henhaobio
Sign in

bioRxivImmunologyPreprint

A high-resolution blood immune cell doublet atlas via imaging spectral cytometry

Cheryl KimiD, Denise Hinz, Aaron J. MiddlebrookiD, Aaron J. TyznikiD, Louise M. D’CruziD, Thomas J. ScribaiD, Bjoern PetersiD, Julie G. BureliD

PDFDOI
In-depth full-text digest available · Jump to digest ↓

Abstract

Circulating cell-cell complexes (doublets) provide critical insights into in vivo immune interactions. Here, we leverage high-dimensional imaging spectral cytometry to systematically map the landscape of immune doublets in human peripheral blood. By integrating morphometric imaging features with a 22-color spectral panel, we established a gating workflow that cleanly segregates genuine, physically interacting complexes from coincidental transits and platelet contaminants, and resolved 12 highly pure homotypic and heterotypic immune doublet configurations in peripheral blood mononuclear cells (PBMC). At steady state, each doublet population displayed unique abundance, affinity and surface phenotypic profiles relative to circulating singlets. Principal component analysis (PCA) of quantitative imaging features further revealed lineage-specific topologies distinguishing singlet cells, homotypic doublets, and heterotypic pairs. Finally, comparative cohort analysis demonstrated that while doublet frequencies, affinities and surface phenotypes remain comparable between healthy donors and patients with tuberculosis (TB) disease, imaging-derived features uncovered disease-specific alterations in doublets morphology. Specifically, high-dimensional clustering using imaging parameters showed that TB disease significantly enriches for a tightly synapsed T cell-monocyte cluster characterized by high CD3-CD14 pixel correlation. Collectively, this study establishes a robust framework for mapping circulating cell-cell networks and demonstrates the superior sensitivity of imaging features over conventional fluorescence for capturing functional immune interactions in health and disease.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 1 (8 Sept 2026) · CC BY 4.0

Image-based gating on an imaging spectral cytometer resolves 12 high-purity immune doublet types in human blood; in tuberculosis, doublet frequencies are unchanged but tightly contacting T cell-monocyte doublets increase.

Flow cytometry usually discards doublets as artifacts, yet some circulating doublets are genuine interacting cell pairs, and earlier work has looked mostly at T cell-monocyte pairs. This work uses real-time imaging spectral cytometry, which pairs a large spectral panel with cell images captured during acquisition, to catalogue many more doublet configurations and to test whether image features add information beyond fluorescence intensity.

Key results

  1. ShownA marker-independent gate on side-scatter and LightLoss image shape and size, followed by lineage-marker gating, resolves 12 homotypic and heterotypic doublet types in peripheral blood mononuclear cells (PBMC). Six further gates (13 to 18) were heavily contaminated with leukocyte-platelet events and were discarded.

    Pairs are named by partner (T, B, NK; M = CD14+ monocyte; M14- = CD14-negative myeloid). Gates 13 to 18 had median CD42b+ (platelet) frequencies of 45.2% to 85.2%, vs 5.6% to 22.6% in populations 1 to 12. In 5 healthy donors the 12 types made up about two-thirds of the morphometric doublet gate (median range 59.7% to 77.2%). Manual image review gave median purities of 80% to 96%. · Fluorescence- and image-guided segmentation…; Steady-state profiling reveals distinct…; Figure legends

  2. SuggestedIn healthy donors, doublets containing T cells are the most common. After normalising for singlet abundance, heterotypic pairs of CD14-negative myeloid cells with B or NK cells have the highest association constants (Ka); T:T and M:M have the lowest.

    Share of live CD45+ events: T:T 1.03%, T:M 0.75%, T:B 0.25%, T:NK 0.21%, M:M 0.02% (n=5). Ka = doublet frequency over the product of the two singlet frequencies. Plotting frequency rank against Ka rank put M:B, T:B and T:M14- in the frequent, high-affinity group, and NK:M14-, B:M14- and B:B in the rare, high-affinity group. · Steady-state profiling reveals distinct…; Figure legends

  3. SuggestedSome heterotypic doublets differ from their matching singlets in sub-lineage make-up. The authors read this as certain subsets preferentially forming contacts in circulation.

    Compared with singlets (Mann-Whitney, healthy n=5): fewer TCRγδ+ T cells in T:M14-, fewer CD20+ B cells in B:M and B:M14-, lower CD16 positivity in T:NK, and more CD16+ intermediate monocytes in M:B. CD4, CD8 and CD27 in T cell pairs, and CD27, CD123, IgM and IgD in B cell pairs, were unchanged. · Doublet combinations display distinct surface…; Figure legends

  4. SuggestedBetween healthy donors and patients with active tuberculosis (TB), doublet frequencies and Ka did not differ, and surface changes mostly tracked those in singlets. Single-channel image features did shift in TB, some only in doublets.

    n=5 per group. In TB, CD3 diffusivity and short-axis moment rose in all T cell doublet types and singlets, and CD3 maximum intensity was broadly reduced. Doublet-only shifts included higher CD3 long-axis moment in T:M and T:M14-, higher CD3 radial moment in T:M and T:T, and higher CD14 radial moment in M:B, M:NK and M:M. CD19 features changed least. · Imaging parameters capture disease-specific…; Figure legends

  5. SuggestedUnsupervised clustering of CD3 and CD14 image features in T cell-monocyte doublets isolates a small tight-contact cluster (high CD3-CD14 pixel correlation, minimal centre-of-mass distance) that is more frequent in TB.

    FlowSOM/UMAP on the combined cohorts gave four T:M clusters. Pop 3 (tight contact) was significantly enriched in TB vs healthy donors (n=5 vs n=5, Mann-Whitney p<0.01). A similar tight-contact cluster among T:B doublets did not differ between cohorts. M:B doublets were not clustered because CD14 signal spilled into the CD19 imaging channel. · Unsupervised clustering of imaging parameters…; Flow cytometry data analysis; Figure legends

Methods, data and reproducibility

PBMC stained with a 22-color panel (CD3, CD14, CD19 imaged) were run on a BD FACSDiscover S8; doublets were gated by image morphometrics, then lineage markers. Doublet types were compared by frequency, association constant (Ka), surface markers and PCA of image features in healthy donors, then healthy and TB donors were compared by these measures, per-feature image medians and FlowSOM/UMAP clustering.

  • Cryopreserved PBMC from healthy controls (no TB history or known exposure; negative interferon-gamma release assay) · human · n = 5
  • Cryopreserved PBMC from patients with active pulmonary TB confirmed by Mycobacterium tuberculosis molecular or culture testing; enrolled at the SATVI field site, South Africa · human · n = 5
  • Healthy PBMC used for antibody titration, dye benchmarking and imaging-channel spillover controls · human · n = 1
Methods & tools
BD FACSDiscover S8 Cell Sorter (SpectralFX, CellView Image Technology), BD FACSChorus v5.4, BD Research Cloud (Complexity Score), FlowJo v10.10, BD CellView Lens plugin v2.1, FlowSOM plugin v3.0, UMAP plugin v3.1, R 4.5 (prcomp), GraphPad Prism v10.4, SpectraComp cell mimic controls, Human BD Fc Block, Ficoll-Paque PLUS
Reagents & environment
BD FACSChorus 5.4; FlowJo 10.10 with BD CellView Lens 2.1, FlowSOM 3.0 and UMAP 3.1 plugins; GraphPad Prism 10.4; R 4.5 (prcomp on centered, scaled features).
Reproducibility gaps
No code or data availability statement. No repository for FCS or image files is named. The panel (Table 1) is cited, but its contents are not reproduced in the text provided.
Limitations (6)
  • FACSChorus v5.4 had no compensation or unmixing for imaging detectors. CD14 RB613 spilled into the CD19 imaging channel (ImgB3), so CD14/CD19 cross-channel metrics were dropped and M:B doublets were left out of clustering. Newer software unmixes only within the imaging channels. · stated by the authors
  • Non-imaging markers cannot be assigned to one partner in homotypic doublets, so homotypic pairs were not sub-phenotyped. Attribution in heterotypic pairs needs lineage-exclusive markers (e.g. CD11c rather than HLA-DR). · stated by the authors
  • IgG staining appeared on myeloid and NK cells despite Fc-receptor blocking. The authors attribute this to incomplete blocking of high-affinity IgG receptors. · stated by the authors
  • Whether Ka values reflect physical bond strength is untested. The authors propose sorting doublets and re-acquiring them under increasing shear. · stated by the authors
  • Cohorts are small (5 healthy, 5 TB), and many Mann-Whitney comparisons are judged at p<0.05 with no multiple-testing correction described. · noted in this summary
  • All measurements are on thawed cryopreserved PBMC. The text reports no control for complexes formed or lost during freezing, thawing or staining. · noted in this summary

Read this if you study immune cell-cell interactions in blood, or want a validated gating scheme for physically interacting doublets on an imaging spectral cytometer such as the BD FACSDiscover S8.

Citations
1

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISBurel '19Burel '20Kang '26Kang '25Ionita '25Rashid '19
76% semantic matchImmunology
J. G. Burel, M. Pomaznoy, C. S. Lindestam Arlehamn+21 more·bioRxiv·Feb 21, 2019