Review Papers Claim by Claim

Visual first — Key quantitative summaries

Visual notes

The first graphic reproduces the paper's qualitative point: when transcription is measured with a rapid, perturbation-minimizing protocol (here using CCND1 induction rather than prolonged FACS sorting), mRNA fold changes align strongly with proteomic fold changes. The second figure captures the live-MS2 result: mean ON (productive) time increases markedly with nuclear/cell size, while OFF-time shortens modestly — burst amplitude (loading rate) changes little.

Evidence summary (visual → concise claims with direct citation)

1. Experimental handle for size: inducible CCND1 knockdown / rescue in hTERT-RPE1 to generate cycling populations spanning ~2× volume with steady-state reached by day 4 — this is the core perturbation enabling fast sampling with minimized sorting artifacts (You et al. Methods: CCND1 system).
2. Proteome remodeling measured by TMT-MS: ~3,800 proteins with per-protein slope values recapitulate prior size-sorted datasets; lysosomal proteins superscale and remain among the strongest size-dependent changes (You et al. proteomics results).
3. Turnover (proteins): pulse-SILAC–TMT across 10-day chase yields T50 half-lives for ~2,000 proteins; half-lives are largely stable across ~2× size, with only short-lived proteins (You et al. protein turnover).
4. Turnover (mRNA): SLAM-seq (s4U pulse 24 h + chase) for ~2.5k genes finds median mRNA half-life ~6.7 h and no strong size-dependent differential stabilization across most genes — i.e., mRNA half-life changes are small and do not explain slope variation (You et al. SLAM-seq).
5. Transcription explains most proteome change: RNA-seq slopes correlate strongly with protein slopes when using rapid CCND1 induction sampling; combined OLS model using mRNA slopes + short-lived protein half-lives predicts protein slope with R^2 ≈ 0.75, with most explanatory power from transcription (You et al. transcript–protein modeling).
6. Live bursting dynamics (mechanism): MS2 live imaging at endogenous loci (HBEC-kt3 cells with intronic MS2) for two scaling genes (RAB7A, RHOA) — analysis of thousands of bursts shows increased ON-time (burst duration) with nuclear/cell size, decreased OFF-time, and stable amplitude; thus scaling arises mainly from longer productive bursts rather than amplitude increases (You et al. MS2 bursting data).

Critical appraisal — strengths

• Comprehensive multi-omic design linking manipulation (CCND1) → transcriptome → mRNA half-life → proteome → protein turnover → live transcriptional bursting; results internally consistent and cross-validated with previous size-sorted datasets (You et al. cross-validation).
• Use of rapid sampling (induction steady-state at day 4) minimizes transcriptomic artifacts introduced by long FACS sorting — plausibly explains higher mRNA–protein correlation relative to prior human sorting studies.
• High replicate numbers in the live imaging bursting analysis (thousands of bursts, many cells) gives robust statistics for kinetic parameters.
• Open code availability (GitHub) improves reproducibility and transparency.

Critical appraisal — limitations & blind spots

• CCND1 manipulation is a strong, non-physiological perturbation: while it yields graded steady-state sizes, CCND1 knockdown/overexpression could change signaling (CDK activity, cell-cycle timing, transcription factor phosphorylation) beyond pure size effects. The authors partially control for cell-cycle phase (G1 vs G2 sorting) but non-size CCND1-dependent effects remain possible (You et al. CCND1 caveats).
• Bursting live-imaging is limited to two genes (RAB7A, RHOA) in HBEC-kt3 cells — both are scaling genes but may not represent the full spectrum of gene classes (highly expressed, lowly expressed, tissue-specific regulators). Genome-wide single-molecule live analyses are currently infeasible, so generality beyond these examples is inferred but not proven (You et al. bursting scope).
• Translation layer was not directly measured (e.g., Ribo-seq or puromycylation): while mRNA explains most protein variation, changes in translation initiation/elongation efficiencies could contribute to gene-specific deviations; the authors recognize translation was not examined and is a potential remaining contributor.
• SLAM-seq and pulse-SILAC assume steady-state kinetics for fits; perturbations in growth rate or induction-dependent effects could bias half-life fits, especially for long-lived proteins/ transcripts. The authors used QC filters (R^2 thresholds) but residual biases may persist (You et al. turnover methods).
• Generality across cell types and in vivo tissue remains to be established: RPE-1 and HBEC-kt3 are diploid epithelial-type lines; other cell types with different chromatin, transcription factor repertoires, or in tissues with mechanical cues may scale differently. The field review highlights diverse mechanisms across organisms and tissues (Skotheim & colleagues review).

Where results could mislead — Caveats for interpretation

1. Correlation ≠ causation at genome scale: although mRNA slopes predict protein slopes (R^2≈0.75), unmeasured post-transcriptional regulation including translation and localized mRNA trafficking could create exceptions. Demonstrating causal sufficiency requires perturbing transcriptional burst regulators and measuring consequent proteome changes.
2. Selection / adaptation in long inductions: the CCND1-induced large cells show gradual size decline over days (authors note possible selection), which might bias steady-state proteomes toward subpopulations better adapted to CCND1 perturbation.
3. MS2 tagging and live-imaging may perturb native kinetics: MS2 stem-loop insertions into introns are minimally invasive in their source study but can alter splicing or elongation in some genes; the authors used previously validated lines but residual tag effects remain possible (You et al. MS2 methods).

What would falsify the paper's main conclusion?

• If an independent system that increases cell size without altering CCND1 (e.g., nutrient-driven growth increase or ploidy-controlled size increase) produced identical proteome remodeling but with mRNA slopes not predictive of protein slopes (i.e., turnover or translation explained most variation), that would contradict the central claim.
• If genome-wide live-burst measurements (future technology) showed burst amplitude increases, not ON-time changes, as the dominant scaling mechanism across most genes, the mechanistic claim (ON-time scaling) would need revision.

Practical recommendations / follow-up experiments

1. Measure translation directly (ribosome profiling or puromycin-labeling) across size bins to quantify translational contributions to slope outliers.
2. Use orthogonal size-increase models (mTOR activation, nutrient upshift, polyploidization) to test whether transcriptional scaling and ON-time increases generalize beyond CCND1 perturbation.
3. Perturb candidate burst regulators (e.g., p-TEFb, pause-release factors, MYC family) with timed, graded inhibition/activation and measure effects on bursting kinetics and proteome slopes to test causality.
4. Genome-scale single-molecule transcription assays (emerging technologies) to test whether ON-time scaling is the dominant mode across gene classes, or whether promoter architecture predicts scaling behavior.

Concise verdict (evidence-weighted)

Conclusion: The paper provides strong, multi-modal evidence that transcriptional modulation — embodied as increased transcriptional ON-time (burst duration) with cell size — is the principal driver of proteome remodeling during a near-twofold increase in proliferating human cells; protein and mRNA turnover play only subsidiary roles. Confidence is high for the studied system and gene subsets, but broader generality across cell types and the full contribution of translation require additional experimental tests (You et al. main claims).

Actions & tools

You can run a fully independent bioinformatics agent to (A) re-fit the authors' OLS model, (B) re-analyze proteomics and RNA-seq slope distributions, or (C) cross-compare with public size-sorted datasets; click below to start an iterative AI scientist analysis.

Author review links

Note: All specific claims above are anchored to the You et al. 2025 preprint; broader context relies on the 2022 Annual Review on cell size control. To dig deeper (reproduce OLS, replot slope distributions, or test alternate size perturbations), click "Run AI Scientist Analysis" to start a reproducible pipeline that will fetch raw data from the authors' GitHub and re-analyze it.