Engineering and evaluation of precision-glycosylated clickable albumin nanoplatform for targeting the tumor microenvironment

Abstract

Rationale: Glycosylation of drug delivery vehicles enables selective tumor microenvironment (TME) targeting but is limited by the lack of precise glycan control and unbiased evaluation of in situ targeting. We developed a clickable albumin nanoplatform engineered by distinct glycosylation for selective in vivo cell targeting (CAN-DGIT) with a defined number of sugar moieties and integrated spatial transcriptomics (ST) to map nanoparticle-TME interactions. Methods: Albumin was functionalized with azadibenzocyclooctyne (ADIBO) at a controlled degree of functionalization (DOF), confirmed by MALDI-TOF and UV-vis spectroscopy, followed by conjugation of azide-functionalized mannose, galactose, or glucose via click chemistry. Nanoparticles were labeled with ⁶⁴Cu or fluorescent dyes for PET imaging and ex vivo analysis in healthy and 4T1 tumor-bearing mice. ST based algorithms, spatial gene-image integration (SPADE), cell-type deconvolution (CellDART), and image-based molecular signature analysis (IAMSAM), were used to define TME clusters, associated cell populations, and glycan receptor gene signatures. Clodronate-loaded glycosylated albumins were tested for tumor-associated macrophage (TAM) depletion. Results: Glycosylation type of CAN-DGIT dictated pharmacokinetics and targeting. Mannosylated albumin (Man-Alb) showed rapid hepatic retention via mannose receptors on Kupffer cells and TAMs; galactosylated albumin (Gal-Alb) exhibited rapid hepatobiliary clearance with the highest tumor-to-liver ratio; glucosylated albumin at the C6 position (Glc(6)-Alb) progressively accumulated in tumors, correlating with glucose transporter 1 (GLUT1)-expressing cancer cells. ST confirmed Man-Alb enrichment in extracellular matrix (ECM)/TAM-rich clusters (mannose receptor C-type 1, Mrc1-high) and Gal-/Glc-Alb uptake in glycolytic/hypoxic tumor clusters (Slc2a1-high). Man-Alb-clodronate achieved potent CD206+ TAM depletion without altering drug release kinetics. Conclusions: Precisely tuned glycosylation enables programmable biodistribution and cell-type targeting of albumin nanoparticles in the TME. Integrating PET with ST provides a robust framework for mechanistic mapping of nanomedicine uptake. The CAN-DGIT platform offers a versatile strategy for developing targeted theranostic agents with immunomodulatory potential.

Keywords: albumin; click chemistry; glycosylation; positron emission tomography; spatial transcriptomics; theranostics; tumor microenvironment.

Figure 1

Construction flow of the nanoplatform for this study and the validation of distinct glycosylation based on positron emission tomography imaging. (A) Schematic flow of CAN-DGIT for confirming the efficient targeting property. (B) Representative PET images of ⁶⁴Cu labeled distinct glycosylated albumin in normal mice following IV administration at various time points (0, 2, 4, 24, and 48 h; n = 4 for each group). The yellow arrow indicates the gallbladder (GB). (C) Time-activity curve of glycosylated albumin in blood and liver. (D) Ex vivo biodistribution of glycosylated albumin at 24 h after injection. All quantification was presented as %ID/g ± SD (**: P < 0.001).

Figure 2

Micro-specificity of mannose to Kupffer cell and galactose to hepatocytes in liver tissue. (A) After simultaneous vascular injection of Man-Alb and Gal-Alb, liver tissue fluorescence imaging acquisition. Schematic figure of targeting for hepatocytes and Kupffer cells. (B) Ex vivo confocal Imaging of Man, Gal-Alb with different fluorescence filters. Upper panel: depicts the periportal region, emphasizing Kupffer cells enriched in the sinusoidal area adjacent to the portal venule. Lower panel: illustrates the liver parenchyma showing Kupffer cells residing within the sinusoidal lumen along hepatocytes. Flamma fluor 648-conjugated mannose albumin and Flamma Fluor 488-conjugated galactose albumin nanoparticles are scanned with appropriate ex/em filters (Red = Flamma fluor 648, Green = Flamma Fluor 488). (C) IVIS imaging of each type of glycosylated albumin after 24 IV injections. (D) Tumor-to-organ ratio of image-based quantification using IVIS. Abbreviations: Li, liver; Ki, kidney; Tu, tumor; He, heart; Lu, lung; Sp, spleen; In, intestine. Error bars represent mean ± SD.

Figure 3

Summary of ST library of Man+Gal sample. (A) The images were derived from spatial transcriptomic (ST) data of the Man+Gal sample. Each image represents the H&E image (upper left), the UMI counts provided by SpaceRanger (upper left), and the spatial clustering analysis of the Man+Gal sample (below). UMAP projections of spatial clusters after data integration with the Man+Gal sample by using the Seurat pipeline (left) and the distribution of spatial clusters according to sample (right). (B) Fluorescence images of Man+Gal sample. Flamma fluor 648-conjugated mannose albumin and Flamma Fluor 488-conjugated galactose albumin nanoparticles are scanned with appropriate ex/em filters (left). Additionally, they mapped with ST spots using the SPADE algorithm (right). (C) Relative fluorescence signals of each cluster mapped by SPADE algorithm to ST libraries according to the albumin nanoplatform. Error bars represent mean ± SD. (D) Mean intensities of min-max normalized albumin fluorescence signals in each cluster. P-values between mannose and galactose albumin fluorescence signals. (E) CellDART results for the Man+Gal sample. The original 4T1 scRNA-seq reference for CellDART execution contained only nine cell types, among which the original 'Anti-inflammatory_macrophages' was divided into the newly defined 'Anti-inflammatory_macrophages' and 'Tumor-associated_macrophages' for non-TAM-like cells and TAM-like cells, respectively.

Figure 4

Analysis of high CAN-DGIT uptake regions of Man+Gal sample. (A) IAMSAM analysis of relatively high mannose albumin uptake region. The characteristic uptake region of mannose albumin was pointed as a gray rectangular box (upper left). When this region is applied in a segment-anything model, distinct albumin uptake patterns in the region are segmented (upper right). Additionally, differentially expressed genes in this ROI were listed and used for gene set enrichment analysis (below). (B) The same analysis was performed in the galactose albumin distribution image. (C) Table of log fold change and adjusted p-values for comparing gene expression levels of each glycan-related gene between segmented ROI from IAMSAM and the rest region. (D) Expressions of mannose receptor, glucose transporter, and asialoglycoprotein receptor-related genes.

Figure 5

In vivo and ex vivo biodistribution of ⁶⁴Cu radiolabeled distinct glycosylated albumins in the 4T1 tumor model. (A) Representative PET imaging of 4T1-bearing mice using Albumin, Man-Alb, Glc(2)-Alb and Glc(6)-Alb following IV administration at various time points (0, 4, 8, and 24 h; n = 4 for each group). (B) Ex vivo biodistribution of ⁶⁴Cu radiolabeled distinct glycosylated albumins in tumor models measured using a gamma counter at 24 h after injection (n = 4 for each group). (C) Representative illustration of in vivo cell targeting in cancer region. The cells highlighted in red represent cancer cells, the green ones are tumor-associated macrophages (TAMs), and the purple ones indicate inflammatory macrophages. D. Time-activity curve of glycosylated albumin in blood, liver, and tumor. Error bars represent mean ± SD.

Figure 6

Summary of ST library of Man+Glc sample. (A) The images were derived from spatial transcriptomic (ST) data of the Man+Glc sample. Each image represents the H&E image (upper left), the UMI counts provided by SpaceRanger (upper left), and the spatial clustering analysis of the Man+Glc sample (below). UMAP projections of spatial clusters after data integration with the Man+Glc sample using the Seurat pipeline (left) and the distribution of spatial clusters according to sample (right). (B) Fluorescence images of the Man+Glc sample. Flamma fluor 648-conjugated mannose albumin and Flamma Fluor 488-conjugated glucose albumin nanoparticles are scanned with appropriate ex/em filters (left). Additionally, they mapped with ST spots using the SPADE algorithm (right). White arrows indicate unreliable FL signals, but yellow arrows indicate seemingly biologically meaningful FL signals. (C) Relative fluorescence signals of each cluster mapped by SPADE algorithm to ST libraries according to albumin nanoplatform. Error bars represent mean ± SD. (D) Mean intensities of albumin fluorescence signals in each cluster. P-values between mannose and glucose albumin fluorescence signals. (E) CellDART results for the Man+Glc sample. The original 4T1 scRNA-seq reference for CellDART execution contained only nine cell types, among which the original 'Anti-inflammatory_macrophages' was divided into the newly defined 'Anti-inflammatory_macrophages' and 'Tumor-associated_macrophages' for non-TAM-like cells and TAM-like cells, respectively.

Figure 7

Analysis of high CAN-DGIT uptake regions of the Man+Glc sample. (A) IAMSAM analysis of relatively high mannose albumin uptake region. The characteristic uptake region of mannose albumin was pointed as a gray rectangular box (upper left). When this region is applied in a segment-anything model, distinct patterns of albumin uptake in the region are segmented (upper right). Differentially expressed genes in this ROI were listed and used for gene set enrichment analysis (below). (B) The same analysis was performed in the glucose albumin distribution image. (C) Table of log fold change and adjusted p-values for comparing gene expression levels of each glycan-related gene between segmented ROI from IAMSAM and the rest region (D) Expressions of mannose receptor, glucose transporter, and asialoglycoprotein receptor-related genes.

Figure 8

Evaluation of the efficacy of the CAN-DGIT approach based on ST analysis in an animal model. (A) Each glycosylated albumin complexed with clodronate, as configured in Figure 1A, to form respective complexes. Drug incorporation and release were assessed through release tests. Additionally, each complex was administered to 4T1-bearing mice to evaluate specific cell targeting in tumor tissues confirmed through ST analysis. (B) Releasing profile of the Man-Alb/clodronate drug complex. (C) Percentages of TAMs (CD45⁺CD11b⁺Ly6g^-F4/80⁺; left) and CD206⁺TAMs (right) in 4T1-bearing mice treated with either Alb/clodronate complex (n = 6 mice), Man-Alb/clodronate complex (n = 4 mice), Gal-Alb/clodronate complex (n = 6 mice), Glc-Alb/clodronate complex (n = 5 mice) or albumin control (n = 5 mice). All data represented as mean ± S.E.M. Statistical significance was determined by two-tailed t-tests. *P < 0.05, **P < 0.01. N.S., nonsignificant.