Does the culture environment move reprogramming efficiency more than delivery or cell fraction?

Answer

The round is undecided. No paper in the corpus breaks the hypothesis under its own condition, and none confirms it cleanly. Lam et al. 2022 is the one paper that varies a culture-environment axis and a starting-fraction axis together. Agitated laminin-521 microcarriers gave 0.97% against 0.02% in a flat dish for PBMC (47.6 ± 7.5-fold), whereas CD34+ cells gave 0.02% against 0.02% in the dish and 0.5% on microcarriers 26454120. That is directionally consistent with the hypothesis, but it is not a clean causal test: the microcarrier and monolayer arms ran different protocols (E8/mTeSR1 on microcarriers, conventional growth medium in the dish), and the microcarrier arm also changed the surface to laminin-521. The environment contrast is therefore a package of vessel, surface, agitation and medium, and medium is outside the hypothesis's closed list. Goh et al. 2013 is the only paper in which delivery outweighs surface (mRNA 1.89% against episomal plasmid 0.012% on feeders, about 157-fold, versus a surface effect of about 8.6-fold), but the arms differ in cocktail, medium and scoring day, so it is a candidate counter-example, not a refutation d234287c.

Summary

The corpus holds two papers that cross an environment axis with a competing axis (Lam, Goh). Neither satisfies "everything else held constant": Lam is directionally supportive but confounded by medium and surface; Goh points against the hypothesis but is confounded by cocktail, medium and scoring day. Every other paper varies the surface alone, or varies a competing axis within one delivery class. Surface effects reported across these papers are small to moderate (no difference reported in Vuoristo; 0.014% against 0.011% in Trokovic; about 1.7-fold in Timilsina; about 8.6-fold in Goh), while the one format/agitation effect is large (47.6 ± 7.5-fold). The corpus does not settle whether a matched design would keep the environment effect ahead. The verdict is undecided, and this is a search result, not a claim that no refuting paper exists.

Evidence

• Lam et al. 2022 — CytoTune 2.0 Sendai on PBMC, TRA-1-60 at day 14 per input cell, n=3. Monolayer 0.02 ± 0.01% against agitated microcarriers 0.97 ± 0.1% (p<0.0001); CD34+ 0.02 ± 0.01% against 0.5 ± 0.1%. No delivery variation. Media differ between arms, so the contrast is directional, not isolated. 26454120
• Goh et al. 2013 — BJ fibroblasts. mRNA 1.89% on feeders and 0.22% on Matrigel; episomal plasmid 0.012 ± 0.008% on feeders and 0.103 ± 0.008% on Matrigel. The surface effect reverses sign between delivery classes. On Matrigel the delivery gap is about 2.1-fold. Arms are unmatched in cocktail, medium and scoring day; replicates for the mRNA arms are not stated. d234287c
• Goh et al. 2013, second reading of the same paper — reaches the same figures and the same conclusion: the pair that looks like a delivery effect larger than the surface effect differs in scoring day (18 versus 30) and denominator, so it is not a fair refutation. A Lam HFF-01 format effect of about 24-fold (0.97 against 0.04) is also recorded there. c803997f
• Trokovic et al. 2014 — non-activated PBMC, SeVdp: 0.014 ± 0.006% on feeders against 0.011 ± 0.006% on Matrigel (n=6), a surface effect inside the scatter. Cytotune 4V gave 0.005% and no colonies from non-activated PBMC, about 2.8-fold below SeVdp. Both vectors are Sendai, activation is a pre-treatment and not a purified fraction, and medium and MOI differ, so this is a candidate counter-example only. 2d7df7a7
• Schlaeger et al. 2015 — fibroblasts: delivery class alone spans mRNA 2.1% to episomal 0.013%, with SeV 0.077% and lentivirus 0.27% (P<0.05 for the pairwise differences reported). Surface is confounded with method, so the paper does not test the hypothesis. ed873f4c
• Chou et al. 2015 — the same episomal class with two vector sets: 10-50-fold more colonies for one (p<.01). Vitronectin against feeders is printed only as "higher", with no number. Wanted row. 590e5c24
• Timilsina et al. 2023 — surface only: 0.37% against 0.22% (p<0.0068, n=3), about 1.7-fold. No competing axis. 19d875b2
• Vuoristo et al. 2013 — retroviral reprogramming of fibroblasts, surface only: the text reports no difference between JAR matrix and Matrigel, with no numeric values in the text. 8bc601a0

Methodology

A paper qualified as a test only if it varied an environment axis (vessel format, surface, agitation, as the task defines them) together with delivery class or starting fraction. Each paper's fold-changes were compared only inside that paper; efficiencies were not pooled across papers, because cell type, scoring marker, day and denominator differ. The qualifying papers were then checked against "everything else held constant" using the held-constant fields of the verified records, and where a record states a confound (medium, cocktail, scoring day) it is carried into the answer. Lam is described as directional and confounded, not as matched. A previous version of this text quoted a Trokovic surface ratio that the reviewer could not trace; this version prints only the two percentages the record gives, 0.014% on feeders and 0.011% on Matrigel.

Limitations

• Only Lam and Goh cross an environment axis with a competing axis, and neither is a matched design. The medium difference in Lam is outside the hypothesis's closed list and cannot be separated from the environment effect.
• Replicates, spread and tests for the mRNA arms in Goh, and the surface effect in Chou, are not stated in the text that was read. Figures and supplements were not read.
• Simara 2014, Ye and Wang 2018 and Liu 2015 were reachable only as abstracts and are wanted rows.
• Vuoristo reports no numeric efficiencies in text; Trokovic's activation axis is not a purified cell fraction.
• The corpus search is finite; a clean matched paper could exist outside it.