Nature Immunology Response to Reviewers: Revision Guide

An added experiment can strengthen one link while leaving another unresolved. State exactly which link changed and which claim was revised.

Create a ledger for tissue, collection time, disease or exposure, treatment, processing, viability, enrichment, gating, marker threshold, batch, donor or animal, repeated sampling, and final denominator. Preserve raw FCS files and analysis settings where applicable.

A marker panel does not define a mechanism. A transcriptomic cluster does not establish lineage. A depletion does not prove cell-intrinsic function if the intervention changes development or another compartment.

Collaborative language is not capitulation. It makes the scientific agreement, disagreement, and remaining uncertainty easier for professional editors and referees to evaluate.

In our pre-submission review work with Nature Immunology revisions, we inspect cohort and animal definitions, tissue provenance, flow and single-cell analysis, perturbations, molecular assays, imaging, host models, endpoints, statistical hierarchy, source data, figures, reporting, abstracts, and claims. We audit each link from immune identity to host-level inference, and we observe the patterns below when cell state, perturbation, statistical unit, and abstract claim refer to different biological boundaries. These are qualitative manuscript patterns, not private Nature Immunology decisions.

Pattern 1: a cell state changes when the gate changes

One marker, threshold, integration choice, or cluster label defines the proposed population. We rerun identity across orthogonal assays and perturb the analysis choices. If the population is unstable, we describe a continuum or context-dependent state rather than a discrete lineage.

Pattern 2: abundance becomes mechanism

A cell type, cytokine, transcription factor, or receptor tracks disease or outcome, and the manuscript calls it a driver. We look for temporal order, cell-specific perturbation, rescue, pathway alternatives, and a functional consequence. Correlation can support a biomarker claim without establishing mechanism.

Pattern 3: developmental and acute perturbations are conflated

A constitutive knockout changes immune development, tissue architecture, or baseline composition, but the phenotype is interpreted as an acute pathway function. We add inducible, blocking, transfer, chimera, or rescue evidence where feasible and state what remains confounded.

Pattern 4: one host context carries a universal claim

One pathogen, tumor, autoimmune model, mouse background, donor group, or treatment condition becomes "immunity" in the abstract. We choose the external context most likely to break the proposed mechanism and narrow the claim when transfer is incomplete.

The distinctive information gain is immune-identity alignment: cell state, compartment, time, perturbation, mechanism, host context, and claim must describe the same biology.

Pattern 5: the denominator changes across modalities

Flow cytometry reports animals, single-cell analysis reports cells, imaging reports fields, and functional assays report wells, while the manuscript compares all four as if they estimate the same biological replication. In our Nature Immunology review work, we rebuild the denominator from donor or animal through tissue, sample, aliquot, field, and cell. We then use the biological unit for inference and retain lower levels as measurement detail. This often preserves the biological pattern while widening intervals and removing an unsupported subgroup claim.

Check the response and revised immune evidence together.

One referee may ask for deeper molecular mechanism while another asks for human validation, another tissue, or a second disease model. Tell the editor which uncertainty controls the main conclusion. A rescue experiment, donor cohort, spatial analysis, and second host model answer different questions.

Do not imply that evidence from one level compensates for another. A stronger mouse mechanism does not establish human treatment response. A larger human association does not establish causal order. A spatial colocalization does not prove a ligand-receptor mechanism.

Revision is not acceptance. The most serious rejection-on-revision risks include unstable cell identity, a perturbation that changes development rather than the proposed acute mechanism, technical units counted as independent biology, a second model that repeats the same dependency, selective donor or subgroup reporting, and an abstract that still generalizes beyond tissue, host, disease, or treatment evidence.

Most failed revision packages answer every numbered comment but preserve the same controlling causal or transport gap. Adding more markers or cells does not repair an invalid biological unit or a non-specific perturbation.

Submit if: every comment is answered and located; cell identity survives orthogonal evidence; tissue and time are explicit; perturbations are specific and interpreted within their limits; donor or animal hierarchy is correct; mechanism language matches necessity, sufficiency, and rescue evidence; and claims stay within the tested host context.

Think twice if: one marker or clustering choice defines the state, constitutive knockout is treated as acute proof, technical cells inflate n, human samples are called clinical validation without design support, or the abstract retains a universal mechanism after context-dependent revision results.

We reviewed current Nature Immunology preparation and reviewer guidance, current Nature Portfolio revision guidance, the PLOS response-writing canon, and Nature Computational Science response guidance. Current journal instructions and the decision letter control required files. The identity-to-host evidence chain is Manusights analysis.

1. Put the editor's controlling issues before referee sections.
2. Answer every point with action, result, claim impact, and exact location.
3. Reconcile cell identity across markers, gates, modalities, lineage, and function.
4. State tissue, time, exposure, treatment, processing, and denominator for every comparison.
5. Separate developmental, cell-extrinsic, and acute perturbation effects.
6. Match causal language to necessity, sufficiency, rescue, and pathway evidence.
7. Use donor or animal as the biological unit and model repeated measures correctly.
8. Test the most important species, host, tissue, disease, or treatment boundary.
9. Synchronize response, figures, legends, methods, source data, code, and supplement.
10. Propagate every narrowed claim through title, abstract, results, and discussion.

Use the first 14 complete Search Console days only as an indexation and query-fit read for this Nature Immunology response owner. At 21 days, decide whether to keep, refine, merge, or remove it using exact-query impressions, clicks, competing Manusights URLs, and qualified response-review starts. One preview start and no journal-impression proxy make demand highly uncertain.

Nature Portfolio sources establish the public revision framework. The immune-mechanism audit is Manusights interpretation.