PASTA: versatile tyramide-oligonucleotide amplification for multimodal spatial biology

Abstract

Spatial proteomics is limited by detection sensitivity, multiplexing and multimodal integration, leaving a gap between discovery and clinical assays. Here we present protein and nucleic acid serial tyramide amplification (PASTA), using horseradish peroxidase-mediated oligonucleotide deposition and cyclical imaging for high-plex, multimodal spatial profiling. Compatible with conjugated antibodies and in situ hybridization, PASTA enables simultaneous protein and RNA codetection from formalin-fixed, paraffin-embedded samples, providing a cost-effective bridge from discovery to clinical validation.

Spatial profiling technologies have revolutionized our understanding of tissue architecture by enabling visualization of biomolecules within their native microenvironment^1,2. Despite recent advances, several technical challenges continue to limit comprehensive application: poor signal-to-noise ratios for low-abundance targets, bioconjugation complications restricting reagent availability, and the complexity of integrating multiple detection modalities.

Current spatial proteomics methods routinely image >20 targets on the same tissue using cyclical imaging^3,4 or lanthanide mass tags^5,6, largely overcoming prior multiplexing limitations. However, sensitivity limitations persist, particularly for low-abundance markers critical to understanding immune responses and disease mechanisms. Oligonucleotide-based amplification approaches have shown considerable promise due to their modular nature for detecting proteins and nucleic acids. Signal amplification by exchange reaction provides strong amplification but is limited in multimodal integration to conventional immunofluorescence^7–10. Hybridization chain reaction enables multimodal detection of proteins, protein–protein interactions and nucleic acids, but is currently restricted to 10-plex^11–13. Rolling circle amplification and amplification by cyclic extension achieve high-plex in situ amplification but similarly lack robust multimodal integration^14,15.

Tyramide signal amplification (TSA) has been widely used to enhance detection sensitivity, though conventionally limited in multiplexing by the types of molecules that can be deposited. Recent work demonstrated that barcoded tyramide oligonucleotides can be deposited downstream of unconjugated antibody staining or in situ hybridization, enabling high-plex imaging¹⁶. However, multimodal integration at high-plex has not been demonstrated. In addition, horseradish peroxidase (HRP)-conjugated oligonucleotides combined with fluorophore tyramides enable antibody signal amplification, but irreversible covalent fluorophore deposition limits plex capacity and precludes sample reimaging¹⁷.

We developed protein and nucleic acid serial tyramide amplification (PASTA) to address limitations in spatial tissue profiling, enhancing detection sensitivity across diverse spatial profiling methods. PASTA uses HRP-catalyzed covalent deposition of oligonucleotide tyramide conjugates on tissue sections (Fig. 1a). Unlike traditional fluorophore-based approaches, deposited oligonucleotides remain intact for extended periods, enabling delayed or repeated imaging and seamless integration with existing spatial technologies.

Fig. 1 |. PASTA enables tunable signal amplification and multimodal integration for high-sensitivity spatial profiling.

a, Overview schematic of the PASTA workflow. HRP recruitment via in situ hybridization (ISH), unconjugated antibodies and PLA enables tyramide radical formation. These radicals deposit oligonucleotides onto nearby proteins through covalent binding to tyrosine residues, creating stable protein–oligonucleotide conjugates that enable serial amplification. b, Concentration-dependent signal amplification of CD3 and Iba1 in FFPE tissues. Images show unamplified detection versus PASTA amplification at increasing tyramide-oligonucleotide concentrations. Scale bars, 100 μm. Right panels show quantification of signal intensity for unamplified and PASTA-amplified signal, and following fluorescent reporter stripping.CD3: N = 282,637 (unamplified), 96,553 (2.5 μM), 94,733 (5 μM), 91,351 (10 μM). Iba1: N = 282,637 (unamplified), 91,351 (2.5 μM), 94,733 (5 μM), 96,553 (10 μM). c, Post-hoc rescue of suboptimal markers with PASTA. The same tissues can undergo PASTA amplification following initial imaging, enabling reimaging of poorly detected markers. Examples show weak initial signal for CAIX in tumor tissue and CD4 in a lymph node, both substantially enhanced after PASTA. Scale bars, 100 μm (main images), 25 μm (insets). d, Viral nucleic acid detection of SIV vDNA and vRNA with PASTA alongside CODEX staining for T cells (CD3), follicular dendritic cell markers (CD21) and PASTA staining for viral protein p17. This demonstrates vDNA+ and vRNA/protein− cells (white arrows), vDNA+ and vRNA/ protein+ (orange arrows), vDNA− and vRNA/protein+ cells (orange arrowheads) and FDC-bound vDNA− and vRNA/protein− virus (white arrowheads). Scale bars, 100 μm (main images), 25 μm (insets). e, Extended ISH multiplexing via sequential application of two 4-plex probe sets on the same tissue section through probe stripping and reapplication, followed by CODEX staining for Ki67 and NaKATPase. Images show distinct spatial patterns across all eight RNA targets and were confirmed strippable in blank cycles imaged at equal exposures and signal thresholds as the signal images. Scale bars, 25 μm. Schematics in a–e created in BioRender; Parmelee, L. https://biorender.com/lywm40r (2026)

We integrated PASTA into the CODEX microfluidics platform for iterative stripping and hybridization of fluorophore-tagged complementary oligonucleotides^3,18. For PASTA amplification, antibodies conjugated with oligonucleotide barcodes (set A) are bound to their targets, followed by oligo-HRP recognition of set A, catalyzing deposition of distinct tyramide-oligonucleotide barcodes (set B), visualized with complementary fluorophore-labeled oligonucleotides. Unamplified detection directly visualizes set A barcodes with their respective fluorophore complements. Crucially, the use of distinct sequences for sets A and B means PASTA does not alter the original unamplified signal. PASTA achieves strong signal enhancement with minimal background in stripped controls (Fig. 1b and Extended Data Fig. 1a), with amplification scaling with tyramide-oligonucleotide concentration for tunable sensitivity. Chemical stripping was confirmed sufficient to remove HRP-oligo activity between depositions (Extended Data Fig. 1b). Signal enhancement was confirmed with clinically relevant markers on the same tissue section (Extended Data Fig. 1c), and PASTA maintained punctate signal integrity for the mitochondrial protein ATP5A (Extended Data Fig. 1d). Here imaging was performed at 0.3775 μm per pixel; the diffraction-limited lateral resolution of the system is approximately 0.86 μm (d = λ/2NA (numerical aperture) = 647 nm/(2 × 0.75)), confirming that spatial resolution rather than pixel sampling is the limiting factor for punctate signal fidelity. An automated protocol on the Opentrons OT-2 liquid handling platform was developed, enabling PASTA with substantially reduced hands-on time and negligible background deposition (Extended Data Figs. 2 and 3a). We further demonstrated repeated oligonucleotide deposition with distinct barcodes and minimal saturation (Extended Data Fig. 3b), and successful application across tissue section thicknesses beyond the standard 5 μm (Extended Data Fig. 4a).

A significant challenge in spatial proteomics is rescuing markers with suboptimal signal-to-noise ratios from precious clinical samples following initial imaging. PASTA enables post-hoc amplification ‘rescue’ on tissues that have already undergone spatial proteomics imaging (Fig. 1c), demonstrated for tumor marker CAIX and immune marker CD4, both showing notable signal enhancement following PASTA amplification. Beyond antibody-based detection, PASTA integrates diverse spatial profiling assays with HRP readout. We applied PASTA to detect simian immunodeficiency virus (SIV) provirus viral DNA (vDNA), transcripts viral RNA (vRNA) and follicular dendritic cell (FDC)-bound viral particles in infected rhesus macaque lymph node tissues (Fig. 1d), combined with CODEX staining for CD3 and CD21, enabling visualization of distinct infection states and FDC-bound virus with high specificity¹⁹ (Extended Data Fig. 4b). PASTA similarly detected Epstein–Barr virus (EBV) episomal DNA in EBV-positive primary central nervous system lymphoma (PCNSL) and spike RNA in severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)-infected rhesus macaque lung tissues (Extended Data Figs. 4c and 5a), with simultaneous multiplex protein codetection including CD20, CD3 and CLDN5 for PCNSL, and CD20, S100A9 and cytokeratin for SARS-CoV-2, enabling correlative analysis of viral distribution and host cellular responses within the same tissue section. PASTA extends the multiplexing capacity of commercial in situ hybridization through serial probe stripping and reprobing, achieving visualization of eight RNA targets across two probe sets on the same formalin-fixed, paraffin-embedded (FFPE) sections with simultaneous protein codetection (Fig. 1e and Extended Data Fig. 5b,c), supporting robust spatial multi-omics with standard microscopy and off-the-shelf reagents. PASTA further accommodates conjugation-resistant antibodies within multiplexed workflows (Extended Data Fig. 5d), validated across basal ganglia (A2AR), tonsil (CD73) and colorectal adenocarcinoma (A2AR, CD73 and PD-L1), and independently of tissue thickness (Extended Data Fig. 4a). We also demonstrated PASTA compatibility with proximity ligation assays (PLA) for spatial detection of CD4–MHC II interactions within a spatial proteomics workflow (Extended Data Fig. 5e).

To demonstrate comprehensive multimodal integration on the same slide, we deployed PASTA–CODEX incorporating (1) in situ hybridization, (2) conjugation-resistant antibodies, (3) automated signal amplification and (4) post-hoc marker rescue on a B cell lymphoma tissue microarray (TMA) cohort comprising Hodgkin’s lymphoma, diffuse large B cell lymphoma (DLBCL) and reactive LNs/tonsil (Fig. 2a). Two serial sections were compared: one underwent all PASTA integration steps, while the other underwent CODEX staining alone for 28 antibodies (Extended Data Fig. 6a). Cell type annotation revealed high concordance in cell type proportions between the PASTA and CODEX-only slides (Fig. 2b,c, Extended Data Fig. 6b, 7 and 8a and Supplementary Figs. 1 and 2). Automated PASTA amplification of CD3 before initial CODEX imaging confirmed signal enhancement consistent with earlier benchmarking (Extended Data Fig. 8b). In situ hybridization integration with CODEX enabled detection of CXCR5, CXCL9 and CCL19, revealing higher CXCR5 RNA and protein expression in DLBCL tumor cells compared with Hodgkin’s and LN/tonsil B cells (Fig. 2d and Extended Data Fig. 9) and elevated CXCL9 RNA in myeloid cells across both lymphoma subtypes compared with LN/tonsil (Extended Data Fig. 10a). CD69 and CTLA-4, which were unavailable in conjugated format or failed bioconjugation, respectively, were integrated using unconjugated primary antibodies and PASTA, confirming T cell expression across all tissue types with elevated expression in lymphomas compared with LN/tonsil (Fig. 2e and Extended Data Fig. 10b). Finally, post-hoc PASTA rescue of CD57 and ICOS, which performed suboptimally in initial CODEX imaging, with CD57 being barely detectable across many tissues, strongly amplified both signals enabling robust detection and analysis (Fig. 2f).

Fig. 2 |. PASTA enables multimodal integration to study the B cell lymphoma tumor microenvironment.

a, Overview of PASTA-enabled multimodal analysis of a clinical TMA integrating seven additional markers beyond a standard CODEX spatial proteomics antibody panel, including transcripts, conjugation-resistant antibodies (Abs) and rescued markers from suboptimal CODEX antibody performance. This is in comparison with a serial section processed without PASTA for multimodal integration. b, Representative multiplex immunofluorescence images and cell type annotation of identical cores for Hodgkin’s and DLBCL from both PASTA and CODEX slides. Scale bars, 500 μm (main images), 100 μm (insets). c, Correlation of cell type annotation proportions of total cells between PASTA and CODEX slides. Each dot represents one cell type from one core (n_core = 14, n_total = 130) with ‘Other’ cell types being displayed as black dots. R² = 0.9874, P = 1.77 × 10⁻¹²³ (Pearson’s correlation, two-sided t-test). d, RNAScope detection of CXCR5-RNA (green) and CXCR5 protein (magenta). The plot shows median expression by tumor cells (for Hodgkin’s and DLBCL) and B cells (for lymph node (LN)/tonsil) where each dot is one TMA core (n = 14). Scale bars, 500 μm (main images), 25 μm (insets). e, Integration of conjugation-resistant antibodies, using unconjugated antibodies with PASTA to detect CD69 (green) and CTLA-4 (magenta). The plot shows median expression by different T cell populations where each dot is one TMA core (n = 14). Scale bars, 500 μm (main images), 25 μm (insets). f, PASTA-enabled post-hoc antibody rescue of suboptimally performing antibodies for CD57 and ICOS on the same slide. Unamplified (Unamp.) images show detection of conjugated, unamplified oligonucleotide barcode; rescued images show detection of PASTA signal on the same cells. Gradient legends show thresholds for image presentation. Plot shows median expression of raw image data in different T cell populations, where each dot is one TMA core (n = 14). The y axis is scaled so that unamplified signal is near 1 to allow easy visual assessment. Scale bars, 500 μm (main images), 25 μm (insets). Schematics in a created in BioRender; Parmelee, L. https://biorender.com/lywm40r (2026).

Discussion

PASTA addresses critical challenges in spatial tissue profiling by providing a versatile, platform-agnostic signal amplification approach compatible with diverse spatial biology modalities. By using HRP-catalyzed oligonucleotide barcode deposition, PASTA enables tunable signal amplification adaptable to target abundance and application requirements.

As an enzymatic TSA-based approach, PASTA has inherent limitations in absolute biomolecule quantification. At high tyramide oligonucleotide concentrations, deposition saturation can produce a plateau effect and may blur spatial resolution of punctate targets. These effects can be mitigated by careful concentration selection and using distinct barcode sequences for unamplified and amplified channels. While PASTA can improve signal-to-noise ratios for dim markers, it is important to note that nonspecific background signal is also amplified proportionally, which may limit PASTA’s ability to rescue markers where poor signal-to-noise ratios arise primarily from nonspecific staining rather than insufficient signal. Some cycle-to-cycle variability was observed, potentially attributable to tyramide oligonucleotide manufacturing purity or hybridization chemistry, although this lies outside the scope of this study. Furthermore, it remains unclear whether the extensive processing steps involved in PASTA workflows may interfere with downstream molecular assays including compatibility with same-slide spatial transcriptomics approaches such as IN-DEPTH²⁰, which has not been evaluated and warrants future investigation.

PASTA’s versatility spans commercially available platforms including Akoya Biosciences’ PhenoCycler, ACDBio’s RNAscope and Navinci’s PLA kits, using off-the-shelf reagents without specialized equipment. Compatibility with in situ hybridization, PLA and conventional antibody staining enables true spatial multi-omics. This extends to pathogen detection, as demonstrated for SIV, EBV and SARS-CoV-2, enabling simultaneous viral nucleic acid detection and host cellular phenotyping within the same tissue section. Experimental design must account for potential incompatibilities; for example, unconjugated antibody staining should precede conjugated staining to prevent nonspecific secondary antibody binding.

A particularly valuable feature is post-hoc amplification for rescuing suboptimal markers from precious clinical samples without requiring additional tissue sections. Importantly, PASTA integration does not significantly alter downstream cell phenotyping, as demonstrated by high concordance between PASTA-integrated and CODEX-only workflows. Compatibility with unconjugated antibodies accommodates antibodies resistant to conjugation chemistries, and applicability across tissue section thicknesses beyond 5 μm further broadens clinical utility.

PASTA requires custom reagents, most notably tyramide and HRP oligonucleotides. While individually inexpensive and stable with appropriate storage (Supplementary Table 5), users should plan barcode diversity in advance for larger panels. An automated liquid handling protocol further reduces hands-on time and technical variability, lowering the barrier to adoption for laboratories without specialized expertise. Future developments could include standardized reagent kits and optimization for additional detection modalities. As spatial biology advances toward clinical translation, PASTA provides a cost-effective and versatile approach to enhance detection sensitivity, rescue suboptimal results and expand multimodal capabilities within established workflows.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41592-026-03200-z.

Methods

FFPE tissues

This research complies with all relevant ethical regulations for the use of human samples. The FFPE tissues used in this study were commercially available tonsil tissues (AMSBio, #AMS6022) for Figs. 1b and 2 and Extended Data Figs. 1–10 and commercially available HeLa cell pellets (ACDBio, #310045) for Fig. 1e and Extended Data Fig. 5b. The basal ganglia (Extended Data Fig. 5d) tissue was sectioned from a block of healthy, human, post-mortem basal ganglia, which was a generous gift from Derek Allison (University of Kentucky). The remaining FFPE tissues were clinical samples used as part of ongoing studies. We thank Jacob D. Estes (OHSU-VGTI) for providing the SIV+ and SIV–tissues (Fig. 1d). The PCNSL tissue (Extended Data Fig. 5a) was part of a TMA generated by M.B. and C.K. The SARS-CoV-2-infected nonhuman primate tissues²¹ (Extended Data Fig. 5a) were a generous gift from Kristina De Paris (UNC, Chapel Hill) and Koen Van Rompay (UC Davis CNPRC). Renal cell carcinoma TMAs with tumor tissue (Fig. 1c, CAIX) and lymph nodes (Fig. 1c, CD4) were provided by N.E.A. and S.S. The colorectal adenocarcinoma tissue (Extended Data Fig. 5d) was part of a TMA provided by Z.X.Z. and T.K.H.L. The B cell lymphoma TMA (Fig. 2) was kindly provided by T.P. and R.C.

FFPE tissues were sectioned at 5 μm thickness onto Fisherbrand SuperFrost Plus Microscope Slides (12-550-15, FisherScientific) or 22 × 22mm #1 coverslips (48366-067, VWR) pretreated with Vecta-Bond (SP-1800-7, Vector Laboratories) following the manufacturer’s instructions. By contrast, the tissues for Extended Data Fig. 4a were sectioned at 10 μm and 20 μm.

The tissues used in this study were either destroyed after data acquisition or subjected to hematoxylin and eosin staining followed by permanent storage.

Tissue processing

FFPE slides or coverslips were baked at 70 °C for 1 h and subsequently deparaffinized twice for 5 min each in Xylenes. To rehydrate the tissues, slides were mounted into a linear stainer (ST4020, Leica Biosystems) and incubated in the following solutions for 3 min each with dipping every 5 s: 3 × Xylenes, 2 × 100% ethanol, 2 × 95% ethanol, 1 × 80% ethanol, 1 × 70% ethanol, and 3 × ultrapure water (10977015, ThermoFisher). Slides were subsequently transferred to Dako pH 9 target retrieval solution (S2367, Agilent) preheated to 75 °C in a PT module (ThermoFisher, A80400012), before being retrieved at 97 °C for 20 min. After the PT module cooled down to 65 °C, the slides were removed and cooled down to room temperature for 10 min. Slides were then washed in ultrapure water before the tissues were surrounded by a hydrophobic barrier pen (H-4000, Vector Laboratories). The pen was left to dry for 5 min while the tissue remained covered with ultrapure water.

CODEX staining of FFPE tissues

The tissue was washed in ultrapure water before being covered with 3% H₂O₂ in 1× Tris-buffered saline (TBS; final, pH 7.5) and incubated for 10 min to inactivate endogenous peroxidases. Afterward, the tissue was washed again in ultrapure water before being washed in 1× TBS-T (0.05% Tween-20, pH 7.5). To block endogenous biotin, the tissue was covered with a few drops of avidin solution (927301, BioLegend) and incubated for 30 min before being washed three times in 1× TBS-T. Then, the tissue was covered with a few drops of biotin solution (927301, BioLegend) and incubated for 30 min before being washed three times in 1× TBS-T.

To prepare the tissue for CODEX staining, the tissue was blocked with CODEX blocking solution (0.75× TBS-T, 3.75% donkey serum (D9663-10ML, Sigma), 0.75% Triton X-100, 0.0375% NaN₃, 1 mg ml⁻¹ sheared salmon sperm DNA (AM9680, ThermoFisher), 100 μg ml⁻¹ mouse IgG (I5381-10mg, Sigma), 100 μg ml⁻¹ rat IgG (I4141-10mg, Sigma) and 100 nM of all unlabeled bottom oligos (Integrated DNA Technologies)) for 1 h in a humidity chamber while photobleaching using strong light-emitting diode lights (Best Buy, 6460231 and Amazon, B07C68N7PC). The temperature was monitored to not exceed 40 °C. Meanwhile, the CODEX antibody cocktail was prepared as described previously¹⁸. In brief, conjugated antibodies (Supplementary Tables 1 and 4) were spun down at 12,000g for 15 min at 4 °C, after which appropriate antibody volumes were added to antibody diluent (1× TBS-T, 5% donkey serum, 0.05% NaN₃). The antibody mixture was then added to a 1× TBS-T prewetted 50-kDa centrifugal filter (UFC5050BK, Sigma) and spun down at 12,000g for 8 min. The concentrated antibody solution was eluted by inverting the filter into a fresh collection tube and spinning at 12,000g for 2 min. The volume of the solution was measured using a pipette. The volume needed to bring the volume of the solution to 0.75× of the final volume was added as antibody diluent to the filter and used to gently rinse through the filter. The filter was then eluted again by inverting it into the same collection tube and spinning again at 12,000g for 2 min. One quarter of the final volume was added as FFPE block (0.325× Dulbecco’s PBS (DPBS) (ThermoFisher, 14190144), 162.5 mM NaCl, 25.35 mM NaH₂PO₄, 39.65 mM Na₂HPO₄, 1.625 mM EDTA, 0.1625% bovine serum albumin (BSA), 0.02% NaN₃, 0.05 mg ml⁻¹ mouse IgG, 0.05 mg ml⁻¹ rat IgG, 0.5 mg ml⁻¹ sheared salmon sperm DNA and 100 nM of all unlabeled bottom oligos). Finally, the solution was filtered through a prewetted 0.1-μm filter (UFC30VV00, Sigma) by spinning at 12,000g for 2 min. After the blocking was completed, the excess blocking buffer was blotted off the slide and the antibody solution added to the tissues. The slides were left to incubate at 4 °C overnight.

The next day, the slides were washed twice in S2 buffer (0.5× DPBS (14190144, ThermoFisher), 250 mM NaCl, 39 mM NaH₂PO₄, 61 mM Na₂HPO₄, 2.5 mM EDTA, 0.25% BSA and 0.02% NaN₃) for 2 min each. Antibodies were fixed by incubating the tissues for 10 min in 1.6% paraformaldehyde (15710, Electron Microscopy Sciences) in S4 buffer (0.9× DPBS, 500 mM NaCl, 4.5 mM EDTA, 0.45% BSA and 0.02% NaN₃) before being rinsed once and washed twice in 1× DPBS (14190144, ThermoFisher). Then, the slides were submerged in ice-cold methanol on ice for 5 min before being rinsed once and washed twice in 1× DPBS. Finally, the tissues were treated with final fixative (4 μg μl⁻¹ (from 200 μg μl⁻¹ BS3 (21580, ThermoFisher) stock solution in dimethyl sulfoxide) in 1× DPBS) for 20 min in the dark before being rinsed once and washed twice in 1× DPBS. The slides were stored in S4 buffer at 4 °C until subsequent steps.

PASTA amplification of CODEX antibody signal

To amplify the CODEX signal, slides were washed in azide-free 1× CODEX buffer (150 mM NaCl, 10 mM Tris–HCl pH 7.5 (15567027, ThermoFisher), 10 mM MgCl₂ and 0.1% Triton X-100) twice for 5 min each. An initial chemical strip was performed to remove blocking oligos bound to the antibodies by incubating the slides twice in azide-free stripping buffer (80% dimethyl sulfoxide, 20% azide-free 1× CODEX buffer) for 5 min each. Slides were then washed in azide-free 1× CODEX buffer. For each cycle of the serial tyramide deposition, the following steps were performed: First, the slides were incubated with HRP conjugated oligos or biotin conjugated oligos (Supplementary Tables 2 and 4) at 200 nM in azide-free plate buffer (0.5 mg ml⁻¹ sheared salmon sperm DNA (AM9680, ThermoFisher) in azide-free 1× CODEX buffer) for 20 min in the dark. In the case of biotin oligos, this was followed by two washes with azide-free 1× CODEX buffer and a 20-min incubation with 1:100 HRP–streptavidin (405210, BioLegend) in azide-free 1× CODEX buffer in the dark. Second, the slides were washed twice for 2 min each in azide-free 1× CODEX buffer and then once in 1× TBS for 2 min. Third, the tissues were incubated for 10 min with tyramide-conjugated oligos diluted in TSA diluent buffer (FP1498, Akoya Biosciences) in the dark. Fourth, the slides were washed twice with azide-free 1× CODEX buffer. Finally, the HRP oligos were chemically stripped by incubating the slides twice for 5 min each with azide-free stripping buffer, followed by two washes in azide-free 1× CODEX buffer for 2 min each. This process was repeated serially until all targets had been amplified.

Testing HRP-oligo strippability for PASTA of CODEX antibody signal

In addition to the initial chemical strip with 2 × 5-min washes in azide-free stripping buffer, the slides were incubated with 3% H₂O₂ in 1× TBS at 40 °C for 30 min in a hybridization oven (ACDBio). Subsequent steps were carried out as outlined above until the end of the tyramide oligo incubation (here oligo 1), at which point the slides were washed twice for 2 min each in azide-free 1× CODEX buffer. The unstripped slide was moved to azide-free 1× CODEX buffer and stored in the dark until a later step. The stripped slide was incubated twice for 5 min each in azide-free stripping buffer, followed by two washes for 2 min each in azide-free 1× CODEX buffer. At this point, the unstripped slide was retrieved and both slides were washed in 1× TBS for 2 min, after which the tissues were incubated with tyramide oligo (here oligo 2) diluted in TSA diluent buffer (FP1498, Akoya Biosciences) in the dark for 20 min. Finally, the slides were washed twice in azide-free 1× CODEX buffer for 2 min each.

Automated PASTA amplification of CODEX antibody signal using the OT-2 liquid handling platform

We have developed custom protocols for carrying out PASTA for CODEX antibody signal amplification on the OT-2 liquid handling platform (OpenTrons). This uses the coverslip or slide OmniStainer C12 or S12 (ParheliaBio) together with the corresponding microfluidic coverslip holder or microfluid slide covers (ParheliaBio). The protocols are available in a GitHub repository (https://github.com/SizunJiangLab/PASTA). In brief, all washing and incubation buffers (azide-free 1× CODEX buffer, azide-free stripping buffer, 1× TBS) are provided in 12-well reagent reservoirs (CT229562, Stellar Scientific) sealed with adhesive metal foil (AB0626, ThermoFisher) to reduce reagent evaporation. All other reagents are prepared per sample in a round bottom 96-well plate (BR781607-100EA, Sigma) such that there is one well per sample per cycle for HRP oligos in azide-free plate buffer, tyramide oligos and TSA buffer. The plates are sealed with adhesive metal foil and placed on a temperature module (OpenTrons) which is set to 4 °C. All wash and reagent application steps are split into multiple lower-volume washes to mitigate the potential for local reagent depletion in the microfluidic system. At the end of the protocol, an automated hydration protocol is included that adds a small amount of buffer at regular intervals for up to 24 h to prevent tissues from drying up.

Automated multiplex imaging

For the Akoya PhenoCycler Fusion, slides were moved into 1× PBS from their storage solution (if applicable) to remove detergents. The edges of the slides were dried, and a flow cell (Akoya Bioscience) was attached using the manufacturer’s flow cell press. The slides were subsequently incubated in 1× CODEX buffer for 10 min to allow the flow cell adhesives to harden. The PhenoCycler Fusion buffers were freshly prepared in-house and loaded into the machine. The slides were loaded into the machine following manufacturer instructions with an output setting for 16-bit images.

For the Akoya CODEX fluidics in combination with the Keyence BZ-X810 microscope, the machine was loaded with fresh 1× CODEX buffer and dimethyl sulfoxide. The coverslips were loaded into the fluidics stage and briefly stained with Hoechst 33342 (ThermoFisher, H3570) diluted 1:1,500 in plate buffer to identify imaging regions of interest. The automated machine was run following manufacturer’s instructions with an output setting for 16-bit images, which were processed using the Akoya Singer software (v1.0.7).

Supplementary Table 1 indicates for each experiment whether imaging took place using the Akoya PhenoCycler Fusion or the Keyence BZ-X810 setup.

In addition, a reporter plate was prepared by adding ATTO 550- or Alexa Fluor 647-conjugated oligos (GenScript, Biomers) complementary to the oligo barcodes on the antibodies or deposited on the tissue at 100 nM into plate buffer supplemented with Hoechst 33342 at a 1:300 dilution to a final volume of 250 μl. The wells were sealed with adhesive foil.

Image quantification of CODEX and PASTA signal

For image processing, most image data was processed to remove autofluorescence background signal via background subtraction with a blank acquired at similar exposure (Figs. 1c,d and 2 and Extended Data Figs. 1d, 2, 3a,b, 4b,c, 5a,d,e and 6–10). For some images, data without background subtraction were used instead (Fig. 1b,e and Extended Data Figs. 1a–c, 4a and 5b,c).

For image alignment, due to the limited number of imaging cycles that can be carried out on the PhenoCycler Fusion at a time (Fig. 1b and Extended Data Fig. 1c), the markers were imaged across four separate but sequential imaging runs. Multiple imaging cycles were also necessary for cases of marker rescue (Figs. 1c and 2f). As a first step, the images from these runs were cropped into a single tonsil per image and subsequently aligned using VALIS²² and combined into a single ome-tiff.

Similarly, for cases where raw, unprocessed data without background subtraction were used (Fig. 1e and Extended Data Fig. 4a), VALIS was used to align the raw images and combine them into a single ome-tiff.

For image segmentation and feature extraction, the images were segmented using MESMER²³ with default parameters and designating nuclear (Hoechst 33342) and membrane using the conjugated, unamplified marker sets indicated in Supplementary Table 6. Single-cell feature extraction for the fluorescent signal for each cell was performed by summing up the pixel values for a marker for each cell and dividing it by the area of that cell.

For data preprocessing (Fig. 2), following staining quality control, cores that stained aberrantly for CODEX antibodies and/or Hoechst 33342 were excluded moving forward. Marker expression data scaled by cell size were first filtered to remove segmented cells without nuclear signal. Next, the signal of each marker was normalized by dividing it by the median nuclear signal for each core. To reduce the skewing of the distribution to aid in annotation moving forward, the data were adjusted with a global inverse hyperbolic sine transformation with a cofactor of 0.1. We then applied a universal percentile normalization on a per-core level to normalize all annotation markers to a [0,1] range and to adjust for staining intensity between cores. For all markers except CD30, the lower percentile cutoff was 10% and the upper percentile cutoff was 99%. For CD30, we adjusted the lower percentile cutoff to 75% to account for high background signal. The corresponding data frame was used for annotation.

For functional markers (Fig. 2d–f and Extended Data Figs. 9 and 10) as well as CD3 PASTA signal (Extended Data Fig. 8b), expression data were filtered and normalized with the median nuclear signal as described above. The resulting data frame without universal percentile normalization was used for markers where the absolute magnitude of signal intensity is important for comparing two markers with each other (Fig. 2f and Extended Data Fig. 8b). For all other functional markers (Fig. 2d,e and Extended Data Fig. 8a,b), the data frame for the PASTA slide was additionally processed with a global universal percentile normalization to a [0,1] range with lower cutoff of 1% and upper cutoff of 99%.

Cell type annotation (Fig. 2) was performed using PhenoGraph²⁴ with unamplified, conjugated signal for CD3, CD4, CD8, FoxP3, CD20, Pax5, CD163, CD11c, CD68, CD56, CD30 and CD31. CD31 clustering was subsequently ignored due to poor staining quality. Multiple rounds of subclustering took place combined with visual assessment of annotation quality. Tumor annotation for Hodgkin’s proved difficult based on clustering alone and was subsequently carried out manually on all Hodgkin’s tumor cores based on CD30 and Pax5 staining. Finally, cells annotated as B cells in the DLBCL tumors were reassigned as tumor. An overview of annotation quality is available in Extended Data Fig. 6 and 7. All annotations were confirmed by a board-certified pathologist (R.C.).

For quantification and data presentation, for the markers of interest, the unamplified and PASTA signal as well as their corresponding stripped signal (if applicable) were plotted on a log₁₀ scale. For the unamplified signal in Fig. 1b, the signal for all three tissues was combined, while for Extended Data Fig. 1c, the unamplified signal was specifically for the same tissue as that shown for the PASTA signal. For the assessment of stripping sufficiency for HRP-oligo removal in Extended Data Fig. 1b, the tissue treated with 10 μM tyramide oligonucleotides was analyzed by plotting the marker expression of all cells for the unamplified oligo, the stripped state and tyramide oligos 1 and 2. To handle cells with an expression value of zero in log₁₀-scaled plotting (Fig. 1 and Extended Data Figs. 1–4), all cells with a zero value were set to the smallest nonzero value for each marker and each condition.

In cases where the median marker expression for specific cell types was plotted (Fig. 2d–f and Extended Data Fig. 8b and 10a), the median expression of the marker of interest for each core for each cell type of the PASTA slide was calculated. For markers that were not normalized to a [0,1] scale (Fig. 2f and Extended Data Fig. 8b), the median values were adjusted by multiplying them with a common factor for all values for each marker so as to set the unamplified signal to 1 to allow easier comparison.

For quantification in tissues that were not segmented (Extended Data Fig. 4a), the images were pseudo-subsampled by placing nine fields of view (FOVs) on the image in matched locations for each tissue. Each FOV had a size of 1,000 × 1,000 pixels (500 × 500 μm). To quantify marker expression, the median pixel value for each marker in each FOV was calculated.

PASTA amplification for in situ hybridization of viral nucleic acids

The processed tissues were covered with RNAScope Hydrogen Peroxide (ACDBio, #322335) and incubated for 10 min at room temperature (SIV DNA/RNA and SARS-CoV-2 RNA) or 20 min at 40 °C (EBV DNA) to block endogenous peroxidase activity. Afterward, the tissue was washed again in ultrapure water. C2–C4 probes were diluted 1:50 in C1 probe solution (Supplementary Table 3) and then prewarmed for an additional 15 min. Fifty microliters of the probe solution was added to the tissues before being placed in a humidity tray which was placed in a hybridization oven (ACDBio) overnight at 40 °C.

The next day, the slides were washed twice for 2 min each in 0.5× RNAScope Wash Buffer (ACDBio, #310091). Next,two drops of Multiplex v2 Amp 1 (ACDBio, #323101) were added to each tissue and incubated for 30 min at 40 °C. The slides were rinsed and washed twice followed by incubation with two drops of Multiplex v2 Amp 2 (ACDBio, #323102) for 15 min at 40 °C. The slides were rinsed and washed twice followed by incubation with two drops of Multiplex v2 Amp 3 (ACDBio, #323103) for 30 min at 40 °C. The slides were rinsed and washed twice followed by incubation with two drops of Multiplex v2 HRP-C1 (ACDBio, #323104) for 15 min at 40 °C. The slides were rinsed and washed twice followed by incubation with two drops of RNAScope Brown Kit Amp 5 (ACDBio, #322315) for 45 min at room temperature. The slides were rinsed and washed twice in 1× TBS-T followed by incubation with 2 drops of RNAScope Brown Kit Amp 6 (ACDBio, #322316) for 15 min at room temperature. The slides were rinsed and washed twice followed by incubation with tyramide oligo diluted to a final concentration of 5 μM in TSA diluent buffer (FP1498, Akoya Biosciences) at room temperature for 15 min. After washing again, two drops of HRP Blocker (ACDBio, #323107) were incubated for 15 min at 40 °C followed by two washes. This process was repeated for the remaining channels as applicable.

Following the full RNAScope protocol, the tissues were washed with 1× TBS-T twice for 2 min each. For the SIV tissues (Fig. 1d), the antibody for SIV p17 protein was integrated as an unconjugated antibody before proceeding to CODEX staining. This was carried out as below with a tyramide oligo concentration of 5 μM. Then, tissues were blocked and stained with CODEX antibodies as described above and incubated overnight at 4 °C. The next day, the tissues were washed and fixed as described above. The slides were imaged on the PhenoCycler Fusion.

Rescue of conjugated antibody signal using PASTA

Following the identification of markers requiring signal amplification for marker rescue (Fig. 1c), the coverslips were rinsed twice for 2 min each in 1× TBS-T. As the tissues had previously been stained with a biotinylated antibody and treated with fluorescent streptavidin, an additional avidin–biotin block was carried out by incubating 20 min with avidin solution (927301, BioLegend), washing twice for 2 min each in 1× TBS-T, incubating 20 min with biotin solution (927301, BioLegend) and finally washed twice for 2 min each in 1× TBS-T. To block the activity of previous HRP conjugated antibodies, the tissues were incubated with HRP Blocker (ACDBio, #323107) for 15 min at 40 °C followed by two washes for 2 min each in 1× TBS-T.

Subsequently, the tissues were stripped, washed, incubated with 100 nM biotin or HRP oligos (Supplementary Tables 2 and 4) in azide-free plate buffer for 10 min in the dark, washed, stained with streptavidin HRP (ThermoFisher, S911) diluted 1:100 in azide-free 1× CODEX buffer for 20 min in the dark, washed, incubated with 3 μM of tyramide oligo diluted in TSA diluent buffer (FP1498, Akoya Bio-sciences) for 10 min, and finally washed again. The process was carried out as above and repeated for all markers to be rescued.

For the rescue on slides in Fig. 2, stripping, washing and staining with 200 nM HRP oligos (Supplementary Tables 2 and 4) was carried out inside the PhenoCycler Fusion imaging flow cell by carefully pushing liquid into the flow cell with a P200 tip. The PASTA reaction took place with 5 μM of tyramide oligos diluted in TSA diluent buffer for 20 min. This was repeated for all rescued markers.

The rescued markers were imaged with fluorescent reporter oligos as indicated in Supplementary Tables 1 and 4 on the same instrument as the original CODEX imaging.

Integration of unconjugated and biotinylated antibodies into CODEX using PASTA

The processed tissues were then rinsed three times with 1× TBS-T for 2 min each. To block endogenous biotin, the tissues were covered with a few drops of avidin solution (927301, BioLegend) and incubated for 20 min, with avidin being refreshed at the 10-min mark, before being washed three times in 1× TBS-T. The tissues were then covered with a few drops of biotin solution (927301, BioLegend) and incubated for 20 min, with biotin being refreshed at the 10-min mark, before being washed three times in 1× TBS-T. Following avidin and biotin blocking, the tissues were blocked with blocking solution (0.75× TBS-T, 3.75% donkey serum (D9663-10ML, Sigma), 0.75% Triton X-100 and 0.0375% NaN₃) for 1 h at room temperature. The primary unconjugated antibodies (Supplementary Table 1) were diluted in antibody diluent and added to the tissues following the blocking. The slides were incubated overnight at 4 °C.

The next day, the tissues were washed three times with 1× TBS-T for 2 min each. The tissues were then covered with 3% H₂O₂ in 1× TBS (final, pH 7.5) and incubated for 15 min at room temperature to inactivate endogenous peroxidases. After, the tissues were washed three times with 1× TBS-T before the tissues were incubated with the first secondary antibody, anti-mouse HRP (Origene, #D37-110) to target the primary antibody derived from a mouse host, for 30 min at room temperature. The tissues were then washed three times with 1× TBS-T before being incubated with tyramide-conjugated oligos in TSA diluent buffer (FP1498, Akoya Biosciences) in the dark for 15 min. The tissues were again washed with 1× TBS-T. To quench the active HRP activity, the coverslips were incubated with 3% H₂O₂ in 1× TBS (final) for 30 min at 40 °C. After rinsing the coverslips three times with 1× TBS-T, the tissues were incubated with the second secondary antibody; anti-rabbit HRP (Origene, #D13–110) to target the primary antibody derived from a rabbit host for 30 min at room temperature. The tissues were then washed three times with 1× TBS-T before being incubated with tyramide-conjugated oligos in TSA diluent buffer (FP1498, Akoya Biosciences) (note: the oligo must be distinct from that from the first round) in the dark for 15 min. The tissues were again washed with 1× TBS-T. Active HRP activity was again quenched by incubating the coverslips with 3% H₂O₂ in 1× TBS (final) for 30 min at 40 °C. The tissues were then washed three times with 1× TBS-T to remove any residual H₂O₂.

After the tyramide oligos were deposited, the tissues were incubated with CODEX blocking solution (as described above) and allowed to incubate while photobleaching for 1 h. CODEX antibodies including the biotinylated antibody were prepared as described above and incubated overnight at 4 °C. After fixation as described above, the tissues were washed with 1× TBS-T and then incubated with HRP–streptavidin (S911, ThermoFisher) diluted 1:100 in 1× TBS-T for 30 min at room temperature. Subsequently, the slides were washed with 1× TBS-T before being incubated with tyramide oligos in TSA diluent buffer (FP1498, Akoya Biosciences) in the dark for 15 min. Finally, the slides were washed twice with 1× TBS-T and twice with azide-free 1 CODEX buffer for 5 min each.

PASTA amplification of PLA signal

Deparaffinization and antigen retrieval were performed on study slides using conditions described above. After the retrieved slides have been cooled to room temperature for 15–20 min the slides were washed once in 1× TBS-T for 5 min, followed by S2 buffer for 20 min, after which the tissues were circled with a hydrophobic barrier pen. Then, the tissues were blocked with blocking buffer (PLA Buffer (Navinci), supplemented with 0.5 mg ml⁻¹ sheared salmon sperm DNA and 25 nM of unlabeled bottom oligos) for 1 h while photobleaching.

The PLA reaction was carried out using a modified version of the manufacturer’s protocol. In brief, primary antibodies were incubated at 1:100 dilution in Navenibody diluent (Navinci) at 37 °C for 1 h. The diluted navenibodies M1 and R2 (Navinci) were diluted to 1× working concentration in Navenibody diluent (Navinci) and added to the tissues followed by 1 h incubation at 37 °C. The slides were rinsed once and washed twice for 5 min each in 1× TBS-T prewarmed to 37 °C. Naveni Buffer 1 and Naveni Enzyme 1 (Navinci) were diluted in ultrapure water to 1× working concentration and incubated on the tissues for 30 min at 37 °C. The slides were rinsed once and washed once in 1× TBS-T. Naveni Buffer 2 and Naveni Enzyme 2 (Navinci) were diluted in ultrapure water to 1× working concentration and subsequently incubated on the tissues for 90 min at 37 °C. The slides were washed twice for 5 min each in 1× TBS before HRP reagent (Navinci) diluted in HRP diluent (Navinci) was incubated for 30 min at room temperature. Finally, the slides were washed twice in 1× TBS. For the TSA reaction, the tyramide-conjugated oligo and tyramide-conjugated digoxigenin were diluted to 5 μM and 1:250, respectively, in TSA diluent buffer (FP1498, Akoya Biosciences) (FP1498, Akoya Biosciences) and incubated on the tissue for 10 min at room temperature. Slides were washed twice for 5 min each in 1× TBS. CODEX antibody cocktail was prepared as described above, added to the slides and incubated overnight at 4 °C. Antibodies were fixed as described above, and slides were imaged on the PhenoCycler Fusion.

Combined high-plex RNAScope in situ hybridization with antibody staining using PASTA

The processed tissues were covered with RNAScope Hydrogen Peroxide (ACDBio) and incubated for 10 min at room temperature to block endogenous peroxidase activity. Afterward, the tissue was washed again in ultrapure water. C2–C4 probes were diluted 1:50 in C1 probe solution (Supplementary Table 3) and then prewarmed for an additional 15 min. Then, 50 μl of the probe solution was added to the tissues before being placed in a humidity tray, which was placed in a hybridization oven (ACDBio) overnight at 40 °C.

The next day, the slides were rinsed once and washed twice for 2 min each in 0.5× RNAScope Wash Buffer. Three drops of prewarmed Multiplex v2 Amp 1 were added and incubated for 30 min at 40 °C. Next, the slides were rinsed and washed as above followed by addition of three drops of prewarmed Multiplex v2 Amp 2 and incubated for 30 min at 40 °C. The slides were rinsed and washed again followed by addition of three drops of prewarmed Multiplex v2 Amp 3 and incubated for 15 min at 40 °C. For the development of the four multiplex channels (C1–C4), the following steps were repeated for each channel: three drops of HRP-C_n were added to each slide and incubated for 15 min at 40 °C. This was followed by a rinse and two washes. Meanwhile, tyramide oligo or fluorophore (for the four-channel positive control, see Supplementary Table 3) were diluted in TSA diluent buffer (FP1498, Akoya Biosciences) to a concentration of 5 μM (tyramide oligo) or 1:100 (tyramide fluorophores) and added to the tissue for incubation at room temperature for 20 min in the dark. Slides were rinsed and washed twice, after which three drops of HRP blocker were added to each tissue and incubated at 40 °C for 15 min followed by a final rinse and wash. For the fluorophore controls, the following tyramide conjugates were deposited in order (Supplementary Table 3): digoxigenin tyramide (Akoya Biosciences, #NEL748001KT), Alexa Fluor 647 tyramide (ThermoFisher, #B40958), Alexa Fluor 568 tyramide (ThermoFisher, #B40956) and Alexa Fluor 488 tyramide (ThermoFisher, #B40953).

For the fluorophore control slides, the slides were washed in 1× TBS-T for 2 min after the final RNAScope deposition. The tissues were blocked with blocking solution (0.75× TBS-T, 3.75% donkey serum (D9663-10ML, Sigma), 0.75% Triton X-100, 0.0375% NaN3) for 1 h at room temperature followed by staining with mouse anti-digoxigenin antibody (Novus Biologicals, #NBP2-31191) diluted 1:500 in antibody diluent (see above) for 1 h at room temperature. After the incubation, the slides were washed twice with 1× TBS-T for 2 min each, followed by secondary antibody staining with anti-mouse DyLight755 (ThermoFisher, #SA5-10175) diluted 1:100 in antibody diluent (see above), and incubated for 30 min at room temperature. The slides were washed twice with 1× TBS-T for 2 min each and finally stained with Hoechst 33342 (ThermoFisher, H3570) diluted in 1× TBS-T followed by two washes in 1× TBS-T for 2 min each. Slides were coverslipped with Prolong Gold (ThermoFisher, #P36930) and allowed to dry before being imaged with the PhenoImager Fusion (Akoya Bioscience).

After all four channels had been developed, the slides were placed in prewarmed stripping solution (90% dimethyl sulfoxide, 10% 0.5× RNAScope Wash Buffer) twice for 10 min each at 42 °C. Slides were washed twice in 0.5× RNAScope Wash Buffer and then placed in ultrapure water. Meanwhile, C2–C4 probes were diluted 1:50 in C1 probe solution and then prewarmed for an additional 15 min. Fifty microliters of the probe solution was added to the tissues before being placed in a humidity tray, which was placed in the hybridization oven overnight at 40 °C. The next day, the channels were developed as described above until all RNAScope markers had been developed. Slides were not stripped after the last RNAScope iteration.

To combine antibody staining, slides were washed in 1× TBS-T for 5 min twice. Then, the tissues were incubated with CODEX blocking solution and blocked for 2 h while photobleaching as described above. The tissues were then incubated with an antibody cocktail overnight at 4 °C. The next day, the tissues were processed and fixed as described above and stored in S4 buffer until imaging on the PhenoCycler Fusion. For increased sensitivity, the raw images (stitched but not aligned or background subtracted) were aligned using VALIS²² and combined into a single ome-tiff.

Statistics and reproducibility

Box plots display the median (center line), interquartile range (IQR; box, 25th–75th percentile) and whiskers extending to the most extreme values within 1.5× IQR of the box bounds (minimum and maximum). Data points beyond 1.5× IQR are not shown. Zero values were replaced with the minimum positive observed value within each group before plotting. Violins without boxplots are plotted with a median line. All violins are scaled to equal maximum width and therefore do not reflect sample size. Violin tails are trimmed to the range of observed data. For the correlation statistics (Fig. 2c and Extended Data Figs. 8a and 9), correlation was computed using Pearson’s correlation coefficient and tested, where applicable, with a two-sided t-test. All experiments in this study were carried out once after optimization. Following extensive quality control, markers of poor performance were excluded from analysis. No statistical method was used to predetermine sample size. The experiments were not randomized. The Investigators were not blinded to allocation during experiments and outcome assessment.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Extended Data

Extended Data Fig. 1 |. PASTA achieves robust signal amplification with high specificity, efficient HRP-oligo stripping, and preserved signal morphology.

(A) PASTA amplification and blank cycles corresponding to data shown in Fig. 1B. The blank cycles are in the cycle immediately following the signal acquisition at equal exposure. Images are displayed with matched minimum and maximum pixel thresholds for each marker and corresponding blanks. Scale bars: 100 μm. (B) Chemical stripping of HRP oligos is sufficient to remove HRP activity. Schematic (top) depicts the PASTA amplification workflow as introduced in Fig. 1 for deposition of oligo 1 followed by treatment with DMSO to chemically strip HRP-oligo or no DMSO strip before carrying out deposition of oligo 2. Chemical stripping is sufficient to reduce oligo 2 signal to below unamplified signal thus demonstrating its sufficiency for HRP activity removal. Scale bar: 100 μm. Plot shows marker expression and unamplified blank on a log₁₀-scale. N = 89971 cells (Not Stripped), 96553 cells (Stripped). Tyramide concentration: 10 μM. (C) Comparison of unamplified versus PASTA-amplified detection for clinically relevant markers (Fas Ligand, CD45RA, PD-L1, CD20, CD11c) on the same slide as in Fig. 1B. Scale bars: 100 μm. Violin plots quantify marker expression of unamplified and PASTA signal. N = 91351 cells (CD20, PD-L1, CD45RA, Fas Ligand), N = 96553 cells (CD11c). Tyramide concentration: 2.5 μM (Fas Ligand), 5 μM (CD45RA, PD-L1), 10 μM (CD20, CD11c). (D) Mitochondrial marker ATP5A maintains its punctate signal after automated PASTA application. Unamplified and PASTA show identical regions on the same slide. Scale bars: 20 μm. Schematics in b created in BioRender; Parmelee, L. https://biorender.com/lywm40r (2026).

Extended Data Fig. 2 |. PASTA negative controls demonstrate high specificity.

(A) Overview of negative control experiment. Two coverslips were stained with a complete antibody panel or a partial antibody panel which lacks three markers (PD-1, Iba1, ICOS). All antibodies on the complete panel coverslip were targeted for PASTA amplification using oligo-HRP. For the partial panel coverslip, only half was amplified (CD3, CD8, FoxP3, CD11c, CD163, CD20) while the other half was split into two controls. Three stained markers (CD4, Pax5, CD57) had no oligo-HRP added to their PASTA cycles. For three cycles, the same HRP-oligo and tyramide-oligos as for the complete panel coverslip (PD-1, Iba1, ICOS) were added. The coverslips were imaged back-to-back with equal exposures and channels on the same instrument. (B) Side-by-side images of all 12 markers on conjugated (unamplified) or PASTA amplified oligonucleotide barcode for the complete panel and partial panel coverslips. Identical thresholds are applied for the same detection condition (unamplified or PASTA) across both slides; thresholds differ between unamplified and PASTA channels for each marker. Scale bars: 100 μm. Plots show unamplified and PASTA signal for both tissues side by side. N = 185120 cells (complete panel), N = 186171 cells (partial panel). Schematics in a created in BioRender; Parmelee, L. https://biorender.com/lywm40r (2026).

Extended Data Fig. 3 |. Automated PASTA robustly amplifies repeatedly on the same cells and slides.

(A) Automated PASTA showing side-by-side unamplified and PASTA signal for identical regions on the same tissue. Presentation min-max thresholds have been adjusted to simplify viewing and are not the same for unamplified and PASTA signal for each marker. Scale bars: 25 μm. Plot shows marker expression of unamplified and PASTA amplified signal on the same tissue. N = 185120 cells. Tyramide concentration: 10 μM. (B) Visualization of the same field-of-view for repeated deposition on the same target cell, here CD3 and CD8 on CD8 T cells, using HRP-oligo PASTA with HRP-oligo stripping between each deposition and a different oligonucleotide being deposited for each marker for each cycle. Matched minimum and maximum display thresholds are applied across all PASTA images for each marker. Scale bars: 100 μm. Plots show marker expression of unamplified and PASTA signal. N = 176256 cells.

Extended Data Fig. 4 |. PASTA is compatible with variable tissue section thicknesses and demonstrates specificity in viral nucleic acid detection.

(A) FFPE tonsil tissue of three different sectioning thicknesses (5 µm, 10 μm, 20 μm) on the same slide was treated with automated PASTA for three markers (CD3, CD8, CD11c) after unconjugated antibody integration of CD69. Scale bars: 100 μm. Plots show median marker expression in 9 matched field-of-views (equal size, as equal position as possible) across the three tissues. N = 9 FOVs.(B) Same field of view and markers as in Fig. 1D but with adjusted threshold for CD3 to demonstrate all vDNA or vRNA/protein positive cells are T cells (orange arrowhead). Scale bars: 100 μm (main images), 25 μm (insets). (C) SIV vDNA and vRNA detection with PASTA is specific as no staining is seen on the lymph node tissue of an SIV-negative animal. Scale bars: 100 μm (main images), 25 μm (insets).

Extended Data Fig. 5 |. PASTA enables multi-modal integration.

(A) Viral nucleic acid detection with PASTA. Left panels show EBV episomal DNA detection in PCNSL tissue (yellow), followed by simultaneous protein phenotyping (CD20 in blue, CD3 in magenta, CLDN5 in red). Right panels show SARS-CoV-2 spike RNA detection in infected rhesus macaque lung tissue (yellow) with protein marker co-staining (CD20, S100A9, Cytokeratin). In uninfected animal tissue, no viral RNA and few infiltrating immune cells were detected; in infected animals, cells positive for viral RNA were also positive for spike protein (SARS-CoV-2 panels). Scale bar: 50 μm. (B) Robustness of the extended multiplexing using PASTA. Left: The two probe sets were used separately with fluorophores to confirm the expected staining. Scale bar: 25 μm. (C) Extended-plex RNAScope is compatible with combined protein staining for high-plex imaging using CODEX conjugated antibodies. CODEX staining for nuclear marker Ki67 and membrane marker NaKATPase is shown on the same slide as 8-plex RNAScope in Fig. 1E. Scale bar: 25 μm. (D) Integration of unconjugated antibodies into multiplexed imaging. Sequential application workflow (left) shows how primary antibodies from different host species (mouse anti-A2AR, rabbit anti-CD73, biotinylated anti-PD-L1) are applied with HRP-conjugated secondary antibodies, with HRP inactivation between steps to prevent cross-reactivity. Comparative imaging of basal ganglia, tonsil, and colorectal adenocarcinoma demonstrates consistent staining patterns between traditional IHC (top row) and PASTA-enabled IHC-style signal integration (bottom row), validating specificity while enabling higher multiplexing capability. Scale bars: 250 μm (A2AR main images), 100 μm (A2AR insets), 100 μm (other main images), 25 μm (other insets). (E) Proximity ligation assay (PLA) compatibility with PASTA and spatial proteomics. Schematic (left) shows the workflow for detecting CD4 interactions in antigen-presenting cells, including target recognition by primary antibodies, PLA probe binding, rolling circle amplification, and HRP recruitment for signal enhancement. Images show nuclear staining (blue), CODEX imaged HLA-DR expression (magenta), CD4 PLA signal (yellow), and a merged view demonstrating the spatial relationships between these markers. Scale bars: 100 μm (main images), 10 μm (insets). Schematics in d and e created in BioRender; Parmelee, L. https://biorender.com/lywm40r (2026).

Extended Data Fig. 6 |. Quality control for CODEX and PASTA staining and annotations.

(A) Marker staining quality control for Fig. 2. Marker of interest shown in white with nuclear signal in blue. The left region is taken from the edge of a germinal center in tonsil tissue. The right region is taken from a myeloid-rich region in the tonsil tissue. All images are taken from the PASTA slide. Scale bars: 100 μm. (B) Heatmaps of annotation marker expression by cell types by disease and slide. Bar graphs on top of heatmap show cell count. Z-scores are calculated row-wise for each marker. N = 225564 (PASTA, Hodgkin’s), N = 279200 (PASTA, DLBCL), N = 1031410 (PASTA, LN/Tonsil), N = 225077 (CODEX, Hodgkin’s), N = 267655 (CODEX, DLBCL), N = 1029007 (CODEX, LN/Tonsil).

Extended Data Fig. 7 |. Cell type annotation phenotype maps.

(A) Complete phenotype maps of cell type annotations for all cores and diseases for both slides. The inset region can be found in Supplementary Note 2. Scale bars: 500μm.

Extended Data Fig. 8 |. PASTA Robustness and Compatibility.

(A) Heatmap of correlation coefficients of cell type annotation proportions of total cells between PASTA and CODEX slides by disease and cell type. Values shown are Pearson’s r. N = 4 cores (Hodgkin’s), N = 5 cores (DLBCL), N = 5 cores (LN/Tonsil). (B) Automated signal amplification of conjugated CD3 before imaging shown in a representative core/region for each disease. Gradients show signal thresholds used for image processing. Plot shows median expression of unamplified (conjugated) and PASTA amplified CD3 in T cell populations (CD4T, CD8T, Treg) where each dot represents one core. Scale bars: 500 μm (main images), 25 μm (insets). N = 4 cores (Hodgkin’s), N = 5 cores (DLBCL), N = 5 cores (LN/Tonsil).

Extended Data Fig. 9 |. Correlation of in situ hybridization and protein signal using PASTA.

(A) Correlation plots of CXCR5-RNA and CXCR5 protein expression separated by cell types and disease. Each dot is one cell. R² based on Pearson’s correlation coefficient. p values based on two-sided t-test.

Extended Data Fig. 10 |. Analysis of PASTA integration of in situ hybridization and conjugation-resistant antibodies.

(A) RNAScope detection of CXCL9-RNA by different myeloid cell populations (DCs, M1-Like, M2-Like) in different disease states. Plot shows median expression where each dot represents one core. Scale bars: 500 μm (main images), 100 μm (insets). N = 4 cores (Hodgkin’s), N = 5 cores (DLBCL), N = 5 cores (LN/Tonsil). (B) Confirmation of conjugation-resistant antibodies (CD69, CTLA-4) staining T cell populations (CD4T, CD8T, Treg) in different disease states. Scale bars: 25 μm.

Supplementary Material

The online version contains supplementary material available at https://doi.org/10.1038/s41592-026-03200-z.

Data availability

The data in this Brief Communication are available via Zenodo at https://doi.org/10.5281/zenodo.15114762 (ref. 25).

Code availability

All code used in this Brief Communication for the plotting of image quantification and phenotype maps, and a template for the automation code for the OT-2, is available via GitHub at https://github.com/SizunJiangLab/PASTA.