Research report: Differential phosphorylation of PHIP phosphopeptides with implications in insulin signaling

A protein called PHIP, long known as a binding partner of the insulin receptor substrate 1 (IRS1), has now been placed under a phosphoproteomic microscope. In a study published in BMC Molecular and Cell Biology, researchers report the results of an unusually large data-integration effort: they screened more than 3,825 articles on PubMed, extracted 185 studies that reported phosphorylation events on PHIP, and catalogued 1,008 individual profiling datasets from them. Of these, 412 datasets, drawn from 89 different studies, showed differential regulation of PHIP phosphopeptides, meaning the abundance of phosphorylated PHIP peptides changed between experimental and control conditions. From this mountain of mass spectrometry data, the team distilled a ranked map of where and how often this protein carries phosphate groups in human cells and tissues.

PHIP, short for Pleckstrin Homology domain Interacting Protein, is an 1,822-amino-acid scaffold containing two bromodomains, epigenetic reader modules that bind acetylated lysine residues, plus multiple WD40 repeats that mediate protein-protein interactions. It was originally identified through its association with the pleckstrin homology domain of IRS1, and the PHIP gene has since been linked to Chung-Jansen syndrome, an autosomal dominant disorder marked by developmental delay and intellectual disability. Because increased acetylation of IRS1 is associated with enhanced insulin signaling, the authors note a hypothesis that PHIP binding to IRS1 may facilitate its phosphorylation by the insulin receptor, positioning PHIP at a junction between chromatin reading machinery and metabolic signaling.

The core technical achievement of the new work lies in how the datasets were harmonized. Because different studies used different mass spectrometry platforms, quantification strategies such as label-free, TMT or SILAC, and different search pipelines, the researchers deliberately avoided direct intensity normalization, which they judged infeasible across so many heterogeneous sources. Instead, they curated data at the level of individual phosphorylation sites, resolving the multiplicity that arises when one peptide carries several phosphate marks, and required high-confidence site localization, accepting only sites with a localization probability of at least 75 percent or an A-score of at least 13. Each time point or experimental condition within a study was treated as an independent dataset with a unique code, so that recurring signals across independent studies, rather than quirks of any single platform, would drive the conclusions.

That ranking produced 45 class 1 phosphorylation sites on PHIP, of which seven stood out as predominant: S1283, S1281, S1783, S1315, S879, S880 and S881, detected with frequencies of 156, 143, 136, 114, 113, 105 and 100 across differential datasets respectively. Structurally, the team observed that these predominant sites sit outside the two bromodomains, a pattern they interpret as pointing toward possible allosteric regulation, in which phosphorylation at a distance could alter the behavior of the acetyl-reading domains. The authors also caution that detection frequency alone can be skewed by peptide-specific mass spectrometry detectability, so their definition of predominant sites rested on the consistency of differential regulation across datasets rather than raw counts.

With the site map in hand, the researchers asked which other proteins rise and fall in phosphorylation alongside PHIP. For every pair consisting of a predominant PHIP site and a phosphorylation site on another protein, they built contingency tables across the 412 differential datasets and applied a one-sided Fisher’s exact test, correcting for the enormous number of comparisons with a Benjamini-Hochberg false discovery rate procedure; associations were considered significant at a q-value below 0.05, and effect sizes were quantified as odds ratios with 95 percent confidence intervals. Positive co-differential regulation meant both sites moved in the same direction, negative meant opposite directions. Additional filters demanded that each pair appear in at least two distinct studies and at least two types of experimental conditions, guarding against overrepresentation of a single stimulus or a small cluster of studies.

The analysis surfaced a constellation of co-regulated proteins with well-established roles in insulin signaling, including NCOA5, ERBB2, SMARCC1, EGFR, C2CD5, MET, APC and SOS1. Most of these showed positive co-differential regulation with PHIP phosphorylation, while SOS1 at site S1136 was negatively co-regulated. Gene set enrichment analysis of the co-regulated proteins pointed to the insulin receptor signaling pathway, the insulin-like growth factor receptor signaling pathway, DNA damage response, chromatin remodeling, cell cycle regulation and DNA repair. Most strikingly, phosphorylation sites on IRS1 and, especially, IRS2, described in the study as the main effector of insulin signaling in the liver, showed significant co-differential regulation with all seven predominant PHIP sites.

The co-regulation analysis also recovered plausible upstream kinases. CLK1 and CHEK2 were identified as kinases acting upstream of PHIP S1315, with CHEK2 phosphorylated at S379 and CLK1 at T138 showing significant conserved co-differential regulation with that PHIP site, while ATM was linked to PHIP S1783. Each of these kinases carries metabolic credentials in the literature: CLK1 inhibition has been reported to improve insulin sensitivity by blocking phosphorylation of the transcriptional cofactor THRAP3 at S243; CHEK2 participates in the regulation of insulin secretion and glucose homeostasis through cell cycle control in pancreatic beta cells; and ATM phosphorylates 4E-BP1 at S111 to support insulin-stimulated translation initiation, a connection that helps explain the insulin resistance seen in ataxia-telangiectasia patients.

A second thread of the analysis connected PHIP phosphorylation to hepatocellular carcinoma. Several PHIP phosphopeptides had been differentially regulated in cell line studies of liver cancer, and among 14 PHIP phosphorylation sites identified in tissue phosphoproteomics datasets, S1281, S1283, S1783 and S1315 appeared predominantly in both hepatocellular carcinoma cell line and tissue data. Cross-checking the CPTAC proteomics data through the UALCAN portal, the team found elevated abundance of PHIP, SMARCA4 and CHEK2 in liver cancer samples, with n equal to 165, and cProSite database mapping showed significantly different phosphorylation levels of PHIP at S879, S880, S1281, S1283 and S1783, and of SMARCA4 at S1417, in tumor samples. SMARCA4, a validated binary interactor of PHIP whose phosphorylation sites positively co-regulated with six predominant PHIP sites, is known to promote tumor cell survival in hepatocellular carcinoma by suppressing accumulation of the lipid peroxidation product 4-hydroxynonenal, and high SMARCA4 levels have been linked to poor patient prognosis.

The authors are careful about what these patterns do and do not establish. The entire study is a large-scale computational meta-analysis; the functional roles it suggests rest on statistically supported co-differential regulation and enrichment, not on direct biochemical confirmation, and none of the reported connections has yet been experimentally validated for its significance in disease biology. The limitations are substantial: publicly available mass spectrometry data differ in sample types, platforms, pipelines and quality, limiting normalization and cross-study comparability; all co-regulation patterns are interpreted as associative rather than causal; and a predominant site may matter less biologically than a rare site critical under conditions poorly represented in the data. Detection frequency, the team reiterates, can also reflect sequence-dependent peptide detectability rather than biological relevance.

Even so, the work sketches a testable framework. Future validation, the authors write, should use phospho-mutant constructs, PHIP knockdown or overexpression models and in vitro kinase assays to confirm whether the predicted regulatory effects hold up, and to clarify PHIP’s mechanistic contribution, if any, to hepatic carcinogenesis. Given that PHIP has already been reported to cooperate with EGFR and ErbB signaling in hepatocellular carcinoma models, where genetic or pharmacological inhibition of PHIP combined with ErbB inhibitors reduces tumor cell proliferation and triggers apoptosis, the phosphorylation-centric map offered here gives wet-lab researchers a concrete, site-specific starting list. What remains an open question is whether the co-regulation captured across a thousand datasets reflects genuine signaling wiring or merely the statistical shadow of shared experimental conditions.

Subject of Research: A computational meta-analysis of PHIP phosphorylation sites and their co-regulation with insulin signaling proteins

Article Title: Differential phosphorylation of PHIP phosphopeptides with implications in insulin signaling

Article References: Sujina, A., Fahma, A., Subair, S., Gopalakrishnan, A. P., Ummar, S., Shivamurthy, P. B., Raju, R., & Ramesh, P. (2026). Differential phosphorylation of PHIP phosphopeptides with implications in insulin signaling. BMC Molecular and Cell Biology, 27(1), Article 34. https://doi.org/10.1186/s12860-026-00590-y

Image Credits: AI Generated

DOI: 10.1186/s12860-026-00590-y

Keywords: PHIP, phosphoproteomics, insulin signaling, IRS1, IRS2, hepatocellular carcinoma, phosphorylation, CHEK2, SMARCA4, ATM, mass spectrometry, data integration