identify miR-128 in controlling the lysosomal master gene TFEB (Science, 2009) and miR-483-5p, which regulate the levels of MECP2, the gene mutated in Rett syndrome (Genes & Development, 2013).  The next challenge will be to characterize, in vivo, individual miRNAs and specific families of miRNAs that are predicted to contribute to the proper CNS function to find new candidate key-disease driving genes in human.

Genes and Development 2013

Gennarino et al., Genome Research 2012

Sphingolipid Homeostasis During Development

In collaboration with Giovanni D'Angelo (EPFL Lausanne, Switzerland) and Thorsten Hornemann (University of Zürich) we recently learned that mutations in CERT1 (ceramide transporter) cause a neurodevelopmental syndrome by distrupting sphingolipid homeostasis. We gathered over 30 patients who showed varying degrees of motor, language, and cognitive delays as well as seizures and mild dysmorphia; the severity of the effects correlate with the degree to which sphingolipid levels rise. Studying the disease-causing mutations in cells and flies led to the discovery of a new coiled-coil domain of CERT that enables it to form dimers (Journal of Clinical Investigation (JCI), 2023).

Gehin et al., JCI 2023

bottom of page