Two genomes, one proteome

In human cells, genes contain numerous introns that need to be excised to form a functional mRNA. We developed direct RNA nanopore sequencing methods, such as nanopore analysis of co-transcriptional processing (nano-COP), to study pre-mRNA processing without the length-associated biases found in techniques that rely on cDNA synthesis and PCR.

We found that 60% of introns are removed co-transcriptionally, while the pre-mRNA is still attached to Pol II. Splicing catalysis, however, largely occurs after Pol II has transcribed thousands of further nucleotides. Examining the sequence of intron removal across pre-mRNAs, we found that introns tend to be excised in one or occasionally two preferred orders that were constant across cell types and alternative splicing, and identified roles for specific cis-elements in controlling splicing order.

Textbook views of the mRNA life cycle typically indicate that all mRNAs are exported. However, the nuclear RNA exosome degrades introns, antisense RNAs and other non-coding RNAs. We recently found that nuclear mRNA degradation may play a broader role in gene expression regulation than previously appreciated.

Using subcellular TimeLapse-seq, we tracked the age of RNAs as they moved across subcellular compartments and determined RNA half-lives at subcellular resolution using a Bayesian modeling framework. This measured the rates at which transcripts were released from chromatin, exported from the nucleus, and loaded onto polysomes for all human mRNAs, as well as nuclear and cytoplasmic degradation rates.

A simple RNA life cycle model did not fit our data for ~10% of genes; instead, a model including substantial nuclear mRNA degradation was required. Hundreds of genes were predicted to undergo nuclear degradation (PUNDs), and more than 85% of the transcripts encoded by these genes are never exported from the nucleus. Many PUND genes participate in gene expression and RNA biology, including ribosome subunits, RNA splicing and nuclear mRNA export factors. PUND transcripts possess features such as incomplete splicing and long poly(A) tails, which may aid their identification by nuclear degradation pathways.

Metabolic labeling separates synthesis from decay, but the chemical recoding step it usually requires damages RNA and does not survive droplet-based single-cell protocols. Analog intrinsic recoding sequencing (AIR-seq) removes that step. Cells are labeled with N4-hydroxycytidine, the active compound of the antiviral molnupiravir, which is incorporated into RNA by RNA polymerase II and base-pairs ambiguously, so new transcripts carry C-to-T and T-to-C mismatches in otherwise standard libraries.

Fraction-new estimates from AIR-seq agree with subcellular TimeLapse-seq, hold across library types and sequencing depths, and work in single cells with the unmodified 10x workflow. A single labeled snapshot orders cells around the cell cycle and gives new RNA a roughly one-hour lead over total RNA, which sharpens RNA velocity estimates. Modeling those rates revealed offset synthesis and degradation programs for nuclear- and mitochondrial-encoded OXPHOS transcripts that leave steady-state abundance unchanged.

Discovered by genetic screens, yeast translation activators are nuclear-encoded and imported into mitochondria to promote the translation of specific mRNAs via their 5′ UTRs. To identify human genes that control mitochondrial gene expression, we developed genome-wide CRISPR screen strategies to identify factors that affect the balance of Complex IV subunits, mitochondrial-encoded COX1 and nuclear-encoded COX4. We identified known regulators such as TACO1, whose mutations cause Leigh syndrome, alongside genes with unclear mitochondrial functions. Among the latter, we are initially focusing on PREPL, a protein highly expressed in the brain, and NME6, a member of the nucleotide diphosphate kinase (NDPK) family.

Neurons are highly dependent on mitochondrial function, and their complex, polarized morphology offers a challenge for balanced nuclear and mitochondrial gene expression. We aim to determine whether this structure makes neurons more susceptible to mitonuclear imbalance due to genetic or physiological stress, and whether this is relevant to mitochondrial malfunction in neurodegenerative disease.