Beckers2023 - Origins and Functions of DN B Cells

Full citation: Beckers, L., Somers, V., & Fraussen, J. (2023). IgD⁻CD27⁻ double negative (DN) B cells: Origins and functions in health and disease. Immunology Letters, 255, 67–76. https://doi.org/10.1016/j.imlet.2023.03.003

Raw file: [[raw/Beckers2023.pdf]]

Summary

A narrative review from the Hasselt University group (University MS Center / Biomedical Research Institute, Belgium) that consolidates the DN B cell literature as of early 2023. It is structured around three questions: what DN B cells are phenotypically, where they come from, and what they do. The authors are themselves primary contributors to the field — a substantial part of the multiple sclerosis (MS) and healthy-donor DN data cited is their own — refs [18] (Fraussen 2019, J Immunol) and [43] (Claes 2016, J Immunol) — which makes those portions a de facto primary source rather than pure synthesis. See §Reference Map below for the verified bracket→reference mapping and for what is not own data.

The review’s central organising claim is that DN B cells are not one population. They comprise at least three (possibly more) subsets with different origins, and the field’s inability to agree on their classification is the primary obstacle to interpreting the literature. Three origin hypotheses are laid out and none is resolved: (1) DN1 cells prematurely exit the germinal centre; (2) DN cells descend from switched memory (SM) cells via CD27 downregulation under chronic antigen stimulation / immunosenescence; (3) DN2/DN3 cells arise entirely GC-independently via the extrafollicular pathway from activated naive or unswitched memory precursors. Critically, single-cell RNA velocity data are cited as supporting flow in both directions between DN1 and SM cells, so the precursor/progeny relationship is not settled.

Functionally, the review separates DN2 (established ASC precursors, TLR7-hyper-responsive, autoreactivity-associated) from DN1 and the other subsets (function largely undetermined), and catalogues antibody-independent functions: antigen presentation, pro-inflammatory cytokine secretion (lymphotoxin-α, TNF-α), granzyme B production, and migration to inflamed tissue via CXCR3/CCR6. Table 1 is the review’s most-cited artefact: a condition-by-condition survey of human peripheral blood DN B cells across ~25 diseases and states — health, aging, thirteen autoimmune diseases, six infections, vaccination, and five “other” conditions including obesity and lung cancer. Dengue appears nowhere in the review.

Study Design

• Type: Narrative review (not systematic; no stated search strategy, inclusion criteria, or quality appraisal). 86 references.
• Sample size: N/A — review. Table 1 aggregates ~70 primary studies of human peripheral blood (and, where noted, synovial fluid, CSF, tonsil, gut, and tumour tissue) DN B cells.
• Setting: Human studies only. The review explicitly restricts its scope to “human B cells defined as CD19⁺IgD⁻CD27⁻ B cells unless mentioned otherwise.”
• Population: Healthy donors (young <60 and aged >60), autoimmune disease patients (SLE, pSS, SSc, RA, JIA, axSpA, MS, NMOSD, myasthenia gravis, Guillain–Barré, Hashimoto’s thyroiditis, Graves’ disease, IBD), infected patients (meningitis/encephalitis, acute sepsis, malaria, rotavirus, HIV, COVID-19), vaccinees (influenza, tick-borne encephalitis), and other conditions (Alzheimer’s, CVID, ALS, obesity, NSCLC).

Key Findings

Scope gap — dengue is absent

• The review surveys DN B cells across ~25 human conditions, including a dedicated “Infections” section (§8.1) and infection block in Table 1 — meningitis/encephalitis, acute sepsis, malaria, rotavirus, HIV, COVID-19 — plus a vaccination block (influenza, tick-borne encephalitis virus). Dengue is not mentioned in the text, the table, or the reference list. As of early 2023, DN B cells in dengue had not entered the field’s landmark disease survey.

Phenotype and memory credentials

• DN B cells are “generally described as antigen-experienced and mostly Ig isotype switched.” The majority of circulating DN cells in young and aged healthy donors (HD) and MS patients are IgG⁺, like SM B cells (Fraussen 2019 / Claes 2016 — the authors’ own data).
• ABCB1 transporter: DN B cells are ABCB1-negative in both aged HD and SLE patients — matching SM B cells and in contrast to naive and unswitched memory (USM) cells. This is the molecular basis for the Rhodamine 123 retention phenotype.
• Telomere length: DN telomeres are highly similar to SM B cells and significantly shorter than naive and USM cells — replicative-history evidence for antigen experience.
• Morphology: DN cells resemble USM and SM cells in FSC/SSC (size and granularity), both significantly higher than naive B cells.
• Developmental markers CD10, CD24, CD38 are broadly similar between DN and SM cells; in HD and MS, DN cells show similar CD10 but higher CD38 and CD5 and lower CD95 than SM cells.
• Increased DN frequencies occur after vaccination against influenza or tick-borne encephalitis virus, arguing for antigen-driven maturation on stimulation rather than a purely degenerate/terminal state.
• Ig VH genes of DN B cells carry significant somatic hypermutation, though at a lower load than SM cells in multiple studies.

Exhaustion / immunosenescence signature

• In young and aged HD, circulating DN B cells show low expression of the anti-apoptotic protein Bcl2 and express senescence-associated secretory phenotype (SASP) markers: TNF-α, IL-6, IL-8 (pro-inflammatory cytokines), p16^INK4 (cell-cycle regulator), and inflammatory microRNAs miR-155, miR-16, miR-96.
• DN and CD20^hi CD27⁻CD21^lo B cells express multiple inhibitory receptors — FcRH3-5, CD22, CD85j — in HIV-infected, malaria-infected, anti-SARS-CoV-2-immunised individuals, and young and aged HD.
• ⚠ But the inhibitory-receptor phenotype splits by disease class: DN B cells lacking inhibitory receptor expression have been reported in the autoimmune diseases SLE and Hashimoto’s thyroiditis. The review’s own synthesis: “Part of the DN B cells shows characteristics of exhausted cells in aged and virally infected individuals, while it is suggested that DN B cells in autoimmune diseases have a higher activation state.”
• CD32b (inhibitory FcγRIIb) is significantly decreased on DN B cells from Hashimoto’s thyroiditis patients — the authors suggest these DN cells are more susceptible to stimulation.
• FOXO1: In SLE, a DN subset with cytoplasmic FOXO1 expression is significantly increased and correlates with disease activity — hypothesised to associate with a high activation state.

Classification — the field has no consensus

• Sanz group (2018): DN1 = CD19⁺IgD⁻CD27⁻CD11c⁻CD21⁺CXCR5⁺; DN2 = CD19^bright IgD⁻CD27⁻CD11c⁺CD21⁻CXCR5⁻ (relayed from Jenks2018 - DN2 B Cells and EF Pathway in SLE).
• 2020: a third subset DN3 = CD19⁺IgD⁻CD27⁻CD11c⁻CD21⁻CXCR5⁻ identified in COVID-19 (relayed from Woodruff2020 - EF B Cell Responses in COVID-19); DN1-3 subsequently confirmed by others in COVID-19 and HD.
• A competing orthogonal scheme (Szelinski et al., Arthritis Rheumatol 2022, ref [37]): DN cells classified by CD19 intensity + CXCR5 into DN^int (CD19^int IgD⁻CD27⁻CXCR5⁺), DN^hi (CD19^hi IgD⁻CD27⁻CXCR5⁻), and DN^low (CD19^low IgD⁻CD27⁻CXCR5⁺). DN^int and DN^hi share phenotypic characteristics with DN1 and DN2 respectively. DN^low is described as a new antigen-experienced DN subset that is increased in SLE, LACKS CD11c expression, yet shares the phenotype and transcriptome with plasmablasts. Szelinski also reports DN^low + DN^hi elevated in SLE, pSS, RA, and COVID-19 (Table 1). The review notes DN^low needs validation and its relationship to DN1-3 is unresolved.
• Terminological overlap is acknowledged: DN cells “overlap to some extent with CD21^-/low or CD11c^hi age-associated B cells (ABCs) or atypical memory B cells.”

Three origin hypotheses (Fig. 1) — none resolved

(1) Premature exit from the GC reaction — proposed for DN1

• DN1 cells express CXCR5, the follicle-homing chemokine receptor.
• Number of cell divisions and Ig mutations for IgG⁺CD27⁻ DN cells is similar to GC B cells, but the mutation load is lower than SM cells — consistent with departure before mutation load accumulates.
• Class-switched (IgA⁺, IgG⁺) CD27⁻ B cells have significantly higher VH mutation than naive and USM cells but lower than SM cells.
• Implication: DN1 could be a precursor of CD27⁺ memory cells caught at an earlier developmental state.
• CDR3 regions of DN cells are smaller, more hydrophilic and basic than naive; but larger, more hydrophobic and more acidic than USM/SM.

(2) Descent from SM B cells via CD27 downregulation

• AIRR sequencing shows a small clonal overlap between DN and SM cells: 0.2–2.2% (the authors’ own study), with a similar mutation load in the clonally related cells.
• Another AIRR study found genealogical trees pointing to DN cells as the progenitor of SM cells — i.e., the arrow runs the other way. No significant differences in VH gene family usage or in the types/locations of Ig mutations between class-switched DN and SM cells; that study concluded most class-switched DN cells are memory cells that acquired mutations by a GC-derived mechanism.
• Transcriptomically, DN1 and SM cells differ by only 22 differentially expressed genes.
• Proposed mechanism for CD27 loss: immunosenescence — chronic antigen stimulation during normal aging, viral infections, and some autoimmune diseases, previously described as “exhausted memory B cells.”

(3) GC-independent (extrafollicular) origin — proposed for DN2/DN3

• The majority of DN and SM cells are clonally distinct, with differences in IgV(D)J family and gene usage — pointing to differential activation pathways or different stimulating antigens.
• DN2 cells lack CXCR5 and CD62L, both required for migration to lymphoid follicles.
• In SLE and HD, DN2 cells are related to activated naive B cells (IgD⁺CD27⁻CD21⁻CD24⁻T-bet⁺); the two subsets share high CD19 and CD11c, lack CD21/CXCR5/CD24/CD38, share a transcriptome and expanded clones, and both respond to TLR7 ligand + IFN-γ + IL-21 in vitro. A proportion of activated naive B cells differentiate into DN2 cells in vitro under that stimulation.

⚠ Bidirectional velocity — the origin question is not a settled arrow. Single-cell RNA velocity analyses are cited on both sides: a high flow from DN1 → SM (supporting DN1 as SM precursor) and a high flow from SM → DN1 (supporting DN as SM progeny), plus a flow from USM → DN2/DN3. The review’s own conclusion: “these hypotheses suggest that DN1-3 cells have different origins although conclusive evidence is still lacking… the origin of DN1 cells is less clear. Some DN1 cells could be precursors of SM B cells, while others could be progeny of SM B cells.”

Responsiveness — an explicitly unresolved debate

The review devotes §5 to a set of directly conflicting stimulation results, and closes it: “the activation potential and responsiveness of DN B cells towards BCR and TLR signaling are still a topic of debate.”

Stimulation	Result	Population
TLR9 (CpG) alone	No proliferation	Young and elderly HD (ref [14])
CpG or anti-BCR+anti-CD40	Significant proliferation	Young HD only — not elderly HD (ref [28])
CpG	Proliferation at a level similar to SM cells	HD (ref [19])
anti-BCR + CpG + IL-4	No proliferation	Young and aged HD (ref [42])
CpG + anti-BCR + anti-CD40 (triple)	Strong activation	Both young and elderly HD (ref [28])

• The review attributes the discrepancies to “the use of different proliferation markers, stimulation conditions, and HD cohorts.”
• Activation-marker deficit: the authors’ own data show a decreased percentage of activated CD86⁺ cells within DN B cells following triple stimulation (CpG+anti-BCR+CD40L) and CD40L stimulation, compared with total B cells, in both HD and MS patients. MS DN B cells, however, showed an increased ability to become activated after CD40L stimulation compared to HD DN cells — a disease-specific rescue of responsiveness.
• BCR signalling is intact: DN1-3 cells of HD, mild and severe SARS-CoV-2 patients, and post-SARS-CoV-2 immunised individuals all maintained BCR signalling after IgG stimulation. DN2 cells of HD showed the highest expression of activation markers (CD69, CD86) and the largest BCR signalling capacity of the DN subsets.
• TLR7 hyper-responsiveness: TLR7 stimulation further increased CD25, HLA-DR and CD86 on DN2 cells, indicating hyper-responsiveness to TLR7 specifically.
• FcRH4 is compartment-restricted: DN cells expressing the inhibitory receptor FcRH4 predominate in tonsil, while peripheral blood DN cells are FcRH4⁻ in both HD and SLE. The authors speculate this is why circulating DN cells are more responsive to activation and expansion than their tissue counterparts.

Function

• DN2 as ASC precursors (established): transcriptomic profile includes IRF4 (essential for ASC differentiation) and lacks Ets-1 and BACH2 (which generally prevent plasma cell differentiation). BCR sequencing shows shared clones between DN2 and plasma cells. TLR7 + BCR ligation + IFN-γ + IL-21 drives plasma cell differentiation in vitro. CD11c⁺ B cells including DN2 also differentiate to ASCs after CpG2006 + Staphylococcus aureus Cowan I + IL-21.
• ⚠ DN1 can also make ASCs. “GC-dependent DN1 cells were also able to differentiate into ASCs in vitro following stimulation with TLR7, BCR ligation, IFN-γ and IL-21. It is not yet clear whether this ASC differentiation by DN1 cells is dependent on TLR7 or BCR signaling.” This qualifies the clean DN1-memory / DN2-effector split.
• Antigen presentation: HLA-DR, CD80 and CD86 expression on DN cells of HD and MS patients was intermediate between naive and SM cells (authors’ own data). In SLE, DN2 cells expressed higher HLA-DR, CD69 and CD86 than SM cells. No direct evidence of DN-driven T cell induction exists — the case is phenotypic only.
• Cytokines — conflicting: DN B cells could not be induced to express IL-10 or TNF-α after total B cell stimulation with anti-CD40+IL-4, or CpG/PMA/ionomycin (ref [17]). In contrast, the authors’ own study found CD40L stimulation of total B cells from HD and MS patients produced lymphotoxin-α and TNF-α from DN cells, with similar LTα⁺ frequencies and higher TNF-α⁺ frequencies in DN than SM cells.
• Granzyme B: DN B cells of young and elderly HD and MS patients produced the cytotoxic molecule granzyme B after in vitro IL-21 + anti-BCR stimulation.
• Migration: DN cells express CXCR3 (young HD) and CCR6 (elderly HD); HIV-infected DN cells express both. Increased CXCR3⁺ DN frequencies in SLE and axSpA. DN cells found in inflamed RA synovial tissue and MS cerebrospinal fluid.

Aging

• DN B cell frequencies are increased in peripheral blood of aged HD (>60 y) vs. young HD (<60 y), and positively correlate with age and with levels of cytotoxic, age-associated CD4⁺CD28⁻ T cells.
• In normal aging, DN cells are mostly described as senescent or exhausted memory B cells. Their role in reduced vaccination efficacy in aged individuals “has not been properly addressed.”

Disease associations (Table 1 highlights)

• SLE: first reported in 2002. Unlike normal aging, DN frequency does not correlate with age in SLE — pointing to premature elevation. DN1 predominates in HD; SLE shows predominant DN2 expansion. Clinical correlates: nephritis, disease flares, severity of renal damage, SLEDAI. DN2 and activated CD95⁺CD27⁻ DN frequencies correlate significantly with SLEDAI. Positive correlations with autoantibodies including anti-Smith, anti-ribonucleoprotein, and anti-9G4⁺ B cells.
• COVID-19: severe (ICU) patients show a shift toward increased DN2 and DN3 and reduced DN1. DN3 levels in severe COVID-19 positively correlate with increased titres of autoreactive antibodies and with clinical/laboratory parameters — ventilatory parameters, leukocytes, neutrophils, CRP, ferritin, D-dimers. The DN2/DN3 increase is transient: frequencies in recovered patients were similar to those who never had severe disease.
• Malaria: CD19⁺IgD⁻CD27⁻CD21⁻ atypical memory B cells increased in people living in malaria-endemic areas and children with chronic malaria exposure; these cells lack CD21, “suggesting overlap with DN2 cells.”
• HIV: tissue-like memory B cells (CD20^hi CD27⁻CD21^lo) resembling DN cells are increased vs. HIV-aviremic and HIV-negative controls. DN frequencies correlate with terminal effector CD4⁺ T cells (CD3⁺CD57⁺CD45RO⁻) that predominate in aging. An expansion of CD21⁻T-bet⁺ B cells resembling DN2 was also observed. DN levels show a negative correlation with immune response after seasonal influenza vaccination.
• MS (the authors’ own disease focus): elevated DN in blood and further increased in CSF; positive correlation with cytotoxic CD4⁺CD28⁻ T cells and absence of correlation with age → premature immune aging. Clonal relationship between class-switched DN cells and intrathecal Ig repertoires suggests periphery↔CNS migration. Only a minority of MS DN cells could be retraced to CD21⁻CD11c⁺ (DN2-like) cells — so the role of DN in MS may differ mechanistically from SLE.
• IBD is the sole inverse case: DN cells are decreased in circulation but enriched in gut-associated lymphoid tissue — interpreted as recruitment from blood to tissue rather than absence.
• Other: pSS (CD27⁻CD21^-/low with autoreactive BCR propensity; not age-correlated), SSc (FcγRIIB expression correlates with severity), RA (elevated in blood, further in synovial fluid; CD11c⁺ DN2 associated with ACPA; no correlation with disease activity), JIA (elevated only in synovial fluid, especially in ANA⁺ patients), axSpA (CD27⁻CD38^low CD21^low elevated; positively correlated with age, ESR, extra-skeletal manifestations), NMOSD (CD20^low DN elevated; CSF AQP4-specific B cells clonally related to circulating DN), Graves’ (correlates with thyroid autoantibody levels), obesity (elevated DN2 in young obese individuals; autoantibody production), ALS (DN predicts shorter survival), NSCLC (increased in lung tumour tissue, correlating with degree of tumour differentiation).

Methods Used

• Conventional Flow Cytometry — the dominant technology in Table 1; the review’s aggregation exposes incompatible DN definitions across studies
• In Vitro B Cell Stimulation — the basis of the responsiveness debate (§5) and the ASC-differentiation evidence (§6.1)
• BCR Sequencing — AIRR/Ig VH repertoire sequencing underpins all three origin hypotheses
• Single-Cell RNA Sequencing — RNA velocity analyses cited on both sides of the DN↔SM origin question
• Phospho-Flow Cytometry — BCR signalling capacity across DN1-3

Entities Mentioned

Double-Negative B Cell, DN2 B Cell, DN3 B Cell, Atypical B Cell, Activated Naive B Cell, Switched Memory B Cell, CD27, TLR7, TLR9, FcRH4

Also discussed but not separately updated from this review (content relayed from primaries already ingested): Age-Associated B Cell, Plasmablast, CD19, CD21, CD11c, CXCR5, CXCR3, CD38, IgD, IgG, T-bet, IRF4, BACH2, IL-21, IFN-gamma, TNF-alpha, CD40L, FCRL5.

Concepts Addressed

Extrafollicular Response, Germinal Center, Atypical B Cell Effector Output

Also touched, no new content beyond ingested primaries: Class Switch Recombination, Somatic Hypermutation, Memory B Cell, Follicular Exclusion, Toll-like Receptor Signaling in B Cells, B Cell Receptor Signaling.

Relevance & Notes

What this adds that the wiki did not have. Three things. First, the memory-credential evidence for DN cells beyond SHM — ABCB1 negativity, short telomeres, and USM/SM-like FSC/SSC morphology. The wiki carried the Rhodamine 123 result from Wei2007 - DN Memory B Cells in SLE but not its molecular explanation or the replicative-history data. Second, the explicit responsiveness debate — four studies giving four different answers to “do DN cells proliferate to CpG?” The wiki’s Double-Negative B Cell page has been carrying Wei2007’s positive result as settled. Third, Szelinski 2022’s DN^low/int/hi scheme, which the wiki had not encountered; DN^low matters disproportionately because it is CD11c⁻ yet plasmablast-transcriptomic — a putative effector subset that a CD11c-gated DN2 panel would miss entirely.

What it corroborates without adding. The DN1/DN2 definitions (Jenks2018 - DN2 B Cells and EF Pathway in SLE, ref [32]), the DN3 subset and its severe-COVID expansion (Woodruff2020 - EF B Cell Responses in COVID-19, ref [34]), and the aNAV→DN2 EF axis are all relayed from primaries already ingested here. Where this review restates them, the wiki cites the primary.

What it qualifies. The wiki has been operating a fairly clean DN1 = GC-derived memory precursor / DN2 = EF effector dichotomy. Beckers qualifies this from two directions: DN1 cells differentiate to ASCs in vitro under the same TLR7/BCR/IFN-γ/IL-21 cocktail that drives DN2, and RNA velocity supports flow in both directions between DN1 and SM. The dichotomy remains the best available organising frame, but it should be held as a working model, not a finding.

The exhaustion/effector split now has a shape. The inhibitory-receptor data (FcRH3-5, CD22, CD85j present in HIV, malaria, vaccinees, and HD; absent in SLE and Hashimoto’s) give a phenotypic axis for what the wiki has been treating as a semantic problem — “exhausted atypical MBC” (malaria/HIV literature) vs. “EF effector DN2” (SLE/COVID literature). Beckers’ framing: exhaustion in aging and chronic viral infection, activation in autoimmunity. This bears directly on how Singh2026 - DENV-Specific Memory B Cell Subsets’s DENV-specific CD27⁻CD21⁻ cells should be read — Singh2026’s temporal correlation data argue against exhaustion, which would place dengue on the autoimmunity side of Beckers’ split rather than the chronic-infection side.

Secondary-source status. This is a narrative review with no stated search strategy. Claims relayed from other groups should be traced to their primaries before being weight-bearing. The exception is the Fraussen/Somers group’s own data — refs [18] (Fraussen 2019, J Immunol) and [43] (Claes 2016, J Immunol) — which is reported here at primary-source level of detail and is cited on wiki pages as review, own data. That set is:

• the 0.2–2.2% DN–SM clonal overlap and the accompanying AIRR/phenotype work (CD10, CD38, CD5, CD95 comparisons; IgG⁺ composition, reported as “we and others”);
• the MS findings — DN elevated in blood and CSF, correlation with cytotoxic CD4⁺CD28⁻ T cells, absence of an age correlation, and only a minority of MS DN cells retraceable to CD21⁻CD11c⁺ (DN2-like) cells;
• the CD86 activation deficit after triple or CD40L stimulation, and the increased CD40L responsiveness of MS DN cells;
• LTα and TNF-α production, granzyme B production (jointly with Bulati 2014), and the intermediate HLA-DR/CD80/CD86 expression.

⚠ Not own data, despite an early draft of this page saying so: ABCB1-negativity (Colonna-Romano 2009 / Wei 2007 / Wirths 2005), telomere length (Colonna-Romano 2009), and FSC/SSC morphology (Wu 2011) are all relayed from other groups — principally the Palermo group (Colonna-Romano/Bulati/Martorana) and Dunn-Walters’ group.

Limitations.

• No systematic search; selection of the ~70 studies in Table 1 is not documented.
• Table 1 aggregates studies using mutually incompatible DN definitions — CD19⁺ vs. CD20⁺ vs. CD20^low vs. CD19^low/int/hi, with or without CD21, CD38, CD10, CD11c, CXCR5, T-bet. Frequencies are therefore not comparable across rows, a caveat the review does not state.
• “Increased in PB” is recorded without effect sizes, n, or p-values anywhere in Table 1.
• The review is Europe/North America-facing in its infection coverage. Malaria and HIV appear; the arboviruses do not.

Reference Map (verified 2026-08-26)

The wiki cites this review’s relayed claims by author-year. Those names were checked against the printed reference list (pp. 75–76) after an initial pass got several wrong. Use this map rather than re-deriving it.

Ref	Source	What the review uses it for
[13]	Huang 2002, Arthritis Rheum	first report of elevated DN in SLE (2002)
[14]	Colonna-Romano 2009, Mech Ageing Dev	DN increased in elderly; telomere length; low Bcl2; CpG study #1
[15]	Agematsu 2000, Immunol Today	CD27 as TNFRSF member, ligand CD70
[16]	Huang 2020, Adv Exp Med Biol	IgD downregulated after isotype switching
[17]	Bulati 2011, Biogerontology	no IL-10/TNF-α induction from DN cells
[18]	Fraussen 2019, J Immunol	OWN DATA — AIRR, 0.2–2.2% DN–SM clonal overlap, DN phenotype
[19]	Wei 2007, J Immunol	ingested as Wei2007 - DN Memory B Cells in SLE; CpG study #3; FcRH4⁻ in blood
[20]	Wirths 2005, Eur J Immunol	ABCB1 discriminates naive from memory
[21]	Wu 2011, Front Immunol	isotype-specific VH mutation; CDR3 physicochemistry; bidirectional DN↔SM genealogical trees
[22]	Ruschil 2020, Front Immunol	DN induced by influenza / TBE vaccination
[23]	Frasca 2017, Exp Gerontol	SASP markers (TNF-α, IL-6, IL-8, p16^INK4, miRs)
[24]–[26]	Moir 2008 / Weiss 2009 / Portugal 2015	HIV and malaria atypical/exhausted memory
[27]	Castleman 2022, Front Immunol	BCR signalling intact in DN1-3; DN3 ↔ autoreactive antibodies
[28]	Martorana 2014, Immunol Lett	CpG study #2 (young but not elderly; triple stimulation)
[29]	Liu 2017, Mol Cell Endocrinol	CD32b reduced on DN in Hashimoto’s
[30]	Stewart 2021, Front Immunol	all three scRNA-velocity claims (DN1→SM, SM→DN1, USM→DN2/3)
[31]	Cancro 2020, Annu Rev Immunol	ingested as Cancro2020 - Age-Associated B Cells
[32]	Jenks 2018, Immunity	ingested as Jenks2018 - DN2 B Cells and EF Pathway in SLE
[34]	Woodruff 2020, Nat Immunol	ingested as Woodruff2020 - EF B Cell Responses in COVID-19
[35]	Reyes 2021, PLOS One	transience of the COVID DN2/DN3 shift (NOT Stewart)
[36]	Sosa-Hernández 2020, Front Immunol	DN3 ↔ CRP, ferritin, D-dimers, ventilatory parameters
[37]	Szelinski 2022, Arthritis Rheumatol	DN^low / DN^int / DN^hi — title reads “antigen-experienced CXCR5⁻CD19low B cells”
[38]	Berkowska 2011, Blood	cell divisions and Ig mutations in IgG⁺CD27⁻ DN ≈ GC B cells
[42]	Frasca 2019, PLOS One	CpG study #4 (no proliferation to anti-BCR+CpG+IL-4); obesity
[43]	Claes 2016, J Immunol	OWN DATA — MS DN/ABC, CD86 deficit, LTα/TNF-α, granzyme B, CSF
[44]	Ehrhardt 2003, PNAS	FcRH4 inhibitory potential (not the 2005 J Exp Med paper)
[48]	Golinski 2020, Front Immunol	CD11c⁺ B cells → ASC with CpG2006 / SAC / IL-21
[51]	Bulati 2014, Exp Gerontol	granzyme B; CXCR3/CCR6 trafficking phenotype
[52]	Jacobi 2008, Arthritis Rheum	activated CD95⁺CD27⁻ DN ↔ SLE disease activity
[53]	Wilbrink 2021, Front Immunol	axSpA CD27⁻CD38^low CD21^low; pSS age non-correlation
[70]	Palanichamy 2014, Sci Transl Med	MS intrathecal Ig repertoire ↔ class-switched DN clonality
[71]	Kowarik 2017, Ann Clin Transl Neurol	NMOSD CSF AQP4-specific B cells ↔ circulating DN
[74]	Carrasco 2019, Front Immunol	IBD — DN reduced in blood, enriched in gut tissue

⚠ One internal inconsistency in the review itself. §3 prints DN^low as CXCR5⁺, while Table 1 of the same review prints it CXCR5⁻ — and Szelinski’s own title says CXCR5⁻. The wiki treats the body text as a typo and reads DN^low as CXCR5⁻, which places it in DN3 territory rather than making it a fourth population. A second inconsistency is recorded on FcRH4: §2 lists FcRH3-5 among inhibitory receptors expressed by DN cells in healthy donors, while §5 states blood DN cells are FcRH4⁻.

Questions Raised

• Is the DN2/DN3 expansion in acute infection always transient? The COVID-19 data show DN2/DN3 normalising after recovery. If dengue behaves the same way, the sampling window in a dengue cohort determines whether the expansion is seen at all. What is the decay constant, and does it differ between primary and secondary infection?
• Would a CD11c-gated DN2 panel miss DN^low? Szelinski’s DN^low is CD11c⁻ but plasmablast-transcriptomic. If a comparable subset exists in dengue, it would be assigned to DN1 or DN3 by the standard CXCR5/CD21/CD11c scheme and its effector character missed.
• Does the exhaustion/activation split by disease class hold in dengue? Dengue is an acute infection (like COVID-19, where DN cells are activated) rather than chronic (like HIV/malaria, where DN cells are exhausted). Do DENV-specific DN cells express FcRH3-5/CD22/CD85j?
• Is DN1’s ASC potential TLR7-dependent or BCR-dependent? The review states this is unresolved. It matters because it determines whether DN1 is a latent effector released by innate signalling or requires cognate antigen.
• What drives CD27 downregulation, mechanistically? “Immunosenescence” and “chronic antigen stimulation” are labels, not mechanisms. No transcriptional or epigenetic pathway is offered.
• Does the DN3 ↔ autoreactive-antibody correlation in severe COVID-19 reflect causation or shared drivers? DN3 also correlates with CRP, ferritin, D-dimers — i.e. with systemic inflammation generally.
• Why is IBD the only condition where DN cells fall in circulation? If tissue recruitment can invert the blood signal, blood-only DN measurements in any tissue-tropic disease may be misleading.