henhaobio
Sign in

Nature MethodsSystems BiologyPeer-reviewed journal article

Integration of Imaging-based and Sequencing-based Spatial Omics Mapping on the Same Tissue Section via DBiTplus

Archibald EnninfuliD, Zhaojun ZhangiD, Dmytro Klymyshyn, Matthew H. IngallsiD, Mingyu YangiD, Hailing Zong, Zhiliang BaiiD, Negin FarzadiD, Graham SuiD, Alev BaysoyiD, Jungmin NamiD, Yao LuiD, Shuozhen BaoiD, Siyan Deng, Nancy Ruonan ZhangiD, Oliver BraubachiD, Mina Luqing XuiD, Zongming MaiD, Rong FaniD

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

Abstract

Spatially mapping the transcriptome and proteome in the same tissue section can significantly advance our understanding of heterogeneous cellular processes and connect cell type to function. Here, we present Deterministic Barcoding in Tissue sequencing plus (DBiTplus), an integrative multi-modality spatial omics approach that combines sequencing-based spatial transcriptomics and image-based spatial protein profiling on the same tissue section to enable both single-cell resolution cell typing and genome-scale interrogation of biological pathways. DBiTplus begins with in situ reverse transcription for cDNA synthesis, microfluidic delivery of DNA oligos for spatial barcoding, retrieval of barcoded cDNA using RNaseH, an enzyme that selectively degrades RNA in an RNA-DNA hybrid, preserving the intact tissue section for high-plex protein imaging with CODEX. We developed computational pipelines to register data from two distinct modalities. Performing both DBiT-seq and CODEX on the same tissue slide enables accurate cell typing in each spatial transcriptome spot and subsequently image-guided decomposition to generate single-cell resolved spatial transcriptome atlases. DBiTplus was applied to mouse embryos with limited protein markers but still demonstrated excellent integration for single-cell transcriptome decomposition, to normal human lymph nodes with high-plex protein profiling to yield a single-cell spatial transcriptome map, and to human lymphoma FFPE tissue to explore the mechanisms of lymphomagenesis and progression. DBiTplusCODEX is a unified workflow including integrative experimental procedure and computational innovation for spatially resolved single-cell atlasing and exploration of biological pathways cell-by-cell at genome-scale.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 1 (11 Nov 2024) · CC BY 4.0

Thermostable RNaseH release of barcoded cDNA keeps sections intact for CODEX imaging, yielding same-section whole-transcriptome and protein maps plus CODEX-guided spot deconvolution in FFPE tissues.

Spatial multimodal data are usually integrated across adjacent sections whose cell composition never fully matches, while imaging methods cover limited gene panels and sequencing-based spatial methods have coarser resolution. DBiTplus runs microfluidic barcoding (DBiT-seq) and ~50-marker CODEX on one section, with registration and CODEX-guided deconvolution that assigns cell types within each transcriptome spot.

Key results

  1. ShownOn an FFPE E11 mouse embryo, transcriptomes recovered by RNaseH matched a standard DBiT-seq run on the adjacent section. The intact section then supported 26-marker CODEX imaging, and its anatomy agreed with the transcriptomic clusters.

    Pearson R=0.99 (p < 2.2e-16) vs standard DBiT-seq on the adjacent section, with 27,884 shared genes; about 1,200 genes and 3,300 UMIs per spot; 10 clusters matched anatomy (liver: Afp, Hba.a1; heart: Myh6, Tnnt2). Mki67, Nefl and Sox2 RNA patterns matched CODEX protein. · Multimodal mapping of mouse embryo; Figure legends

  2. SuggestedThree ways of releasing cDNA after barcoding were tried. Only RNaseH digestion left mouse spleen sections usable for CODEX; NaOH also disrupted the morphology of embryo sections.

    Three serial spleen sections (0.1M NaOH for 5 minutes, 90% DMSO for 30 minutes, 10U RNaseH at 37°C for 30 minutes) were stained with a 25-marker mouse immune panel. The RNaseH section gave 14 positive markers, including CD45R and CD169; the NaOH and DMSO sections stained poorly. · Workflow, development, and optimization in fresh…; Figure legends

  3. ShownIn OCT-frozen mouse embryo sections, enzymatic release captured fewer genes than standard tissue lysis. Even so, replicates correlated well, and spatial clusters and marker genes matched reference atlases (Allen ISH, Stereo-seq).

    20,089 vs 23,295 genes detected (RNaseH vs proteinase K lysis). With 0.5% Triton X-100 added, mean genes per spot were 1,029 vs 2,312. Test vs control correlations were R=0.87 (first replicate) and R=0.72 (second replicate), with R=0.84 between the two test samples; 1109 genes overlapped between matched clusters 0 (control) and 2 (test). · Workflow, development, and optimization in fresh…; Figure legends

  4. ShownIn human lymph node, running DBiTplus first preserved CODEX signal for cell-type-specific markers, compared with a CODEX-only section two sections away. Markers expressed across many cell types agreed less well.

    35-plex CODEX after DBiTplus vs a CODEX-only section, compared over ∼4000 binned regions: Pearson > 0.85 for CD3ɛ, CD20 and Podoplanin, lower for CD45 and Beta-actin. Five transcriptomic clusters (50 μm device) lined up with CODEX zones, e.g., MS4A1/CXCL13 with CD20+ follicles. · Multimodal mapping of normal human lymph nodes; Assessing quality of CODEX data after DBiT-seq; Figure legends

  5. SuggestedUsing the CODEX cell composition from the same section as the deconvolution prior, then splitting spots into pure-cell-type sub-spots and integrating with WNN, separated cell types better than deconvolution without CODEX or with TACCO.

    Lymph node: ASW 0.45 and ARI 0.15 with the matched CODEX prior, reported as higher than without CODEX (comparator values not given). Mouse embryo: epithelial cells separated after WNN but were lost with TACCO (qualitative only). CODEX WNN weight was about 0.55 for germinal center and plasma B cells. · Development of Computational framework for…; Multimodal mapping of mouse embryo; Multimodal mapping of normal human lymph nodes

Methods, data and reproducibility

After in situ reverse transcription, spatial barcodes are ligated using two perpendicular 50-channel microfluidic chips. Thermostable RNaseH then releases the cDNA, leaving the section for CODEX and H&E. CODEX cells are segmented, annotated by MaxFuse plus an SVM, and registered to spots, which are split into pure-cell-type sub-spots for Seurat WNN integration.

  • OCT-frozen C57 E13 mouse embryo sagittal sections (Zyagen MF-104-13-C57) used to optimize RNaseH retrieval against a proteinase K lysis control · mouse (C57)
  • OCT-frozen C57 mouse spleen serial sections comparing NaOH, DMSO and RNaseH cDNA retrieval, followed by 25-marker CODEX · mouse (C57) · n = three serial sections
  • FFPE C57 E11 mouse embryo sagittal section (Zyagen MP-104-11-C57) profiled by DBiTplus with 26-marker CODEX, plus an adjacent standard DBiT-seq section · mouse (C57)
  • FFPE human cerebellum section (Zyagen HP-202) imaged with a 29-marker CODEX panel after DBiTplus · human
  • Archived FFPE reactive human lymph node from Yale Pathology Tissue Services, barcoded at 50 μm and 25 μm pixel sizes, with 35-plex CODEX and H&E · human
  • FFPE diffuse large B cell lymphoma (DLBCL) transformed from marginal zone lymphoma; 2.5mm by 2.5mm region at 25 μm with 44-marker CODEX · human
  • Mouse Organogenesis Cell Atlas (MOCA) scRNA-seq reference, restricted to cells at 10.5 to 11.5 days post-coitum · mouse
  • Cell2location integrated lymphoid organ scRNA-seq reference with the original cell type annotations, used for MaxFuse label transfer · human
Methods & tools
DBiT-seq / Patho-DBiT microfluidic spatial barcoding, CODEX on Akoya PhenoCycler-Fusion, Thermostable RNaseH (NEB M0523S), STAR v2.7.8a, ST_Pipeline v1.8.1, Seurat v4.3.0 (SCTransform, WNN), Mesmer / DeepCell whole-cell segmentation, MaxFuse, Support vector machine classifier, Scanpy, scikit-image and SciPy (image registration), TACCO (deconvolution comparator)
Reagents & environment
STAR V2.7.8a against GRCm39 or GRCh38, ST_Pipeline V1.8.1, Seurat V4.3.0, QuPath V0.5.1; Python scikit-image and SciPy for registration; Illumina NovaSeq 6000 sequencing with paired-end 150bp reads.
Reproducibility gaps
The text gives no code link and no deposition accession for the new DBiTplus sequencing or CODEX data; only the public scRNA-seq reference links are provided.
Limitations (6)
  • Capture depth is shallow, and RNaseH release appears less efficient than tissue lysis, so low-abundance transcripts and rare cell types may be missed. RNaseH concentration and incubation times still need optimization. · stated by the authors
  • The demonstrated spot sizes of 20 μm to 50 μm fall short of single-cell resolution. Single-cell assignment depends on CODEX-guided spot splitting rather than on native transcript capture. · stated by the authors
  • H&E staining done after DBiTplus and CODEX showed lower nuclear detail than an untreated section, although trained pathologists could still identify general cell types from size, shape and tissue location. · stated by the authors
  • The lymphoma application rests on a single DLBCL sample transformed from marginal zone lymphoma. The authors say the tumor microenvironment needs further analysis in a patient cohort. · stated by the authors
  • Sample sizes are small: the FFPE embryo, cerebellum and lymphoma each rest on one DBiTplus section and the lymph node on adjacent sections. The TACCO comparison is qualitative, ASW/ARI values without CODEX are not reported, and CARD, RCTD or Cell2location are not benchmarked quantitatively. · noted in this summary
  • The main text and figure legends disagree on embryo stage (E11 vs E13), panel sizes (26 vs 21 markers; 29 vs 28 markers) and lymph node device pitch (20 vs 25 μm), which makes it hard to match results to samples. · noted in this summary

Read this if you want transcriptome-wide sequencing and multiplexed protein imaging on one FFPE section, or need a protocol for releasing cDNA without destroying tissue plus CODEX-guided spot deconvolution.

Citations
34
FWCI
55.80
Field percentile
top 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
THISLiu '20Barnett '22Duchini '26Venkatraman '26Battistoni '24Wirth '23
71% semantic matchGenomics
Yang Liu, Mingyu Yang, Yanxiang Deng+13 more·Cell·Nov 13, 2020· 966 citations