Review Papers Claim by Claim

Visual first — Key experimental summaries (data-derived)

Visuals explained (short)

• The panel breakdown (plot 1) uses the cell-line categorization reported by the authors: 8 lines showed augmented PI3K inhibitor activity when co-treated with OSI-906, 9 lines showed little or no additional effect, and 1 line had decreased sensitivity with IGF1R inhibitor co-treatment — n = 18 total Panel response breakdown (Fig.3)
• The GI50 dose–response example (plot 2) reproduces the reported HCC1806 shift (GI50 ≈ 4.2 µM → ≈1.4 µM with 2 µM OSI-906) as a conceptual demonstration of how co-inhibition left-shifts the viability curve in sensitive lines HCC1806 GI50 shift (Fig.2A)

Evidence synthesis — what the experiments show (organized)

1) Discovery: genetic screen → IGF1R

A pooled TRC kinome shRNA screen in HCC1806 (moderately PI3K-sensitive, PIK3CA-amplified) identified IGF1R as the top sensitizer (9 independent hairpins depleted under 1 µM GDC-0941). The screen pipeline (MOI≈0.2, 8 population doublings, barcode sequencing, DESeq hit calling) is standard and appropriate for finding context-specific synthetic vulnerabilities RNAi screen methodology and IGF1R hit

2) Validation: genetic and pharmacologic

Stable IGF1R knockdown increased sensitivity to GDC-0941 in HCC1806 (by doubling-time and colony assays). Pharmacological IGF1R inhibitors (OSI-906, NVP-AEW541) reproduced the effect and in HCC1806 reduced GI50 ≈3-fold when combined with GDC-0941. These orthogonal validations (shRNA + two distinct inhibitors) substantiate on-target biology rather than off-target compound effects IGF1R genetic/pharmacologic validation

3) Mechanism: residual PI3K signaling & apoptosis

Sensitive lines (HCC1806, CAL-51) show that GDC-0941 reduces p-AKT rapidly but p-AKT rebounds at 24 h; addition of OSI-906 further reduces residual p-AKT and p-PRAS40 and doubles caspase-3/7 apoptosis compared with PI3K inhibitor alone. Combination-insensitive lines (CAL-148, MFM-223) do not show this additional biochemical suppression. Thus, IGF1R provides upstream input that becomes rate-limiting when PI3K is inhibited; co-inhibition prevents reactivation and increases apoptotic response in responsive lines Biochemical mechanism: p-AKT/p-PRAS40 and apoptosis

4) Context dependence: ligand and RNA-binding protein modulation

IGF2 ligand availability changed outcomes: adding IGF2 decreased sensitivity to GDC-0941 in sensitive lines and enhanced colony outgrowth under PI3K blockade; neutralizing anti-IGF2 antibodies increased sensitivity. IGF2BP3 (IMP3) mRNA/protein correlated best with ΔAUC sensitization across the panel and is enriched in TNBC in TCGA; authors propose IGF2BP3 as a candidate predictive biomarker because it can increase IGF2 translation in other systems — but causality remains to be demonstrated in breast cancer IGF2 ligand and IGF2BP3 association

Critical appraisal — strengths, limitations, and blindspots

1. Strengths: orthogonal validation (genetic + multiple inhibitors); use of an 18-line TNBC panel (captures heterogeneity); biochemical time-course showing rebound signaling; plausible mechanistic model linking ligand-driven RTK input to residual PI3K activity; candidate biomarker identification coupled to TCGA analysis Paper strengths summary
2. Key limitations / blindspots:
   • All data are in vitro; no in vivo (xenograft or PDX) validation is presented in the manuscript text provided — pharmacokinetics, insulin/IGF endocrine feedback, stromal/immune interactions and drug exposure differ in vivo and could alter efficacy or toxicity (IGF1R/IR cross-reactivity of OSI-906 is clinically relevant) In vitro limitation noted by authors
   • Quantitative biomarker claim (IGF2BP3 predicts response) is correlative across 18 lines — small sample for predictive biomarker discovery and vulnerable to selection bias; prospective validation in clinical samples or independent panels is required before clinical use IGF2BP3 biomarker caveat
   • Synergy quantification limited by experimental constraints: Loewe model could be applied only where OSI-906 alone produced a full curve (rare), so authors used ΔAUC; ΔAUC is useful but less standardized across panels and may be sensitive to assay conditions and single-point OSI-906 effects (they used 4 concentrations) Synergy metric limitations
   • Overexpression of IGF1R did not induce resistance — authors interpret this as ligand-limitation rather than receptor abundance; mechanistically plausible but leaves open alternative resistance routes (e.g., p110β compensation, RTK coactivation, glycosylation/state changes) that were not exhaustively explored in this work IGF1R OE negative result and interpretation
3. Potential biases and confounders to watch: selection of cell lines (18 chosen lines may under/overrepresent certain TNBC molecular subtypes), in vitro ligand concentrations (IGF2 at 50 ng/mL may be supraphysiologic), shRNA off-target effects (although nine hairpins argue against single-hairpin artifacts), and lack of patient-derived validation (tumor microenvironment, insulin feedback) Methodological caveats and in vitro dosing

Practical implications & translational path

• Rationale supports testing PI3K + IGF1R inhibitors in TNBC with stratification (IGF2BP3-high / IGF2-high) but only after pharmacodynamic and toxicity assessments in vivo and early-phase clinical trials that monitor systemic insulin/IGF levels and metabolic adverse events (OSI-906 has IR activity) Translational caution re OSI-906/IR cross-reactivity
• IGF2BP3 is an attractive candidate predictive biomarker but requires: (a) orthogonal validation in patient-derived xenografts or organoids, (b) retrospective analysis of patient cohorts treated with PI3K or IGF1R agents if available, and (c) prospective biomarker-driven early-phase trials IGF2BP3 biomarker development path

Concrete next experiments (testable, prioritized)

1. In vivo xenograft/PDX experiments using combination GDC-0941 (or clinical PI3K inhibitor analog) + OSI-906 (or selective IGF1R inhibitor) in IGF2BP3-high and -low PDXs, measure tumor growth, p-AKT dynamics, systemic insulin/IGF levels, and tolerability (glucose/insulin) — to validate efficacy and safety in a physiologic context.
2. Mechanistic rescue: in sensitive cell lines, knock down IGF2BP3 or neutralize IGF2 and test whether combination sensitivity is lost; conversely, overexpress IGF2BP3 in insensitive lines and assess whether IGF2/IGF1R dependence increases — to move from correlation to causation.
3. Single-cell RNA-seq or spatial transcriptomics on combination-treated vs control PDXs to identify microenvironmental modulation (e.g., stromal IGF2 production) that may govern response.

Quick reproducibility checklist (from methods)

• Use HCC1806 as initial genetic-screen validation model; TRC kinome library or focused kinome libraries; MOI ~0.2; select with puromycin; treat ±1 µM GDC-0941 for ~8 doublings; recover shRNA barcodes and sequence (HiSeq); analyze with DESeq.
• Viability assays: 72-h CellTiter-Blue; four replicate wells; GI50 by 4-parameter curves where possible; ΔAUC for sensitization where OSI-906 single-agent curve absent.
• Biochemistry: time-course p-AKT/p-PRAS40 Westerns (1 h and 24 h) to detect rebound; caspase-3/7 IncuCyte reporter for apoptosis kinetics.
• Gene expression: RNA-Seq aligned to hg19; extract read counts for candidate gene panels; correct for multiple testing (Benjamini-Hochberg) when correlating with ΔAUC.

Primary source for all claims and data visualized above: the original paper De Lint et al. 2016 (MCT)