Quantitative Spatial Profiling of Antibody Delivery and Target Engagement Using Optically Labeled Antibodies

Visual Abstract

Abstract

Visualization of drug distribution in human tumors is critical for optimizing biologic therapeutics, yet current clinical workflows cannot colocalize drugs and biomarkers on the same formalin-fixed, paraffin-embedded section. Methods: We present a reproducible workflow that integrates near-infrared imaging of optically labeled therapeutic antibodies with multiplex immunostaining and computational registration to enable quantitative, single-cell analysis. Results: Using panitumumab–IRDye800CW, we demonstrated heterogeneous accumulation in epidermal growth factor receptor–positive regions and quantified penetration gradients relative to tumor architecture. Conclusion: We established a multiplex imaging platform that enables single-cell analysis of therapeutic antibody delivery and pharmacodynamics in human formalin-fixed, paraffin-embedded tissues.

The importance of identifying the optimal biologic dose in the development of antibody-based therapeutics has been increasingly recognized, prompting the Food and Drug Administration to launch Project Optimus, an initiative aimed at improving dose optimization for targeted therapies (1,2). This program advocates a shift from the traditional paradigm of maximum tolerated dose to optimal biologic dose, defined as the minimal dose that achieves maximal biologic effect, and emphasizes the integration of pharmacodynamic biomarkers to guide dose selection (1,2). However, despite this paradigm shift, appropriate clinical trial methodologies to reliably identify optimal biologic dose have yet to be established.

To address this challenge, we propose direct visualization of drug distribution using optically labeled therapeutic antibodies (OLTAs) (3–5). Without this labeling before systemic administration, it is difficult to determine whether lack of therapeutic efficacy arises from inadequate drug delivery to tumor cells or from intrinsic drug resistance, and accurate assessment of intratumoral receptor saturation remains elusive. Although fluorescently labeled proteins have been widely used in preclinical studies (6–8), OLTAs have not been broadly adopted for clinical drug evaluation. A major barrier has been the technical difficulty of multiplex visualization of OLTAs and protein biomarkers in formalin-fixed, paraffin-embedded (FFPE) tissue, the standard specimen type in clinical pathology. Specifically, receptor-bound OLTAs are largely lost during antigen retrieval required for immunofluorescence, and tissue autofluorescence in FFPE sections obscures true fluorescent signals, limiting quantitative analysis. Consequently, prior approaches have relied on chromogenic 3,3′-diaminobenzidine staining on serial sections, precluding single-cell-level correlation and precise quantification (3,4).

These limitations can be overcome by separating the acquisition of OLTA and immunofluorescence images and integrating them using computational image registration (9), together with tyramide signal amplification to enhance signal intensity (10). In this study, we established an integrated workflow based on this approach, enabling single-cell-level, quantitative analysis of OLTAs and protein biomarkers on standard FFPE human tumor sections.

MATERIALS AND METHODS

Patient and Panitumumab–IRDye800CW (pan800) Administration

A 68-y-old man with base-of-tongue squamous cell carcinoma who received a 50-mg intravenous infusion of pan800 and underwent surgical resection 42 h later was analyzed as a representative case. Tumor tissue was processed as FFPE sections for pathologic evaluation. This case was derived from a fluorescence-guided surgery clinical trial (NCT05945875) approved by Vanderbilt University Medical Center’s institutional review board (230327).

FFPE Processing, Multiplex Imaging, and Image Registration

Resected tumor tissues were processed into FFPE blocks using standard histology procedures and sectioned at a thickness of 5 µm. Sections were deparaffinized under conditions optimized to preserve near-infrared fluorescence, stained with a nuclear counterstain, and imaged for pan800 before antigen retrieval using a VS200 whole-slide scanner (Evident Scientific) at x20 magnification. After pan800 imaging, cover slips were removed by incubation in phosphate-buffered saline, and multiplex immunofluorescence staining was performed using the Opal 3-Plex Detection Kit (NEL820001KT; Akoya Biosciences) with tyramide signal amplification for epidermal growth factor receptor (EGFR), α-smooth muscle actin (αSMA), and CD31 on an automated staining platform. Protein marker images were acquired under matched imaging conditions. pan800 and protein marker images were imported into QuPath, where regions of interest (ROIs) were defined, exported, and spatially registered using the Warpy algorithm (9). The resulting coregistered images were used for subsequent quantitative analyses. A detailed step-by-step protocol, including practical tips and notes for all procedures, is provided in the Supplemental Protocol, available at http://jnm.snmjournals.org.

Precision of Image Registration

Registration accuracy was assessed by measuring Euclidean distances between corresponding nuclear centroids after overlaying nuclei channels from pan800 and immunofluorescence images. Cells were detected in QuPath, centroid coordinates were exported in GeoJSON format, and nearest-neighbor distances were computed in R version 4.4.2 using the sf package (Supplemental Script 1).

pan800 Signal in EGFR-Positive and -Negative Tumor Regions

Merged images were analyzed in 5 1-mm² ROIs selected in QuPath. Tumor tissue was segmented into EGFR-positive and -negative regions by pixel classification. Background-subtracted pan800 signal was calculated for each region and compared using a paired t-test.

pan800 Signal in EGFR-Positive and -Negative Tumor Cells

Cells were detected using the positive cell detection function in QuPath and classified as EGFR-positive or -negative using an EGFR mean fluorescence intensity threshold of 1200. Background-subtracted pan800 signal was extracted for each cell and compared between groups using the Mann–Whitney U test.

pan800 Signal and Distance from Vasculature

CD31-positive regions were identified by pixel classification and annotated as vasculature. For EGFR-positive cells, the pan800 signal and shortest Euclidean distance from the cell centroid to the vasculature annotation were calculated using a QuPath script (Supplemental Script 2). Associations were evaluated by simple linear regression in Prism (GraphPad Software).

pan800 Signal and Distance from Tumor Nest

EGFR-positive regions identified by pixel classification were subdivided by distance from the tumor margin into 4 zones (0–20, 21–40, 41–60, and 61–80 µm; Supplemental Script 2). The pan800 signal was normalized to the 0- to 20-µm zone and compared across distance groups using the Dunn multiple comparisons test.

Statistical Analysis

All statistical analyses were performed using Prism version 10.6.1. A P value of less than 0.05 was considered statistically significant.

RESULTS

Spatial Mapping of pan800 and Tissue Architecture

Multiplex imaging of pan800, immunofluorescence markers (EGFR, αSMA, and CD31), and hematoxylin and eosin staining from the same tissue section demonstrated pan800 localization relative to tumor cells, stroma, and vasculature (Fig. 1A). Image registration achieved accurate single-cell alignment between pan800 and immunofluorescence images, with a median centroid drift of 1.16 µm (interquartile range, 0.76–1.74 µm), corresponding to less than 10% of the median cell diameter (Fig. 1B), supporting reliable cell-level analysis.

FIGURE 1.

Representative FFPE sections showing OLTA localization with protein markers. (A) Coregistration and merging visualize pan800 distribution relative to tumor cells (EGFR), fibroblasts (αSMA), and vasculature (CD31) within the same section; hematoxylin and eosin (H&E) confirms fluorescence accuracy. (B) Postregistration nuclei align at cellular level, with minimal nuclear drift, indicating high registration accuracy. IF = immunofluorescence.

pan800 Distribution in EGFR-Positive and -Negative Regions

Tumor tissue was segmented into EGFR-positive and -negative regions (Supplemental Figs. 1A–1C). The pan800 signal was significantly higher in EGFR-positive regions than in EGFR-negative regions (Supplemental Fig. 1D), although the difference was modest (median, 2.2-fold; range, 1.4-fold to 3.7-fold), indicating substantial intratumoral heterogeneity.

Single-Cell Analysis Reveals Heterogeneous pan800 Uptake Among EGFR-Positive Cells

Single-cell analysis across 5 ROIs (median, 6982 cells/ROI) showed significantly higher pan800 signal in EGFR-positive tumor cells than in EGFR-negative cells (Fig. 2D). EGFR-positive cells exhibited a bimodal pan800 distribution, whereas EGFR-negative cells showed a unimodal pattern. Using a pan800 intensity cutoff of 100, a median of 48.8% (range, 39.2%–72.4%) of EGFR-positive cells were pan800-positive, indicating heterogeneous drug delivery. A small subset of EGFR-negative cells displayed a high pan800 signal, suggesting possible off-target uptake (Figs. 2C and 2D).

FIGURE 2.

pan800 signal in EGFR-positive and -negative cells. (A) Microscopic EGFR signal distribution. (B) Computational classification of cells as EGFR-positive or -negative. (C) Overlay of pan800 signal on cell annotations from (B). (D) Comparison of pan800 signal across 5 ROIs using Mann–Whitney U test. ****P < 0.0001.

pan800 Signal Shows Weak Association with Distance from Vasculature

Because intratumoral antibody delivery is generally expected to positively correlate with blood flow, we analyzed the relationship between distance from the vasculature and pan800 signal. Analysis of EGFR-positive cells revealed a statistically significant but weak negative correlation between the pan800 signal and distance from the vasculature across all ROIs (median r, 0.098; r range, 0.037–0.14), indicating minimal dependence of pan800 distribution on vascular proximity in this case (Fig. 3D; Supplemental Fig. 2).

FIGURE 3.

Relationship between pan800 signal and distance from vasculature. (A) EGFR expression and CD31-positive vasculature in representative tumor section. (B) pan800 signal overlaid with nuclei in same region. (C) Composite image showing pan800 in EGFR-positive tumor cells with annotated CD31-positive vasculature. (D) Scatter plot of pan800 signal versus distance from vasculature for individual EGFR-positive cells. Red line indicates linear regression.

pan800 Signal Intensity as a Function of Distance from Tumor Nest

Given that pan800 signal heterogeneity among EGFR-positive cells may depend on their spatial location, we examined the relationship between pan800 intensity and distance from the tumor–stroma border. EGFR-positive regions were stratified by distance from the tumor nest margin (0–20, 21–40, 41–60, and 61–80 µm; Fig. 4). The pan800 signal decreased progressively with increasing distance, with a reduction of approximately 50% observed just 20 µm inward from the tumor periphery (Fig. 4D), consistent with limited intratumoral penetration attributable to a binding-site barrier (11,12).

FIGURE 4.

pan800 signal and distance from p tumor nest periphery. (A) Microscopic distribution of EGFR signal. (B) Computational classification of EGFR-positive and p-negative regions. (C) EGFR-positive regions stratified by distance from EGFR-negative areas, with pan800 signal visualized at 20-µm intervals. (D) pan800 intensity versus distance from tumor nest margin across 5 regions, normalized to the 0- to 20-µm zone. Points and bars indicate median and IQR using Dunn multiple comparisons test. *P = 0.0429; ***P = 0.0007.

DISCUSSION

Here, we demonstrated that single-cell-level correlation of OLTAs with protein biomarkers can be achieved on standard FFPE sections. By integrating established techniques—including sequential imaging, signal amplification, and computational image registration—we created a highly reproducible workflow.

Using this approach, we demonstrated that OLTAs selectively accumulate within tumor regions; however, their intratumoral distribution was highly heterogeneous. Quantitative analysis revealed that OLTA penetration was limited to approximately 20 µm from the tumor nest periphery, consistent with the influence of a binding-site barrier. In addition, the correlation between the OLTA signal and distance from the vasculature was unexpectedly weak, suggesting that vascular proximity alone does not fully explain intratumoral drug distribution. Because FFPE tissue is the gold standard in human pathology and we have previously found that fixation has a limited impact on OLTA signal intensity with minimal information loss (13), this platform provides a practical and robust framework for analyzing the distribution of antibodies in human tissue. Moreover, beyond drug delivery, the cellular resolution of this method enables spatial correlation of OLTAs with pharmacodynamic biomarkers, including receptor phosphorylation, proliferation, and apoptosis, thereby supporting biologic activity assessment in window-of-opportunity trials (14).

Several limitations should be acknowledged. Although we used routine processes for this analysis, the reproducibility and broader applicability of this platform will require validation across multiple antibodies and additional patient samples. Second, the pan800 dose evaluated in this study was below the therapeutic dose and assessed several days after systemic administration; evaluation at clinically relevant treatment doses and at multiple time points will be necessary to fully interpret the biologic significance of the observed distribution patterns. Third, this workflow is primarily applicable to specific research settings—such as window-of-opportunity trials—where tissue sampling after OLTA administration is feasible.

CONCLUSION

We established a multiplex imaging platform that enables single-cell analysis of therapeutic antibody delivery and pharmacodynamics in human FFPE tissues.

DISCLOSURE

This work was supported by grant R01CA266233 from the National Institutes of Health. Hidenori Tanaka reports support from the Japanese Society of Otorhinolaryngology Head and Neck Surgery 2024 Overseas Study Support Program and the Society for Promotion of International Oto-Rhino-Laryngology (SPIO), Japan, through the SODA TOYOJI SPIO Scholarship 2023 (SS23001). No other potential conflict of interest relevant to this article was reported.

KEY POINTS

QUESTION: Can antibody therapeutics and protein markers be simultaneously visualized and quantitatively analyzed in human FFPE sections?

PERTINENT FINDINGS: A reproducible pipeline integrating OLTA imaging, multiplex immunostaining, and image registration enabled single-cell spatial analysis in human FFPE tissue.

IMPLICATIONS FOR PATIENT CARE: This scalable approach supports assessment of drug delivery and pharmacodynamics in clinical specimens, facilitating optimization of antibody therapeutics in window-of-opportunity trials.