EFB Dengue Literature Review

Mechanistic Case for DN and DN2 Cells in Dengue

21 min read

Mechanistic Case for DN and DN2 Cells in Dengue

Research Question

What is the mechanistic justification for measuring DN and DN2-phenotype B cells in acute dengue — and what can and cannot be claimed from a cohort in which DN2-like cells rise as a proportion of non-plasmablast B cells but not as a proportion of total B cells?

Purpose. Written as source material for the Discussion of the ABC stat analysis manuscript (HC n=13, DF n=12, DHF n=15; WHO-1997 severity). It is not a summary of that analysis — the SAP, memory.md and log.md in that project are the authority for the result. This page supplies the mechanistic literature the Discussion needs, with the provenance of each claim visible.

The finding this has to be written around

Stated as the frozen analysis states it, because the framing determines which mechanistic literature is relevant:

  • Primary endpoint met. DN2-like cells (CD11c⁺CD21⁻ within DN) as % of non-plasmablast B cells: 0.91% (HC) → 1.58% (DF) → 1.87% (DHF), Kruskal–Wallis p=0.0115.
  • The pre-specified denominator check diverged, and that is the substantive result. ASC (CD27⁺CD38⁺) rise from 0.75% of total B cells in controls to ~40–50% in dengue, roughly halving the non-ASC denominator. On a total-B denominator the effect does not hold (HC vs DHF p=0.32; HC vs DF reverses sign).
  • Much of the apparent elevation is therefore compositional. No claim of residual absolute elevation in DHF is made.
  • Post-hoc: DNQ4 (IgD⁻CD27⁻CD21⁻CD11c⁻) was the only population surviving the total-B denominator — exploratory, not pre-specified.

The honest one-line version: the DN2-like compartment is proportionally enriched among non-secreting B cells in dengue, in a severity-ordered way, at the same time as a massive antibody-secreting-cell expansion — and the current data cannot separate enrichment from redistribution.

That is a weaker claim than “severe dengue has more DN2 cells.” It is also, as §3 argues, the more mechanistically interesting one.

1. What the DN2 gate is measuring — and its known limits

  • The canonical definition is DN2 = IgD⁻CD27⁻CD38⁻CD24⁻CD21⁻, T-bet⁺CD11c⁺FcRL5⁺SLAMF7⁺CXCR5⁻, annotated as “extrafollicular ASC precursors”; DN1 = IgD⁻CD27⁻CD38⁺CD24⁺CD21⁺, CXCR5⁺FcRL5⁻, annotated “memory precursors” (see Sanz2019 - Consistent Classification of Human B Cell Populations, review — the nomenclature authority for this manuscript)
  • A CD11c⁺CD21⁻-within-DN gate without T-bet, FcRL5 or CXCR5 is a DN2-phenotype gate, not a confirmed DN2 identification. See DN2 Gating Strategy.
  • Undercounting: a CD21⁻CD27⁻ gate captures only 44.7% of transcriptomically defined atypical B cells; CD11c is the best single surface marker (see Sutton2021 - Alternative Lineage B Cells in Vaccination and Infection, human CITE-seq, core n=4)
  • Overcounting: a CD21⁻-anchored gate over-calls ABC by roughly 50% — only ~2/3 of CD21⁻CD23⁻ cells are T-bet⁺ and about half of those CD11c⁺ (see Cancro2020 - Age-Associated B Cells, review, murine)
  • CD21^low^ is not disease-specific. It marks activated memory in normal vaccination responses, HIV, malaria and checkpoint-inhibitor-expanded memory; early transitional T1 cells can also be CD21^low^ and require CD38/CD24/CD10 to exclude (see Sanz2019 - Consistent Classification of Human B Cell Populations, review)
  • CD27 is not a fixed marker. CpG stimulation upregulates CD27 on DN cells, so the DN gate boundary is activation-sensitive (see Wei2007 - DN Memory B Cells in SLE, human). In dengue’s high-TNF environment CD27 shedding is an additional concern.

For the Discussion: the two gate biases run in opposite directions, so the direction of the effect is more trustworthy than its magnitude. This argues against over-interpreting the absolute percentages and in favour of the severity ordering.

2. Why DN2 cells are worth measuring at all — the mechanism, in dependency order

The DN2 pathway is now specified at receptor level, in humans. This is the substance of the “so what?”

The generating signals, and their order (see Zumaquero2019 - IFN-gamma Programs T-bet-hi B Cells for ASC Differentiation, human in vitro + n=40 SLE):

  • Priming (early): BCR + IFN-γ. Omitting IFN-γ left >80% of cells T-bet^neg/lo^ and unable to upregulate IRF4 — IFN-γ is obligate for forming the T-bet^hi^ pre-ASC in this system.
  • Differentiation (late): IL-21. No ASCs formed at all without IL-21; late IL-21 alone was sufficient.
  • Throughout: TLR7/8. Early for survival, late for proliferation.
  • BCR must be transient. Continuous anti-Ig gave 2.8% ASCs versus 49% when restricted to days 0–3 — a ~17-fold penalty for continuous engagement.
  • Mechanism of the IFN-γ effect: epigenetic remodelling opening chromatin around T-bet, NF-κB, STAT5, IRF4 and BLIMP1 motifs, remodelling of the PRDM1 and IL21R loci, and a 5.5–6-fold rise in IL-21R protein with significantly increased IL-21-induced phospho-STAT3. See IFN-gamma, IL-21R, STAT3.

Why the cells are hyper-responsive — TLR7 hyper-responsiveness attributed to loss of the negative regulators TRAF5 and TNFAIP3, with R848 withdrawal causing >95% death (see Jenks2018 - DN2 B Cells and EF Pathway in SLE, human phospho-flow n=5–10). See TRAF5, Toll-like Receptor Signaling in B Cells.

They are not exhausted. DN2 cells retain intact proximal BCR signalling (BLNK phosphorylation), explicitly contrasted with FCRL4⁺ HIV cells; and the SLE (FcRL5⁺CD11c⁺) and HIV (FcRL4⁺) DN populations are phenotypically reciprocal, i.e. probably different cells (see Jenks2018 - DN2 B Cells and EF Pathway in SLE; Sanz2019 - Consistent Classification of Human B Cell Populations). See B Cell Receptor Signaling.

★ The dengue-specific problem the Discussion should name. The canonical priming signal is IFN-γ, but the early dengue response is dominated by type I IFN: type I IFN signalling was the top predicted upstream regulator of high-viral-load genes in acute dengue whole blood (see Kwissa2014 - Monocytes Drive Plasmablast Differentiation in Dengue, n=28). Whether type I IFN substitutes for IFN-γ in the priming window is untested — the source that raises the possibility notes only that IFNα- and IFN-γ-regulated gene sets overlap substantially. See Type I Interferon. This is a genuine gap, and naming it is more defensible than assuming the SLE mechanism transfers.

What is established in dengue — two independent blockade experiments, both on the plasmablast readout rather than DN2:

Neither has been run with a DN2 readout. That is the single most obvious next experiment and worth stating as such.

3. ★ The ASC expansion is not a nuisance — it is the mechanistically expected co-finding

This is the most useful thing the mechanistic literature does for this manuscript, and it should probably lead the Discussion.

DN2 cells are pre-antibody-secreting cells, and the evidence is quantitative:

Therefore: in an infection with a 50-fold ASC expansion, the pre-ASC compartment is being consumed as fast as it is being made. A DN2 pool that is proportionally enriched among non-secreting cells while not expanding against total B cells is precisely what a high-flux pre-ASC → ASC pipeline looks like. A static DN2 percentage under those conditions implies high throughput, not absence of activity.

What this licenses saying: that the compositional result is consistent with, and predicted by, the pre-ASC model — the DN2 compartment behaving as a transit population rather than an accumulating one.

What this does not license saying: that there is a real absolute expansion being masked. A flux argument is a hypothesis about rates, and frequencies at one timepoint cannot measure a rate. Two things would be needed: absolute counts (this cohort has no counting beads) and either serial sampling or a proliferation/differentiation marker. Do not use the flux argument to reinstate the retired sentence. Its legitimate use is to explain why the denominator divergence is interesting rather than merely disappointing, and to motivate the design that would settle it.

A concrete, cheap test for the next cohort: Ki-67 within the DN2 gate, and surface IL-21R. Under the pre-ASC/flux model DN2 cells should be Ki-67⁺ and IL-21R^hi^ in acute dengue; under a static-bystander model they should not. IL-21R is the better-motivated of the two because it is the specific node IFN-γ priming upregulates (5.5–6-fold), and because plasma IL-21 is uninformative — see §5.

4. Why an extrafollicular readout is worth having in dengue at all

5. The interpretive hazards this Discussion must concede

(a) Severity is a treacherous outcome variable. Three independent reasons, all already in the wiki:

This cohort is WHO-1997 DF/DHF with no WHO-2009 severe cases, so the finding is an association with DHF, not with severe dengue as currently defined. Worth stating explicitly. See Dengue Severity Classification.

(b) Day of illness is a confounder, not a covariate to mention in passing. The wiki’s own council downgraded the severity association in Ansari2025 - Peripheral T Helper Subset Drives B Cell Response in Dengue for day-of-sampling confounding. Kwissa2014 - Monocytes Drive Plasmablast Differentiation in Dengue sampled once per patient across days 2–9 and found viral load inversely correlated with duration of illness (r²=0.4272) — so any cross-sectional cohort mixes kinetics with severity. A single-timepoint design cannot separate “more DN2 in DHF” from “DHF patients sampled at a different point on the same curve.”

(c) Blood frequency may be a mobilisation readout. Two independent reasons:

A blood frequency can therefore rise because cells left a tissue. This is a further reason to be cautious about the flux argument in §3 — the same observation has at least two mechanistic readings.

(d) Serum cytokines are the wrong measurement for locally delivered signals. Plasma IL-21 showed no correlation with DN2 frequency in SLE (r=0.087) even though IL-21 is functionally required for ASC formation and IL-21R blockade removes 60% of the dengue plasmablast response (see Zumaquero2019 - IFN-gamma Programs T-bet-hi B Cells for ASC Differentiation; Ansari2025 - Peripheral T Helper Subset Drives B Cell Response in Dengue). Serum BAFF/APRIL likewise did not correlate with plasmablast magnitude (see GarciaBates2013 - Plasmablast Response and Dengue Severity) despite BAFF blockade reducing differentiation (see Kwissa2014 - Monocytes Drive Plasmablast Differentiation in Dengue) — see BAFF for the unresolved three-way tension. Do not add serum cytokine panels expecting them to corroborate a cellular finding.

(e) The DNQ4 result is post-hoc. It is worth reporting as exploratory, but there is no mechanistic literature in this wiki about a CD11c⁻CD21⁻ DN population under a Sanz2019 framework — Sanz2019 has no DN3, and its third DN row is defined by FcRL4, which this panel does not measure. Whether DNQ4 corresponds to anything in the published taxonomy is genuinely unknown. Flagged rather than interpreted. See DN3 B Cell for the competing DN1–DN4 scheme from Lamprinou2026 - ABCs and DN B Cells, which is not the same partition.

(f) The effector-function literature is weaker than it reads. All claims that atypical B cells secrete inflammatory cytokines trace to a single murine review, and the antigen-presentation claim is asserted by three reviews and mechanised by none. No human primary anywhere in this wiki measures cytokine secretion by sorted DN2 cells. See Atypical B Cell Effector Output. Avoid “DN2 cells drive inflammation through cytokine production” — it is not supportable from primary human data.

(g) ★ “Extrafollicular” is a location claim this study cannot make — and a consensus panel has now said so in print. Eisenbarth2025 - A Roadmap for Defining Extrafollicular B Cell Responses (Immunity, Nov 2025; twelve authors, Ignacio Sanz among them) concludes that “currently there are no flow cytometry-based means alone that can distinguish EF B cells nor their progeny from activated cells in earlier phases,” and that the term EF “should be avoided unless proliferation of antigen-specific B cells outside of a follicle is observed.” Applied by name to the cell this analysis is about: “the EF designation of this human DN2 cell refers to its presumed GC-independent origin rather than its location.”

This is narrower than it first reads, and the distinction is the whole point:

  • What survives untouched. Every mechanistic claim in §2 is an origin/process claim — IFN-γ/TLR7/IL-21 specification, the aNAV→DN2→ASC ordering, clonal connectivity, the epigenetic trajectory, CD40L antagonism. None of these assert a location. The low-SHM argument in §4 is likewise an origin argument. The mechanistic case does not weaken.
  • What must change is the wording. Claims of the form “dengue drives an extrafollicular response” are not licensed by a blood panel. The licensed form is “GC-independent”, stated with the criteria used. See GC-Independent Response for the evidence-to-claim mapping.
  • A second-order hit that compounds hazard (c). The same paper cautions that “low expression of CD21 and CXCR5 and increased CD11c could indicate recent B cell activation rather than a permanent state.” The DN2 gate is built from exactly these markers. So a single acute-timepoint frequency has now three candidate readings, not two: compartment enrichment, tissue egress (hazard c), or transient activation state. Serial sampling across fever days — which acute dengue uniquely permits — is the only design in reach that separates the third from the first.
  • One thing the wiki should not over-correct into. The paper does not claim DN2 cells are GC-derived. It withdraws a location inference, not the GC-independence inference. The supportable sentence is “as of Nov 2025, a consensus panel holds that EF should be reserved for imaged responses and that no flow panel alone establishes EF origin” — not “DN2 cells are not extrafollicular.”

(h) ★ The only direct human tissue data in this wiki say the CD11c⁺ DN compartment is scarce in tissue — and that is a hazard for a blood-only pilot built on it. Allard-Chamard2023 - DN3 B Cells Infiltrate Inflamed Tissues quantified DN subsets by immunofluorescence in COVID-19 thoracic lymph node (n=6) and IgG4-RD submandibular gland (n=10 vs n=7 sialadenitis controls). Stated on the axis this analysis can actually use — CD11c⁺ vs CD11c⁻ within the tissue DN pool, which is gate-independent — CD11c⁺ DN cells are ~4% of tissue DN in COVID-19 lymph node (~24 vs ~630 cells/mm²) and ~10% in IgG4-RD gland (~25 vs ~218 cells/mm²). The CD11c⁻ subsets dominate both tissues. The result is robust to the unresolved DN4 dispute — the primary gates DN4 as CXCR5⁺CD11c⁺ while Lamprinou2026 - ABCs and DN B Cells relays it as CXCR5⁺CD11c⁻ (recorded on Double-Negative B Cell) — because on the second reading the CD11c⁺ share falls further, to ~1% and ~2.5%.

Why this is stated as a pooled CD11c contrast rather than as “DN2 is 7 cells/mm²”: that paper gates DN on CXCR5 × CD11c with no CD21 in the panel, so its DN2 (CXCR5⁻CD11c⁺) is not this cohort’s DN2-phenotype (CD21⁻CD11c⁺). Quoting its per-subset numbers against a CD21-based gate would be a cross-gate comparison presented as like-for-like — the exact error this wiki has flagged elsewhere (see DN3 B Cell Contradictions; DN2 Gating Strategy). Pooling on CD11c is the one comparison both panels support.

What it does and does not do to the case in §2. It does not touch any mechanism claim — those are origin/process arguments about blood cells, and the paper measures tissue. What it does is remove an unstated assumption: that a rising blood DN2-phenotype frequency is a proxy for DN2 cells accumulating somewhere that matters. In the only human tissue anyone has counted, the CD11c⁺ DN population is a small minority while CD11c⁻ DN cells are abundant. This compounds hazards (c) and (g) into a single problem — a blood frequency at one timepoint may reflect a compartment, an egress event, a transient activation state, or a population that is largely confined to blood. Four readings, one number.

Four reasons not to over-read it, all from the source. (i) The whole comparison rests on CD11c detection in FFPE tissue, and the paper’s own open questions ask whether CD11c is simply lost on fixation — if it is, the tissue CD11c⁺ deficit is partly artefactual. (ii) The COVID-19 autopsy cohort is late-phase only (15–36 days), so it says nothing about the acute d5–8 window this pilot samples. (iii) No significance markers are printed on the subset-level panels (Fig. 5C, 6D); n=6–10 per arm, uncorrected, and Fig. 5C has no control lymph nodes at all. (iv) Two diseases, neither of them an acute arboviral infection. The honest form is a hazard, not a refutation: the tissue behaviour of the CD11c⁺ DN compartment in acute dengue is unmeasured, and the assumption that blood frequency indexes tissue presence is now known to fail in two chronic inflammatory diseases.

6. What can be claimed, and what cannot

Supportable from the literature plus this cohort:

  • DN2-phenotype cells are proportionally enriched among non-secreting B cells in acute dengue, with ordering HC < DF < DHF (p=0.0115)
  • The enrichment coincides with a large ASC expansion that roughly halves the non-ASC denominator, so it is substantially compositional
  • DN2 cells are mechanistically specified pre-ASCs in humans, generated by IFN-γ/TLR7/IL-21 with a defined temporal order, and are hyper-responsive rather than exhausted
  • Both known dengue B-cell-helper mechanisms — Tph/IL-21 and monocyte/BAFF-APRIL-IL-10 — converge on the extrafollicular output that DN2 cells feed
  • Dengue’s antibody response bears extrafollicular hallmarks (low SHM), and EF-derived antibody is cross-reactive and ADE-competent
  • A compositional enrichment in a high-flux pre-ASC system is consistent with the DN2 compartment operating as transit rather than accumulation

Not supportable, and should not be written:

  • That DHF patients have an absolute expansion of DN2 cells (retired by the curator’s own decision; the total-B denominator does not support it)
  • That the finding relates to WHO-2009 severe dengue (no such cases in the cohort)
  • That the cells measured are confirmed DN2 (no T-bet, FcRL5 or CXCR5 in the gate) — “DN2-like” or “DN2-phenotype” throughout
  • That DN2 frequency is a validated biomarker in any infection (no ingested source tests it prospectively)
  • That the dengue IFN environment primes the DN2 pathway (type I, not IFN-γ; substitution untested)
  • That these cells produce the autoantibodies or the ADE-competent antibodies in dengue — no functional antibody output has ever been measured from sorted DN cells in any infection, which Sutton2021 - Alternative Lineage B Cells in Vaccination and Infection states as its own limitation
  • That a rising blood DN2-phenotype frequency indexes DN2 cells accumulating in tissue. The only human tissue counts available show the CD11c⁺ DN compartment is a small minority of tissue DN cells in two chronic inflammatory diseases (see hazard (h), Allard-Chamard2023 - DN3 B Cells Infiltrate Inflamed Tissues) — measured on a CD11c axis, late-phase, and with CD11c-in-FFPE unresolved, so it is a reason for caution in wording rather than a finding to assert about dengue
  • That the response measured is extrafollicular. A blood panel cannot establish location; the licensed claim is GC-independent, and even that is an inference from converging evidence rather than a determination (see hazard (g), GC-Independent Response, Eisenbarth2025 - A Roadmap for Defining Extrafollicular B Cell Responses)

Open Questions

  • Does type I IFN substitute for IFN-γ in the priming window? The single most consequential unknown for transferring the DN2 mechanism into dengue.
  • Are dengue DN cells FcRL5⁺ (SLE-like effector) or FcRL4⁺ (HIV-like)? Sanz2019 - Consistent Classification of Human B Cell Populations shows these are reciprocal; adding FcRL4 and FcRL5 would place dengue on that axis and is the highest-information panel addition available.
  • Is the DN2 compartment in acute dengue proliferating (Ki-67⁺) and IL-21R^hi^, as the pre-ASC/flux reading predicts?
  • Would absolute counts (counting beads) plus serial sampling separate enrichment from redistribution?
  • Does either dengue blockade system — Tph/IL-21R or monocyte/BAFF — actually generate DN2 cells, or only plasmablasts?
  • What is DNQ4 under a Sanz2019 framework, and does it correspond to the FcRL4⁺ DN row?
  • Do human DN2 cells undergo marginal-zone-type retention, and could blood frequency be tracking egress?
  • Is the CD11c⁺ DN compartment scarce in inflamed tissue in acute infection as it is in late-phase COVID-19 and IgG4-RD — or is that a chronic-disease and late-timepoint phenomenon? And is the scarcity real or a CD11c-epitope-in-FFPE artefact? Nothing in this wiki answers either, and both bear directly on what a blood DN2-phenotype frequency is a measure of.
  • Can serial sampling across fever days separate a durable DN2 compartment from a transient CD21ˡᵒCD11c⁺ activation state? This is the one discriminator acute dengue affords that the SLE and vaccination cohorts defining these gates never used.
  • If “extrafollicular” is not claimable from blood, what is the strongest available surrogate for GC-independence in a dengue cohort — SHM load, isotype distribution, CXCR5/CD21 kinetics across timepoints, or paired GC-marker absence?

Why DN B Cells Matter - Disease Relevance and Infectious Disease Case, DN2 B Cell, Double-Negative B Cell, DN3 B Cell, Atypical B Cell, Plasmablast, Extrafollicular Response, Atypical B Cell Effector Output, Toll-like Receptor Signaling in B Cells, B Cell Receptor Signaling, Follicular Exclusion, Extrafollicular T Cell Help, IFN-gamma, Type I Interferon, IL-21R, BAFF, Inflammatory Monocyte, DN2 Gating Strategy, Dengue Severity Classification, Research Plan - DN B Cell Expansion in Dengue, Thesis Objectives and Grant Pitch, GC-Independent Response, Conventional Flow Cytometry

Sources Used

Dengue primaries: Kwissa2014 - Monocytes Drive Plasmablast Differentiation in Dengue, Ansari2025 - Peripheral T Helper Subset Drives B Cell Response in Dengue, GodoyLozano2016 - Lower IgG SHM Rates in Acute Dengue, GarciaBates2013 - Plasmablast Response and Dengue Severity, Priyamvada2016 - Cross-Reactive Memory Plasmablasts in Secondary Dengue, Appanna2016 - Plasmablasts as Subset of Memory B Cell Pool, Singh2026 - DENV-Specific Memory B Cell Subsets, Narvaez2011 - Evaluating WHO Dengue Severity Classifications, Morra2018 - Defining Warning Signs and Severe Dengue

Mechanism primaries: Zumaquero2019 - IFN-gamma Programs T-bet-hi B Cells for ASC Differentiation, Jenks2018 - DN2 B Cells and EF Pathway in SLE, Song2022 - Tfh Outside Germinal Centers Drive T-bet CD11c B Cells, Scharer2019 - Epigenetic Programming in SLE B Cells, Wei2007 - DN Memory B Cells in SLE, Tipton2015 - ASC Diversity and Origin in SLE, Woodruff2020 - EF B Cell Responses in COVID-19, Kaneko2020 - GC Loss and TFH Block in COVID-19, Sutton2021 - Alternative Lineage B Cells in Vaccination and Infection, Allard-Chamard2023 - DN3 B Cells Infiltrate Inflamed Tissues (human tissue immunofluorescence; the CD11c⁺-DN-scarcity hazard (h) — note its DN gate is CXCR5 × CD11c, not CD21 × CD11c)

Reviews (Tier B — nomenclature and framing only): Sanz2019 - Consistent Classification of Human B Cell Populations, Sanz2025 - Human Atypical B Cells Overview, Cancro2020 - Age-Associated B Cells, Lamprinou2026 - ABCs and DN B Cells, Glaros2025 - Multilayered Identity of B Cell Memory, Eisenbarth2025 - A Roadmap for Defining Extrafollicular B Cell Responses (consensus Perspective — bounds the EF claim to origin, not location; drives hazard (g))

Graph View

Backlinks