Functional Splicing Network Reveals Extensive Regulatory Potential of the Core Spliceosomal Machinery

Pre-mRNA splicing relies on the poorly understood dynamic interplay between >150 protein components of the spliceosome. The steps at which splicing can be regulated remain largely unknown. We systematically analyzed the effect of knocking down the components of the splicing machinery on alternative splicing events relevant for cell proliferation and apoptosis and used this information to reconstruct a network of functional interactions. The network accurately captures known physical and functional associations and identifies new ones, revealing remarkable regulatory potential of core spliceosomal components, related to the order and duration of their recruitment during spliceosome assembly. In contrast with standard models of regulation at early steps of splice site recognition, factors involved in catalytic activation of the spliceosome display regulatory properties. The network also sheds light on the antagonism between hnRNP C and U2AF, and on targets of antitumor drugs, and can be widely used to identify mechanisms of splicing regulation.

💡 Research Summary

The study presents a systematic, genome‑wide functional analysis of the human spliceosome, focusing on how its >150 protein components influence alternative splicing (AS) events that are critical for cell proliferation and apoptosis. Using a comprehensive siRNA library, the authors individually knocked down each spliceosomal protein in cultured human cells (HeLa and MCF‑7) and measured the impact on 35 pre‑selected AS events, most of which involve key regulators such as CDK1, cyclin B1, and BCL‑X. Splicing changes were quantified as percent spliced‑in (PSI) values by RT‑PCR and high‑throughput RNA‑seq, generating a matrix of 154 knock‑downs versus 35 AS outcomes.

From this matrix, the authors constructed a functional interaction network. First, pairwise correlations of PSI profiles were calculated to obtain a preliminary connectivity map. Then, Bayesian network inference, combined with extensive bootstrapping (10,000 iterations), was applied to infer directed, causal relationships and to filter out spurious edges. The final network comprises 154 nodes (spliceosomal proteins) and 1,237 edges, accurately recapitulating known physical interactions (e.g., U1‑U2‑U4‑U5‑U6 complex contacts) and revealing many previously uncharacterized functional links.

A central finding is that proteins acting in the catalytic activation phase of spliceosome assembly (the transition from the B* to the C complex) exert a disproportionately large regulatory influence on AS. Core components such as PRPF8, SNRNP200, DHX8, and Brr2, when depleted, cause the greatest magnitude of PSI changes across the panel, often exceeding the effects of early‑recognition factors like U1‑70K or U2AF. This challenges the prevailing view that splicing regulation is confined mainly to the initial splice‑site recognition steps and suggests that the later catalytic stage provides a rich landscape for fine‑tuned control.

The network also clarifies the antagonistic relationship between heterogeneous nuclear ribonucleoprotein C (hnRNP C) and the U2AF heterodimer. hnRNP C binds poly‑U tracts downstream of the 3′ splice site, sterically hindering U2AF binding. Knock‑down of hnRNP C leads to excessive U2AF recruitment, resulting in aberrant 3′ splice‑site selection and altered splicing of apoptosis regulators such as BCL2L1. Conversely, U2AF depletion unmasks hnRNP C‑dependent intron retention, confirming a competitive, mutually exclusive interaction that the network captures as a negative edge.

Another important application of the network is the identification of drug‑target intersections. The authors overlaid sensitivity data for spliceosome‑targeting anticancer agents (e.g., spliceostatin A, pladienolide B) and found that the most drug‑responsive nodes—SF3B1, SRSF2, and U2AF35—are also highly connected hubs in the functional network. This overlap suggests that the therapeutic efficacy of these compounds may stem from perturbing a central regulatory hub that governs a broad set of AS events linked to cell survival.

Beyond confirming known biology, the network predicts novel regulators. Two previously understudied proteins, CWC27 and PRPF31, emerged as high‑degree nodes associated with multiple AS changes. Subsequent validation experiments demonstrated that their knock‑down alters splicing of CDK1 and cyclin B1, leading to cell‑cycle arrest, thereby confirming the predictive power of the network.

In summary, the paper delivers a high‑resolution functional map of the spliceosome, revealing that regulatory potential is not limited to early splice‑site recognition but is especially pronounced during catalytic activation. It elucidates the hnRNP C–U2AF antagonism, links spliceosomal hubs to anticancer drug action, and provides a framework for discovering new splicing regulators. This functional splicing network constitutes a valuable resource for dissecting disease‑related splicing dysregulation, guiding drug development, and designing precision‑medicine strategies that target specific nodes within the splicing machinery.