henhaobio
Sign in

bioRxivSystems BiologyPreprint

The diverse functions of viral microproteins

Yi Hua CheniD, Louisa GrauvogeliD, Adam S Zhang, André C. MichaelisiD, Tim HeymanniD, Elisabeth Zollbrecht, Paulina Engel, Christian Pensl, Kathryn E. YostiD, Wesley SueniD, Sumita SinhaiD, Edwin N. NeumanniD, Stephen Joseph ElledgeiD, Karl‐Peter HopfneriD, Benjamin E. GewurziD, Matthias ManniD, Jonathan S. WeissmaniD

PDFDOICodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

A large fraction of the uncharacterized proteome consists of microproteins. Viruses, rich in microproteins with potent activities, provide a unique system for exploring their functions and evolution. Here, we systematically annotate viral microproteins through a phenotype-first approach: construction of a >50,000-element small ORF library spanning the microprotein-coding potential of human-infecting viruses, followed by broad screens that enrich candidates for high-resolution profiling by Perturb-seq, mass spectrometry, AI-guided structure prediction, and mechanistic studies. This atlas identified >2,000 active microproteins that modulate host pathways through diverse mechanisms, including molecular mimicry, hijacking of key regulators, and altering subcellular localization. We further show the oncogenic potential of BNLF2b, an uncharacterized Epstein-Barr virus microprotein linked to nasopharyngeal carcinoma. Three principles emerge: viral microproteins are diverse and often multifunctional, show widespread convergence toward shared functions, and serve as reservoirs of evolutionary innovation. These findings elucidate how viral microproteins occupy a large functional landscape, establish them as versatile tools for host manipulation, and provide a framework for functional annotation of uncharacterized proteins, contributing toward predictive systems virology.

In-depth summary

Claude Opus 5.5 · xhigh effort · 7 Nov 2026 · from the full text of version 1 (24 Aug 2026) · CC BY 4.0

A >50,000-element viral smORF library screened by growth, MHC-I, ER stress, Perturb-seq and mass spectrometry yields >2,000 active microproteins and shows EBV BNLF2b inhibits PKA by mimicry.

The paper describes a viral dark proteome (translated but unannotated, or annotated but uncharacterized sequences) that is largely unexplored beyond HIV and SARS-CoV-2. It builds a phenotype-first functional atlas of microproteins across human-infecting viruses, rather than relying on sequence or structure homology.

Key results

  1. ShownA funnel from 52,979 smORFs (502 human-infecting virus genomes) to a 4,107-element enriched library yielded 2,028 growth hits and, across all assays, >2,000 active microproteins.

    Primary growth screens in K562 and RPE-1 gave 4,329 hits (8.2%) at 2 SD from negative controls. The 4-replicate enriched screen gave 2,028 hits (FDR < 0.05). Toxic hits correlated between cell lines (Pearson R = 0.78), and 70% of hits were unannotated. · A funnel strategy for the identification of…; Design and curation of the pan-viral smORF…; Growth screens identify toxic and pro-growth…; Discussion

  2. ShownTargeted MHC-I screens found hundreds of modulators, many independent of growth. AE-MS on five hits showed that microproteins engage several distinct steps of antigen presentation.

    748 of 8,806 smORFs altered surface MHC-I (FDR < 0.05); 40% of MHC-I hits did not score in growth screens. Known TAP inhibitors (ICP47, BNLF2a, CPXV012) were recovered. The five examined hits engage at least four pathway steps. · Additional microprotein functions captured by…; Identification of uncharacterized viral MHC…; AE-MS reveals that MHC-I inhibitors interact…; Microproteins act through short “molecular keys”

  3. ShownPerturb-seq on the enriched library found strong transcriptional phenotypes for many smORFs. Guilt-by-association clustering grouped homologous and non-homologous elements with shared functions.

    3,412 smORFs profiled; 1,185 passed the energy test and 691 had >10 DEGs. Negative controls had <5 DEGs; all 18 positive controls passed. Clustering gave 24 clusters (433 smORFs) and 752 singletons. · Perturb-seq identifies hundreds of smORFs that…; Perturb-seq classifies smORFs into functional…; ORF perturbation calling; Clustering, gene program analysis, and manual…

  4. ShownPhylogenetically distant microproteins (HIV-1 Vpr, SARS-CoV-2 ORF6, poxvirus MPXVgp154 and YLDV132R, an unannotated astrovirus smORF) co-cluster and block interferon signaling via nuclear transport. Three share a Rae1-binding motif.

    All cluster elements reduced IFN-α/γ reporter response in arrayed assays. Met-to-Ala mutants (ORF6 M58A, MPXVgp154 M69A, YLDV132R M75A) lost IFN inhibition. The astrovirus smORF has no BLAST hits and sits in the +1 frame of ORF2. · Perturb-seq cluster 4 (Nuclear shuttling…; Identification of a shared Cluster 4 molecular…; De novo emergence of an unannotated smORF from…; Discussion

  5. ShownEBV BNLF2b promotes epithelial growth by mimicking a human PKA inhibitor motif, activating MAPK. It also drives PKA-independent anchorage-independent growth that the PKA inhibitor PKIA does not.

    AE-MS: PRKACA/PRKACB were top interactors; AF3 model ipTM = 0.79. RRNAL and proline-rich motif mutants lost the fitness advantage. RAS or MEK inhibitors abolished BNLF2b and PKIA advantage but not E7. BNLF2b, not PKIA, grew in soft agar. · EBV BNLF2b is a singleton that promotes…; EBV BNLF2b inhibits PKA to drive proliferation…; Identification of an EBV BNLF2b molecular key…; EBV BNLF2b activates an additional oncogenic…

Methods, data and reproducibility

Synthesized a pooled lentiviral library of viral smORFs under a doxycycline-inducible vector (pANDORA) and screened it for growth, MHC-I surface levels and ER-stress survival, then ran Perturb-seq (10x Flex) on an enriched subset. Hits were followed up with affinity-enrichment mass spectrometry, AlphaFold 3 modelling, mutagenesis, RNA-seq and arrayed assays.

  • Pan-viral smORF library (52,979 elements) from complete genomes of human-infecting viruses in VirusHostDB; enriched library of 4,107 elements · virus · n = 52,979 smORFs from 502 genomes; 4,107 in enriched library
  • Pooled growth, MHC-I and ER-stress screens in RPE-1, K562 and A375 cells · human (cell lines) · n = 3 replicates (50K library), 4 replicates (enriched library) · PRJNA1516588
  • Perturb-seq of enriched library in RPE-1 cells · human (RPE-1 cell line) · n = 1,231,800 cells; 780,731 singlets; 3,412 smORFs · GSE344535
  • Arrayed high-throughput AE-MS of ALFA-tagged viral baits in HEK293 cells, plus RPE-1 AE-MS · human (cell lines) · n = 380 viral baits; 2,304 runs (smORF library), 40 runs (RPE-1) · PXD08253
  • Bulk RNA-seq of RPE-1 cells expressing PKIA, BNLF2b wild type or mutants, E7 or controls · human (RPE-1 cell line) · n = 4 biological replicates per condition · GSE342977
Methods & tools
pANDORA, 10x Genomics Flex, Cell Ranger 9.0.1, STARsolo, scanpy 1.11.5, pertpy 1.0.3, HDBSCAN, pymde 0.1.18, AlphaFold 3, DIA-NN, directLFQ, MrBayes 3.2.6
Reagents & environment
Versions stated for bowtie2 2.4.2, SAMtools 1.11, Cell Ranger 9.0.1, STARsolo 2.7.1a, scanpy 1.11.5, pertpy 1.0.3, pymde 0.1.18, DIA-NN (1.9.1 and 2.1.0), directLFQ 0.3.3, MUSCLE 5.1, MrBayes 3.2.6, Cytoscape 3.10.4.
Limitations (6)
  • Microproteins that need other viral co-factors are missed; EBV BDLF3.5, which acts within the viral pre-initiation complex, did not score in any screen. · stated by the authors
  • Many functions are context-specific, so only a subset of physiology is captured. Pro-growth phenotypes depend on cell type and media, and transformed lines can mask oncogenic effects. · stated by the authors
  • The initial growth-phenotype filter of the funnel misses smORFs with growth-independent functions, as the MHC-I screens illustrate. · stated by the authors
  • Transgenic expression may fail to accumulate very short or unstable microproteins (false negatives) or produce overexpression artifacts. · stated by the authors
  • Mechanistic models (Rae1-binding motif, PKA mimicry, beta-hairpin STAT1 binding) rest largely on AlphaFold 3 predictions, restricted to high-confidence models, and the screens use ectopic expression in cell lines rather than infection. · noted in this summary
  • Perturb-seq was done in one cell line (RPE-1), and the conservative clustering leaves 752 of 1,185 phenotype-positive smORFs as unassigned singletons. · noted in this summary

Read this if you want a pan-viral functional atlas of small ORFs, or a template for combining pooled overexpression screens, Perturb-seq and AE-MS to annotate uncharacterized proteins.

Citations
1

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISWang '21Nomburg '24Santos '23Enard '15Ng '22Profeta '26
58% semantic matchPlant Biology
L. Wang, H. Tan, L. Medina-Puche+14 more·bioRxiv·Apr 7, 2021