Decision letter

Decision Letter

VERDICT: major

Summary of Evaluation

This manuscript combines a substantial paired-biopsy spatial transcriptomic dataset (22 RA patients pre- and post-treatment, 2 healthy donors, ~2M cells) with a mechanistic chain running from NOTCH3 → NGF induction → neurotrophin receptor expression → fibroblast acquisition of mural-cell markers and contractility, and closes with pharmacology in human RA synovial explants using two FDA-approved TRK inhibitors. The panel was consistent in judging the underlying biology interesting and the experimental triangulation unusually broad for a study of this kind: three orthogonal modalities support neurotrophin receptor expression on synovial mural cells, and gain- and loss-of-function approaches (DLL4, siRNA, CRISPR, NGFR overexpression, receptor agonists and antagonists) converge on the same directionality. The bidirectional pharmacology in explants, combined with the viability data in Fig. S9B, makes nonspecific toxicity an unlikely sole explanation of the drug effects. This is a real contribution and the therapeutic lead is concrete.

Two problems, however, sit directly on load-bearing claims and cannot be resolved by wording alone.

First, the paper's motivating claim — that pathological vascular maturation persists despite 6 months of immunosuppression — is not unambiguously supported by the statistics as reported. The Results text gives paired p-values per cell type in which only the second is explicitly labelled "compared to healthy" (e.g. "pericytes p = 1.6e-05, p = 0.029 compared to healthy"), so the referent of the first value is undetermined in the text, and the figure legend states only that a paired Wilcoxon test was used "for paired patient samples." No multiple-comparison correction is disclosed for a family of roughly twelve tests across six vascular subtypes and three group contrasts. The debate did not resolve this: the advocate's reading (first value = paired pre-vs-post) is plausible and several effects would likely survive Bonferroni, but no passage in the manuscript settles it, and the abstract-level claim depends specifically on the within-patient contrast. Because the outcome of an explicit, corrected reanalysis could in principle alter which conclusions stand, this is a major revision rather than a labelling fix. Relatedly, the statement that the increase occurred "regardless of whether or not patients reached criteria for clinical remission" is asserted without any shown stratified analysis or clinical data table.

Second, "reversal of pathological vascular maturation" overstates what the explant data establish. Three reviewers independently converged here, each from a different angle. Three days of ex vivo drug exposure reducing αSMA and PECAM1 signal is consistent with acute suppression of a maturation program; it does not distinguish dedifferentiation of established mural cells from selective loss of recently differentiated ones, does not demonstrate movement toward a normal vascular state (no healthy-donor explant comparator is reported), and does not establish durability. The advocate conceded that "reversal" should become "acute suppression." I agree, and I am treating this as a required claim revision rather than a demand for new experiments — but the title, abstract, Fig. 6 title and Discussion all currently carry the stronger word.

Beyond these, the panel and the compliance audits identified a dense set of reporting deficits that are individually fixable but collectively serious for a venue whose reviews are read by people deciding whether to trust a preprint: sample sizes (n biological replicates) are not stated for any in vitro or explant experiment; the quantification metrics behind the headline Fig. 6 numbers ("aSMA intensity," "vascular density," "PECAM1-positive vascular structures") are not defined; the concentrations of larotrectinib and entrectinib used in the explant experiments are not given; no accession numbers or code repository are provided for the Xenium or bulk RNA-seq data, and the RNA-seq methods (platform, pipeline, DE thresholds, the 461-gene list) are essentially absent; Tables 1, 2 and S1–S3, which carry the primer, probe and custom-panel content, are referenced but not supplied. The citation audit also flags one apparent misattribution: reference 2 (Tamiato et al., cardiac pericyte RGS5) is cited for endothelial–fibroblast crosstalk in RA, and reference 13 (Veale & Fearon, psoriatic arthritis) is cited alongside it for the same RA claim. These need correction, not defence.

I want to be explicit about what I am not requiring. I am not requiring an animal model, lineage tracing, promoter-occupancy assays, or a validation cohort as conditions of acceptance. Those would strengthen the paper and are listed as suggestions. What is required is (i) an unambiguous, corrected reporting of the Fig. 1 comparisons with the conclusions adjusted to whatever survives, (ii) the stratified remission analysis or removal of that claim, (iii) claim language matched to the evidence for the explant pharmacology and the NOTCH3/NGFR mechanism, and (iv) the reporting, deposition and citation repairs. If the corrected Fig. 1 analysis holds, the paper's architecture is unchanged and this is a straightforward revision; the verdict reflects the fact that the outcome is not knowable from the manuscript as written.

Ethics and compliance are otherwise in order: IRB approvals for both sites are stated and funding is disclosed. An explicit informed-consent statement and inclusion/exclusion criteria are missing and should be added.

Required Revisions

1. Disambiguate and correct the Fig. 1G–H statistics. For every vascular subtype, state explicitly which two groups each reported p-value compares (healthy vs. pre-treatment RA; healthy vs. post-treatment RA; paired pre- vs. post-treatment RA), which test was applied to which contrast (paired Wilcoxon is not valid for the healthy comparisons), and apply a stated multiple-comparison correction across the pre-specified family of tests. Report effect sizes with confidence intervals alongside p-values. If any of the pericyte, VSMC, capillary or arteriolar contrasts do not survive correction, revise the abstract, title claim ("sustains"), Results and Discussion to match what remains supported, and label the remainder exploratory.

2. Support or remove the remission-independence claim. Provide the analysis stratifying vascular density change by 6-month DAS28-ESR remission status, together with a clinical table for the cohort (age, sex, disease duration, seropositivity, treatment arm, baseline and 6-month DAS28-ESR, per-patient assignment). If the subgroup is underpowered, say so and delete the sentence. Also state the number of healthy donors (n = 2) as an explicit limitation wherever healthy comparisons are drawn.

3. Report sample sizes and define quantification metrics for all cell-culture and explant experiments. For each panel in Figs. 3–6 and Figs. S3–S9, state n as the number of independent biological replicates and, for tissue work, the number of distinct patient donors versus sections per donor. Define precisely how "aSMA intensity," "vascular area," "aSMA-positive vascular structures per section" and "PECAM1-positive vascular structures" were computed (numerator, denominator, thresholding, whether per-structure or per-tissue-area), and report confidence intervals for the Fig. 6E–G effect sizes. State the concentrations of larotrectinib, entrectinib, DAPT, GNF5837, GW441756, ANA-12, LM22B-10 and 7,8-DHF actually used in the explant experiments, and the treatment duration for each.

4. Match the claim language for the TRK-inhibitor experiments to the evidence. Replace "reverse/reversal" with language describing acute suppression of vascular maturation markers in short-term ex vivo culture (title, abstract, Fig. 6 title, Results and Discussion). State plainly in the Discussion that the current data do not distinguish dedifferentiation of established mural cells from selective loss of recently differentiated cells, that no healthy-donor explant comparator was tested so RA-specificity is unestablished, and that durability after drug withdrawal was not assessed.

5. Temper the NOTCH3→NGF and NGFR-sensitization mechanistic claims to the level of evidence, and complete their reporting. For Fig. 5C–E, state n, report confidence intervals for the 1.2-fold NGF induction, clarify whether panels C and D are independent replicates or the same samples assayed two ways, and note that DAPT is a pan-γ-secretase inhibitor. Since no promoter-occupancy or reporter data are presented, avoid "transactivation" and "directly activates" and describe NOTCH3 as required for NGF induction. For Fig. 5F–I and S8, provide densitometry quantified across replicates rather than representative blots alone, report lentiviral transduction efficiency and whether selected/sorted populations were used, and either report endogenous NGFR and TRKA protein levels in RA synovial fibroblasts or state explicitly that the sensitization model is inferred from overexpression.

6. Deposit the data and code, and supply the missing tables. Provide accession numbers for the Xenium spatial transcriptomic data (raw and processed matrices) and for the bulk RNA-seq, plus a public repository for the analysis code. Add a data and code availability statement; "available on request" is not sufficient here. Supply the full 461-gene NGF/NGFR signature with fold changes and adjusted p-values, and complete bulk RNA-seq methods (platform, library kit, read configuration, depth, reference genome, alignment and DE tools, thresholds). Include Table 1 (primers), Table 2 (RNAscope probes) and Tables S1–S3 (custom panel gene lists) in the submitted materials.

7. Correct the citation problems. Reference 2 (Tamiato et al., cardiac pericyte RGS5) does not support the statement that endothelial cells orchestrate synovial fibroblast differentiation in RA; replace or remove it. Reference 13 (Veale & Fearon, psoriatic arthritis) is likewise not an RA source for that claim. Verify that reference 36 (Donovan et al. 2000, BDNF/endothelial survival) supports the stated pericyte-migration and VSMC-function phenotypes in TRKB-null mice, and that reference 37 (Tessarollo et al. 1994, muscle sensory neurons) supports the stated vascular abnormalities in NT3-null mice; substitute appropriate primary sources where they do not. Give reference 27 a full citation. Finally, state concretely which parts of the spatial cohort, panel design and analysis pipeline are newly generated here versus carried over from Bhamidipati et al. (ref. 1), so readers can judge the incremental contribution.

8. Complete the Methods to the point of reproducibility. Add: an explicit informed-consent statement and patient inclusion/exclusion criteria; Xenium QC thresholds (transcripts and features per cell, any excluded samples or cells) and the clustering parameters actually used (number of PCs, Harmony settings, Louvain resolution for both the lineage and vascular-subtype levels); antibody host species, clonality and working dilutions for all Western blot, immunofluorescence and whole-mount applications; the specific antibodies, catalog numbers and dilutions used for IHC (currently delegated to a core facility and therefore untraceable); the ELISA kit(s) used for NGF/BDNF/NT3; siRNA and CRISPR guide RNA sequences with evidence of NOTCH3 knockout validation; and a self-contained collagen gel contraction protocol (collagen source, gel geometry, cell density, duration, measurement method). Add mycoplasma testing status for the fibroblast lines and HUVECs, and the HUVEC source catalog number.

9. State the novelty position explicitly. In the Introduction or Discussion, say what is new relative to the established developmental literature on neurotrophins in mural-cell biology (refs. 36–38): whether the NOTCH3→NGF→NGFR/TRKA axis described here is proposed as RA-specific, or as a developmental program sustained pathologically in RA synovium. The current framing leaves this unresolved and invites the criticism that known biology has been relocated to a new tissue.

Minor Suggestions

• Test larotrectinib/entrectinib on healthy-donor synovial explants. This is the single most informative addition for the therapeutic argument and would let the RA-specificity claim be made rather than hedged.
• Single-cell or lineage-tracing analysis of treated explants to distinguish mural-cell dedifferentiation from selective loss, and a withdrawal arm to address durability.
• Dose–response and time-course for the pan-TRK and FDA-approved inhibitors in explants; at present a single concentration and timepoint carry the pharmacology.
• Determine which TRK isoform(s) mediate the explant effect; larotrectinib and entrectinib are pan-TRK, and entrectinib additionally hits ROS1/ALK. A brief statement on plausible off-target kinases expressed in synovium would help readers weigh the result.
• Report a sensitivity analysis for the Xenium clustering resolution and, if feasible, an assessment of annotation accuracy (manual curation of a held-out subset) for the fine-grain vascular subtypes.
• Test whether csDMARD or TNFi exposure alters NGF or NOTCH3 expression in the paired biopsies. This would connect the mechanism directly to the treatment-resistance framing that motivates the paper.
• Compare the neurotrophin-stimulated fibroblast transcriptome to native pericyte and VSMC profiles, and assess persistence after neurotrophin withdrawal, to substantiate "differentiation" over "marker induction."
• Overlay individual data points on bar plots throughout (Fig. 3D–E and several supplementary panels currently do not), and state the R version and Cellpose/scikit-image/ImageJ versions.
• Clarify the fibroblast:endothelial ratio used in each co-culture experiment; the Results state 1:3 while the Methods list 1:1, 1:5 and 1:10.
• Consider softening the Discussion paragraph linking neurotrophin signaling to sensory nerve coupling, or label it explicitly as hypothesis; it is not tested here.