Non-invasive imaging with ICOS-targeting monoclonal antibody for preclinical diagnosis of rheumatoid arthritis in a humanized mouse model

Abstract

Background

Activated T cells play a pivotal role in rheumatoid arthritis (RA) pathogenesis, and imaging of activated T cells may provide a non-invasive tool for RA detection. Here, we first developed an optical probe targeting human inducible T cell co-stimulator (ICOS) and tested its capacity in RA diagnosis by capturing ICOS⁺ activated T cells in vivo in a humanized mouse model.

Methods

The humanized arthritis model, Human peripheral blood mononuclear cells- adjuvant induced arthritis (HuPBMC-AIA) was established, and flow cytometry and immunofluorescence were employed to determine ICOS expression in huPBMC-AIA model. Anti-human ICOS monoclonal antibody (mAb) was conjugated to Cy7 via NHS ester amine reaction. A cell uptake study was used to confirm the specificity of Cy7-ICOS mAb to activated T cells. 4-view near-infrared fluorescence (NIRF) imaging study was performed to test Cy7-ICOS mAb in detecting RA in vivo.

Findings

ICOS was confirmed as an indicator of RA pathogenesis via RNA-seq, flow cytometry and immunofluorescence data. An in-vitro cellular uptake study validated the specificity of Cy7-ICOS mAb to activated T cells. Cy7-ICOS mAb could detect ICOS⁺ activated T cells in vivo through 4-view NIRF imaging. The receiver operating characteristic (ROC) curve created based on NIRF imaging quantification could distinguish the huPBMC-AIA group from the control group at all time points imaged.

Conclusion

In this study, we first developed an optical imaging probe targeting human ICOS, Cy7-ICOS mAb. The 4-view NIRF imaging with Cy7-ICOS mAb could detect pathogenic ICOS⁺ activated T cells with high sensitivity and specificity in vivo, which indicated the great potential of this imaging probe in RA early diagnosis.

Highlights

the first to target human ICOS, a critical aspect in advancing towards humanized immunology.

ICOS as a promising biomarker for early RA detection.

the validity of the huPBMC-AIA model in mimicking the complex microenvironment of RA.

Context and significance.

Our findings offer a paradigm shift in the approach to RA research, suggesting that humanized models, particularly those targeting ICOS, provide a more clinically relevant platform for early diagnosis. The simplicity of utilizing a single biomarker for imaging in the complex microenvironment of RA holds promise for facilitating easier and more effective diagnosis, thus potentially improving RA patient management.

Significance

Our findings offer a paradigm shift in the approach to RA research, suggesting that humanized models, particularly those targeting ICOS, provide a more clinically relevant platform for early diagnosis. The simplicity of utilizing a single biomarker for imaging in the complex microenvironment of RA holds promise for facilitating easier and more effective diagnosis, thus potentially improving RA patient management.

Graphical abstract

Introduction

RA is a chronic autoimmune condition characterized by persistent debilitating joint inflammation [1]. Early diagnosis and timely intervention may lead to better prognosis of rheumatoid arthritis patients [2]. Due to the low sensitivity and specificity, current approaches often fail in RA early detection [3]. For instance, anatomic imaging techniques, can only diagnose RA when cortical bone destruction has occurred. On the other hand, blood tests of RA-specific autoantibodies, such as rheumatoid factor (RF) and anti-cyclic-citrullinated protein antibodies (ACPA), often lack specificity [4]. To overcome these challenges, there is an urgent need to develop novel tools for RA early detection.

Autoreactive T cell activation gone awry has been proposed as the leading cause of autoimmune disorders, such as multiple sclerosis (MS) and rheumatoid arthritis (RA). During RA pathogenesis, CD4⁺ T cells stimulate osteoclast and macrophage activation, autoantibody synthesis, lymphoid neogenesis, and pannus development in the inflammatory joint [5]. It is thought that activation of major histocompatibility complex (MHC) class II-restricted autoreactive T cells, which specifically assist B cells in producing autoantibodies, is the cause of RA years before the disease manifests clinically [6]. Moreover, T-cell activation is enriched in synovial tissue, which predates the onset of RA [7]. As consequently, techniques to precisely identify and monitor activated T cell reactions in vivo may assist with RA diagnosis, staging, and treatment monitoring. Molecular imaging has shown great potential for achieving this goal [8]. A nucleoside analogue, 2′-deoxy-2′-[18 F]fluoro-9-β-D-arabinofuranosylguanine ([18 F]F-AraG), was tested in RA detection in a murine AIA-induced arthritis model [9]. Despite the high uptake of activated T cells, [18 F]F-AraG also accumulated in other immune cells, such as macrophages and dendritic cells. This unexpected result may be attributed to the same nucleic acid metabolic pathways shared by immune cells [10].

Thus, we diverted our attention to T-cell surface costimulatory receptors. In our previous study, we developed an NIRF imaging probe targeting murine 4-1BB, and we successfully diagnosed RA pathogenesis at early stages via in vivo activated T cell tracking [11], which demonstrated the feasibility of imaging T-cell surface costimulatory receptors in RA early diagnosis. In consideration of further clinical translation, as well as to select the most sensitive biomarker for RA imaging, we analyzed RNA-seq data from the SRA database, and inducible T cell co-stimulator (ICOS or CD278) was chosen for this study [12]. ICOS is a member of the CD28 superfamily, which is mainly expressed on activated T cells. The ligation of ICOS and its ligand (ICOSL) could activate the downstream pathway and regulate T-cell proliferation, as well as cytokine secretion [13]. It had been proved as an ideal biomarker for in vivo activated T cell tracking in many murine models [14,15,16], whereas no humanized model or clinical studies were reported. Here, we first assessed human ICOS imaging in a humanized rheumatoid arthritis model. ICOS expression was validated via flow cytometry and immunofluorescence staining. NIRF probes have great potential for early detection and diagnosis of liver malignancies and thyroid cancer [17, 18]. Similarly, NIRF imaging is becoming a noninvasive alternative to radionuclide imaging in joints of small animals [19]. An optical imaging probe targeting human ICOS, Cy7-ICOS mAb was synthesized and evaluated via both in vitro cell uptake and in vivo NIRF imaging studies. Our findings demonstrated that ICOS targeted NIRF imaging is a promising strategy for rheumatoid arthritis early diagnosis.

Materials and methods

Data source

Raw RNA-seq data for Gene Expression Omnibus (GEO) datasets GSE89408 and GSE90081 was downloaded from the SRA database and quantified to gene-level counts using the ARCHS4 pipeline [20]. Expression profiling by array data for GEO datasets GSE82107, GSE77298 was downloaded from GEO. The platform GPL570 (HG-U133_Plus_2) The Affymetrix Human Genome U133 Plus 2.0 Array was utilized for annotation. Series included 57 early RA samples and 54 normal samples, were selected as the validation sets. Gene counts were downloaded from the ARCHS4 gene expression matrix v5.For more information about ARCHS4, as well as free access to the quantified gene expression matrix, visit the project home page at the following URL: http://amp.pharm.mssm.edu/archs4/download.html.

Data normalization

Raw counts were normalized to log10-Counts Per Million by dividing each column by the total sum of its counts, multiplying it by 10⁶, followed by the application of a log10-transform.

Volcano plot analysis

Volcano plot was plotted by https://www.bioinformatics.com.cn (last accessed on 10 July 2023), an online platform for data analysis and visualization. Openbiox Hiplot (ORG) was used to create gene differential expression sequencing plots.

Identification of immune cell-specific expressed genes

To further understand the immune cell-specific expression of these DEGs, the web program BioGPS (http://biogps.org/) was utilized to analyze cell expression. Screening criteria were as follows: transcripts matched to a single organ system with an expression value of > 10 multiples of the median. The genes identified using these criteria were thought to be immune cell-specific genes.

Immune infiltration analysis

The deconvolution method was employed to process marker gene expression values to estimate the proportion of different types of immune cells in the glomerular tissues of both normal and diabetic nephropathy. The CIBERSORT algorithm was used for analysis. These cells comprise 22 different types of immune cells, including M1 and M2 macrophages, plasma cells, static memory CD4 + T cells, γδT cells, and mast cells. For every sample, a boxplot representing the percentage of T cell was created.

Cell culture

Human peripheral blood mononuclear cells (huPBMC) from Charles River (formerly HemaCare). huPBMC was kept at 37 °C in a humidified incubator with 5% CO₂. Harvested during the exponential growth phase, the cells were resuspended at a density of 10 × 10⁶ cells/mL in full growth media.

huPBMC-AIA model

All the animal studies were conducted under protocols approved by the Local Ethical Committee of Harbin Medical University Animal Care and Use. Five-week-old NOG male (Charles River, USA) mice were infused intravenously with 3 × 10⁶ huPBMC. Peripheral blood was collected, and flow cytometry was used to evaluate the immune cell subpopulations to verify the humanization. The humanized mice were included in the Adjuvant induced arthritis (AIA) induction protocol after achieving sufficient immune blood reconstitution (engraftment efficiency (huCD45 > 25%)).

For the humanized arthritis model, the periarticular space surrounding the ankle was injected with 10 µL of either complete Freund’s adjuvant (CFA) 4 mg/mL (Chondrex, Redmond, WA) or PBS. To deplete murine neutrophils, anti-Gr-1 antibody (250 µg) (BP0075, Clone: RB6-8C5, BioXCell, West Lebanon, US) was given intraperitoneally (IP) on days 1 and − 1. Ankle joint processing and arthritis were carried out as previously mentioned [21]. A vernier caliper was used to measure the thickness of the paw.

Bulk RNA-seq analysis and KEGG

Total RNA was extracted from paw of huPBMC-AIA model followed by library preparation according to Illumina standard instruction (VAHTS Universal V6 RNA-seq Library Prep Kit for Illumina^®). Agilent 4200 bioanalyzer was employed to evaluate the concentration and size distribution of cDNA library before sequencing with an Illumina novaseq6000. The protocol of high-throughput sequencing was fully according to the manufacturer’s instructions (Illumina). The raw reads were filtered by Seqtk before mapping to genome using Hisat2 (version:2.0.4). The fragments of genes were counted using stringtie(v1.3.3b) followed by TMM (trimmed mean of M values) normalization. Significant differential expressed genes (DEGs) were identified as those with a False Discovery Rate (FDR) value above the threshold (Q < 0.05) and fold-change > 2 using edgeR software [22]. An online platform (chiplot) was used for the KEGG pathway and Reactome enrichment analyses of DEGs.

Flow Cytometry

Flow cytometric analyses were performed on a flow cytometer CytoFLEX (Beckman Coulter, USA). The peripheral blood, spleen, bone marrow, and paw were processed in the same manner as previously described [21]. To collect single-cell suspensions from ankles with joints were dissected. The joint capsules were sliced to allow enzymatic dissociation. Tissue was digested in RPMI 1640 with 1.5 mg/mL collagenase IV (Worthington), 0.2 mg/mL dispase II (absin), and 0.1 mg/mL DNase I (absin). The samples were incubated three times at 37 °C for 20 min, with continual shaking (180 rpm). red blood cell (RBC) lysis buffer (BioLegend) was used to dissolve total blood, spleen, and bone marrow. The cell suspensions were filtered to less than 40 μm. Single cell suspensions were stained with a mixture of fluorochrome-conjugated antibodies: PerCP/Cyanine5.5 anti-mouse CD45 (1:100, 157208, Clone: S18009F, BioLegend), PE anti-human CD4 (1:100, 344606, Clone: SK3, BioLegend), APC anti-human CD8 (1:100, 344722, Clone: SK1, BioLegend), APC/Cyanine7 anti-human/mouse/rat CD278 (ICOS) (1:100, 313530, Clone: C398.4 A, BioLegend), Pacific Blue™ anti-mouse/human CD11b (1:100, 101224, Clone: M1/70, BioLegend), FITC anti-mouse Ly-6G/Ly-6 C (Gr-1) (1:100, 108405, Clone: RB6-8C5, BioLegend), and PE/Cyanine7 anti-human CD45 (1:100, 304016, Clone: HI30, BioLegend), after being fixed with the Zombie Aqua™ Fixable Viability Kit (1:100, 423102, BioLegend, San Diego, CA). All samples were examined using a high-sensitivity flow cytometer (Beckman Coulter Cytoflex S, Krefeld, Germany). FlowJo software (Treestar Inc, OR) and online platform (Cytobank, Beckmen, US) was used for the analyses.

Conjugation of Cy7-ICOS mAb

The supplier of the in vivo monoclonal antibody (BE0353, clone: C398.4 A) against human ICOS was BioXCell, located in West Lebanon, USA. The Cy7 NHS ester was supplied by APExBIO. Anti-ICOS mAb was diluted in PBS (pH 8.9) to 1 mg/mL. Cy7 NHS ester was dissolved in dimethyl sulfoxide to a final concentration of 5 mg/mL before being mixed with anti-ICOS mAb. The ICOS mAb to Cy7 NHS ester molar ratio was 1:10. The mixture was incubated at room temperature for 1 h. Cy7-ICOS mAb was purified using a Vivaspin2 50 KDa cutoff MWCO spin filter (GE Healthcare, Piscataway, USA). NanoDrop (Thermo Fisher, Waltham, MA, USA) was used to determine the final concentration of Cy7-ICOS mAb.

Cell uptake study

The 500X concentrate of the Cell stimulation Cocktail (Invitrogen, Carlsbad, USA) was added straight to the Iscove’s Modified Dubecco’s Medium (IMDM) at a rate of 1 × (2 uL/mL) to activate T cells. Probe incubation was performed out in 1mL IMDM with 200,000 active, resting, or blocked cells. Following 1-hour incubation at 37 °C, 2 µg of Cy7-ICOS mAb was added after 48 h of stimulation. Before the binding assay, activated T cells were treated with 100 µg of unconjugated ICOS mAb and incubated for 1 h at 37 °C for the blocking group. A multiscan spectrum (Tecan, Switzerland) and a CytoFLEX flow cytometer were used to analyze the data.

In vivo 4-view Near-infrared fluorescence imaging (NIRF) studies

An OPTICAL IMAGING IN VISIBLE & NIR I (Biospace, California, USA) was applied to collect the NIRF imaging data. On the 3rd day following the induction of arthritis, 50 µg of Cy7-ICOS mAb was administered via the mouse tail vein following anesthesia with 2–3% isoflurane gas. On days 3, 4, 5, 6, and 7, NIRF imaging was performed using the following parameters: Acquisition mode: FLI integration (1000 ms per frame − 5 frames) Fluorophore: Cy7, Excitation = 737 nm, Background = 687 nm, Emission = 797 nm, Filter cut off = 780–815 nm (band pass) nm; Lens: 50 mm f/1.2; Aperture = 1.2; Focus = 93.0. Overlaying x-ray images and optical signals is possible with the PhotonIMAGER Optima’s X-ray module. Improved optical signal localization and quantification are provided by the 4-View – 3D module (Acquisition mode: FLI integration (1000 ms per frame − 5 frames Fluorophore: Cy7, Excitation = 737 nm, Background = 687 nm, Emission = 797 nm, Filter cut off = 780–815 nm (band pass) nm Lens: 35 mm f/1.4; Aperture = 1.4; Focus = 94.0). Region of interest (ROI) of NIRF images was drawn using software (M3 vision, Biospace, California, USA).

Ex vivo biodistribution

To validate the ROI data, mice were euthanized after the imaging session on day 7 and subjected to the standard ex vivo biodistribution analysis described above. The following organs were harvested: the heart, liver, lung, spleen, kidney, intestine, right paw (RP), left paw (LP), bone (tibia), muscle, skin, and blood. Parameters for acquisition: 50 mm f/1.2 lens; 1.2 aperture; 96.0 focus.

Immunofluorescence

Immunofluorescence (IF) staining antibodies: Abcam (Boston, USA) provided the IgG H&L (Alexa Fluor^® 488) (ab173003) and IgG H&L (Alexa Fluor^® 594) (ab150108). BioLegend provided a purified anti-human CD3 antibody (317301, Clone: OKT3) and an anti-human/mouse/rat CD278 (ICOS) antibody (313502, Clone: C398.4 A). After the mice were sacrificed, the ankle joints were fixed with 4% paraformaldehyde and decalcified for 40 days with 10% EDTA. After embedding in paraffin After deparaffinization, antigen retrieval, and BSA blockade, ankle joint sections were stained overnight at 4 °C with CD3 or CD278 antibodies. The fluorescent-labeled secondary antibodies were then added and incubated for 1 h. After staining the nuclei with DAPI (Beyotime, Shanghai, China), the sections were examined under a fluorescent microscope (Nikon, Japan).

Safety evaluation

During the procedures, the body weight was recorded. Blood was collected three days after the last administration, and serum was separated by centrifugation. Standard kits were used to measure serum levels of ALT, AST, BUN, and Crea. Major organs (heart, liver, spleen, lung, and kidney) were removed, fixed with 4% paraformaldehyde, and stained with hematoxylin and eosin (H&E). A Light microscope (Nikon, Japan) was then used to image the sections.

Statistical analysis

An online platform for data analysis and visualization, https://www.chiplot.online, was used for data management and analysis. For the statistical analyses, GraphPad Prism 9.0v (GraphPad, CA, USA) software was utilized. The Shapiro-Wilk test was used to determine data normality. This test was used to see if the data distribution varied substantially from a normal distribution. A p-value < 0.05 indicates non-normality. There were only two groups compared using the t test. Three or more groups were subjected to a one-way ANOVA. P-values < 0.05 were regarded as statistically significant.

Results

T cell portion profile and differential expression analysis (DEG) in early rheumatoid arthritis

To select the most sensitive imaging biomarker for RA, the GSE89408 dataset was chosen for T cell portion analysis and DEG identification. It contained 57 early RA and 28 healthy control (HC) samples from synovial tissues. We employed the CIBERSORT to study immune infiltration in this dataset. Compared to HC group, activated CD4⁺ memory T cells (P < 0.001) were found to be increased in early RA group, along with decreased CD4⁺ memory resting(P < 0.05) and CD4 naive T cells (P < 0.001) (Fig. 1A), which indicated that activated T cells are driven factors of rheumatoid arthritis at an early stage [23]. Further DEG data was visualized using volcano plot analyses (Fig. 1B), ICOS was the most differential activated T cell genes in RA samples, compared to other T cell genes (TNFRSF 9, TNFRSF 4, CD69 and CTLA 4). Gene expression sequencing is in light of the critical function of T cells in the early development of RA. We concentrated on the expression changes of CD molecule family and costimulatory factors in differential genes. ICOS belongs to TOP10 up-regulated genes (Supplementary Fig. 1A and 1B). T cells express ICOS specifically, according to the BioGPS database (Supplementary Fig. 1C). To validate the expression of ICOS on activated human T cells, human PBMCs were incubated with cell-stimulating kit (PMA/Iono) for 72 h. According to the flow cytometry study, a significant peak shift could be observed in PMA/Iono group (Fig. 1C). Higher frequency of ICOS could be detected on both CD4 and CD8 T cells, compared to PMA/Iono + blocked, blocked, and resting groups, P < 0.001 (Fig. 1D and E).

Fig. 1

A Bar plots indicating the differences of T cell fractions between the RA and healthy control group (HC). B Volcano plot of differentially expressed genes between early RA samples and normal samples. C and D Flow cytometry analysis of ICOS expression on huCD45 + CD4 + and huCD45 + CD8 + T cells in huPBMCs (containing PMA/Iono, PMA/Iono + blocked, blocked, vehicle). E Flow cytometry gating strategy. One-way ANOVA and t test was conducted for statistical significance. Error bars represent mean ± SEM, ****, p < 0.0001; ***, p < 0.001; **, p < 0.01; *, p < 0.05

Establishment and validation of huPBMC-AIA rheumatoid arthritis model

The animal study scheme is shown in Fig. 2A. Generally, huPBMC was injected into NOG mice via the tail vein to establish the huPBMC-AIA model. The experimental group, huPBMC-AIA, allowing for the study of immune responses in a humanized inflammatory environment. Weekly peripheral blood was taken to confirm humanization using flow cytometry. According to our data (Freq.huCD45(%) -2week: 0.68 ± 1.66; -1 week: 10.34 ± 5.55; day0: 35.61 ± 13.26), the engraftment efficiency reached to a high level after three weeks post human PBMC injection (Fig. 2B and C), which validated the success of NOG mice humanization. The depletion of Gr-1 was carried out at day 1 and − 1 to minimize the intervention of myeloid cells on inflammation [20]. The CD11b⁺Gr-1⁺ cell frequency from in vivo depletion of Gr-1⁺ group was much lower than that in the control group, P < 0.0001 (Fig. 2D). Following CFA injection, the symptoms, such as swelling and redness, could be observed in inflamed joint from huPBMC-AIA group (Fig. 2E). To monitor the severity of joint arthritis, paw thickness was measured every day after CFA injection, and the paw thickness in the huPBMC-AIA model was higher than that in the control group at all time points examined (P < 0.01) (Fig. 2F). To determine whether huPBMC-AIA model could accurately mimic RA pathogenesis, RNA-seq was performed using CFA injected paw. To reveal all the changes (especially immunology-related reactions) and define KEGG disease, we performed transcriptome sequencing on treated paws 3 days after adjuvant injection. huPBMC-AIA mode could promote the expression of specific genes related to RA. (Fig. 2G).

Fig. 2

A The schematic Illustration of animal studies. B Human immune system multilineage development: FACS analysis of the human CD45 + cell fraction in the peripheral blood of humanized mice − 2, -1 and 0 weeks after humanization. C Engraftment efficiency (%graft) between the different groups. %graft was calculated as the proportion of human CD45⁺cells in the total CD45⁺cell population (human and mouse) in peripheral blood. D In-vivo depletion of endogenous Myeloid-derived suppressor cells (MDSCs) with Gr-1 antibody, and the proportion of MDSCs (CD11b and Gr-1 positive) decreased after intervention. E The photographs images of ankles between huPBMC-AIA and control group (0 day, 3 days, 7days). F Average paw thickness in huPBMC-AIA and control group after different treatments, summary of 3 independent replicate experiments (n = 15). G Top 5 enrichment analysis of biological processes in Kyoto Encyclopedia of Genes and Genomes (KEGG) of huPBMC-AIA RNA-seq. t test was conducted for statistical significance. Error bars represent mean ± SEM, ****, p < 0.0001; ***, p < 0.001; **, p < 0.01; *, p < 0.05

ICOS is an indicator of RA pathogenesis at early stage in huPBMC-AIA model

IF staining was conducted on the day3 to determine ICOS expression in injected paw of both control and huPBMC-AIA model. Confocal imaging confirmed positive expression of both CD3 and ICOS in huPBMC-AIA group, whereas no obvious signal could be detected in the control group. Good co-localization of CD3 and ICOS was also observed in huPBMC-AIA group (Fig. 3A). To re-confirm the consistent role of ICOS during RA pathogenesis, flow cytometry study was performed at day7. The most striking finding beyond our thought was that ICOS expression in bone marrow (BM) was much higher in huPBMC-AIA group (huCD45⁺CD4⁺ICOS⁺|Freq.of Parent: P < 0.01; huCD45⁺CD8⁺ICOS⁺|Freq.of Parent: P < 0.05) (Fig. 3B), which indicated CFA injection may also lead to systemic immune response. ICOS⁺CD4⁺ frequency in CFA injected paw from huPBMC-AIA group was considerably higher compared to control group, which was in line with IF staining data. tSNE plot demonstrated ICOS is strict to T cells (Fig. 3C). As an indicator of RA pathogenesis, there was a significant positive correlation between paw thickness and huCD45⁺CD4⁺ICOS⁺ T cells (P value = 0.0001; R square = 0.8548) (Fig. 3D). Principal component/components analysis (PCA) significantly separated animals from the huPBMC-AIA group from other mice throughout PC1, which accounted for 52.4% of the variance (Fig. 3E). The major contributors in identifying mice in the huPBMC-AIA group from other cohorts over PC1 were paw CD4⁺ICOS⁺, BM CD4⁺ICOS⁺, and BM CD8⁺ICOS⁺. These findings suggested that the invasion of ICOS⁺ activated T cells was primarily responsible for the early modeling of huPBMC-AIA.

Fig. 3

A Representative immunofluorescence staining of paw tissues showing the expression of CD3 (red) and ICOS (green) at day3 (scale bar: 100 μm). White arrows indicate cells that colocalize. B Pie Chart and FACS analysis of ICOS expression on CD4 and CD8 T cells in BM of huPBMC-AIA (n = 6) and control group (n = 4) at day7. C tSNE plot and FACS analysis of ICOS expression on CD4 and CD8 T cells in paw of huPBMC-AIA (n = 6) and control group(n = 4) at day7. D Correlation between paw thickness measurements and percentage of human CD45⁺CD4⁺ICOS⁺ cells at the day7 time point. E Principal component analysis identifies major mediators distinguishing huPBMC-AIA mice from control mice. Each dot represents an individual mouse(left). Each line represents key drivers of the delineation. All values represent the mean ± SEM unless otherwise specified. t test was conducted for statistical significance. ****, p < 0.0001; ***, p < 0.001; **, p < 0.01; *, p < 0.05; ns p > 0.05

Conjugation and characterizations of Cy7-ICOS mAb

The anti-human ICOS mAb was conjugated with Cy7-NHS ester, then purified using a Vivaspin2 50 KDa cut-off spin filter (Fig. 4A). The multiscan spectrum was used to test the probe absorbance (Fig. 4B). Based on the results of three separate trials, the chemical yield of the probe was 65%. To assess the binding of Cy7-ICOS mAb to activated T cells, in vitro cell uptake assays were conducted. There was no obvious difference at 1nM and 10nM of Cy7-ICOS mAb concentration, whereas significant difference could be detected at 100nM level (p < 0.01). In contrast to blocked or naïve cohorts, the ICOS MFI was much higher in activated T cell group (p < 0.001) (Fig. 4C and D).

Fig. 4

A Scheme of Cy7 NHS ester conjugation with anti-human ICOS monoclonal antibody (mAb). B The characterization of labeled antibody at different concentrations. Relative fluorescence unit (RFU). C Cell uptake of Cy7-ICOS mAb in PMA/Iono activated, blocked, and resting T cells in huPBMCs at 1-hour incubation (n = 3). D MFI was measured to assess cellular uptake of the ICOS mAb labeled with Cy7 fluorophore at a concentration of 100nM. All values represent the mean ± SEM unless otherwise specified. One-way ANOVA and t test was conducted for statistical significance. ****, p < 0.0001; ***, p < 0.001; **, p < 0.01; *, p < 0.05; ns p > 0.05

Cy7-ICOS mAb detected activated T cells in huPBMC-AIA model via NIRF imaging

The 4-view NIRF imaging design and ROI atlas were shown in Fig. 5A. On day3, each mouse was received 50 µg Cy7-ICOS mAb administration via the tail vein, and longitudinal NIRF/X-ray imaging was carried out 6, 24, 48, 72, and 96 h post injection. 4-View images and Top NIRF/X-ray images were listed in Fig. 5B and C. As we expected, significant optical signals could be observed at injected right paw (RP) of hu-PBMC AIA group at all time points examined. The Cy7-ICOS mAb pharmacokinetics at RP from 4-view images are shown in Fig. 5D. From 6 h to 96 h, the RP fluorescence intensity of huPBMC-AIA was consistently higher than that in control group (6 h: p < 0.01; 24 h: p < 0.01; 48 h: p < 0.01; 72 h: p < 0.001; 96 h: p < 0.0001). The ROI profiles of the huPBMC-AIA and control groups were compared using fluorescence intensity ratio of RP to LP. The huPBMC-AIA group RP/LP ratio was noticeably higher than that in control group (huPBMC-AIA: 6 h = 1.62 ± 0.23; 24 h = 1.59 ± 0.23; 48 h = 1.43 ± 0.19; 72 h = 1.27 ± 0.09; 96 h = 1.25 ± 0.08; control: 6 h = 1.04 ± 0.04; 24 h = 0.87 ± 0.06; 48 h = 1.08 ± 0.11; 72 h = 1.06 ± 0.09; 96 h = 1.06 ± 0.10 and P value = 0.000171, 0.000024, 0.002716, 0.006593, respectively) (Fig. 5E). A strong correlation between paw thickness and RP/LP ratio could be detected with all time points ROI data (P value < 0.0001; R squared = 0.71) (Fig. 5F). As an optical imaging probe based on monoclonal antibody, Cy7-ICOS mAb should not exacerbate RA pathogenesis. To assess Cy7-ICOS mAb safety, the same dose of control mAb was also administrated. While there was no difference of paw thickness and body weight observed between two cohorts (Supplementary Fig. 2A and 2B). Blood urea nitrogen (BUN), creatinine (Crea), aspartate aminotransferase (AST), and alanine transaminase (ALT) were also measured as representative blood biochemical indicators (Supplementary Fig. 2C and 2D). The major organs were further stained for histological analysis (Supplementary Fig. 2E), and no pathological change was observed.

Fig. 5

A Reference of 4-view Regions of Interest (ROI) drawing was implemented, encompassing perspectives from the TOP, Bottom, Left, and Right orientations. B 4-view NIRF images at all time points (24 h, 48 h, 72 h, 96 h) between the huPBMC-AIA (n = 6) and control groups (n = 4). C X-ray and photographs images merge NIRF imaging following ICOS mAb-Cy7 injection at 6 h, 24 h, 48 h, 72 h and 96 h. D Fluorescence Intensity (FI) of four distinct views was systematically assessed at various time points. E ROI quantification of RP and the RP/LP ratio at all time points examined between huPBMC-AIA and control. F Correlation between paw thickness and RP/LP ratio in NIRF. Significant correlation was observed in both huPBMC-AIA and control group. All values represent the mean ± SEM unless otherwise specified. One-way ANOVA and t test was conducted for statistical significance. ****, p < 0.0001; ***, p < 0.001; **, p < 0.01; *, p < 0.05; ns p > 0.05

Evaluation of ICOS imaging metrics for RA diagnosis at early stage

Ex vivo NIRF imaging was performed immediately after the final imaging at day 7 to validate the accuracy of our manual ROI measurement (Fig. 6A). Among all the organs collected, the RP FI in huPBMC-AIA group was higher than control group (P < 0.05), which is consistent with in vivo NIRF imaging data (Fig. 6B). Intra-group correlation and cluster analysis of in vitro distribution data showed that RP was clustered significantly with liver, while RP was negatively correlated with kidney (p < 0.05) (Fig. 6C). This indicates that ICOS-mAb exhibits selectivity in the paw caused by arthritis and complies with the hepatic absorption metabolism. The NIRF and BioD data showed good concordance at all RP locations (Pearson correlation: 6 h = 0.91; 24 h = 0.88; 96 h = 0.67) (Fig. 6D). PCA showed that PC1 contributed 92.95% and PC2 contributed 4.04%, which indicated that the NIRF signal from RP was the major mediator to distinguish RA from control cohort. (Fig. 6E). ROI values from dimension reduction analysis at various time stages demonstrated that the preclinical imaging of the control and huPBMC-AIA groups could be clearly identified. The huPBMC-AIA could be distinguished from control group using ROC curves based on FI RP/LP ratio at all time points photographed (P < 0.0001; area under the curve (AUC) = 0.9778) (Fig. 6F), which demonstrated the capacity of ICOS NIRF imaging in RA early diagnosis.

Fig. 6

A Ex vivo imaging in major organs (1 heart, 2 liver, 3 lung, 4 spleen, 5 kidney, 6 intestine, 7 right paw (RP), 8 left paw (LP), 9 bone, 10 muscle, 11 skin, and 12 blood) at day 7 right after the last in vivo NIRF scan (X-ray and photographs images). B Fluorescence intensity (FI) statistics were analyzed for both paw tissues and major organs following the intravenous administration of ICOS mAb labeled with Cy7 fluorophore at the 96 h time point. C Intra-group correlation analysis, the uptake and metabolism of the anti-human ICOS-Cy7 were evaluated. Row/column order determined by unsupervised hierarchical clustering. Side bars represent log10 (FI of organs) at 96 h. D Correlogram depicting r2 Pearson correlation between FI measured in the indicated ROI, compared with the log10 (fold-change RP/LP)at various time points. E Principal Component Analysis (PCA) involved the transformation of the data into principal components, enabling a comprehensive exploration of patterns and variations within the normalized ICOS NIRF signals. F ROC analysis showing sensitivity against 100%specificity for distinguishing the huPBMC-AIA from control group based on FI fold change imaged at 24 h and 96 h. All values represent the mean ± SEM unless otherwise specified. One-way ANOVA and t test was conducted for statistical significance. ****, p < 0.0001; ***, p < 0.001; **, p < 0.01; *, p < 0.05; ns p > 0.05

Discussion

The life quality of rheumatoid arthritis patients could be improved via early diagnosis and intervention [24]. Given the low specificity and sensitivity, current approaches often fail in RA early detection. Thus, the development of novel diagnostic tools is essential. In this study, we identified ICOS as an indicator of RA pathogenesis, and NIRF imaging targeting ICOS enables precise detection of RA at early stages.

For selecting the most sensitive and specific cell portion and biomarkers for RA imaging, we first employed RNA-seq omics data. Activated T cells were identified as the primary cell portion in early RA samples, which was in line with our knowledge. In addition to outperforming other T cell costimulatory receptors like TNFRSF 9 (4-1BB) and TNFRSF 4 (OX40), ICOS was identified as one of the most differential T cell genes among the DEGs found in the same dataset. This will lessen the debate over the choice of biomarkers for tracking activated T cells. Previous studies noted that abnormal ICOS expression was correlated with the pathogenesis of RA [25]. Further flow cytometry and immunofluorescence staining data from protein level matched with the omics data from genetic level. A strong correlation between paw thickness and ICOS expression level strengthened the hypothesis that ICOS was an ideal biomarker for RA diagnosis and severity monitoring. Based on the data listed, the NIRF imaging data acquired should be attributed to the massive infiltrated activated T cells in the ankle joint from huPBMC-AIA model, which makes ICOS imaging with high specificity. The NIRF imaging pattern we chose for this study is widely applied in surgery guiding and superficial disease imaging [26], and other advantages, such as non-radioactive, safe and low cost, enables NIRF imaging more suitable for clinical translation. Through NIRF imaging, we could visualize the dynamics of activated T cells in a non-invasive manner. With ROI quantification, we could easily distinguish RA mice from control cohort at all time-point imaged, which demonstrated NIRF imaging targeting ICOS was a reliable tool for RA detection. Ex vivo Pearson, PCA and ROC analysis further reinforced the thesis. Thus, with this non-invasive imaging technique, physicians could optimize RA patient management in clinic.

Fundamental distinctions of immune systems between mice and human hamper the translation of basic research results to the bedside [27]. Multiple editorials and assessments on the present status of immunology research have recently emphasized the demand for humanized immunology rather than murine immunology [28]. Previous T cell costimulatory receptors imaging studies were all targeting murine activated T cells, and to our knowledge, this is the first study of targeting human ICOS. Due to its better similarity to the human immune system, the creation of humanized systemic RA models will provide us more hope for ICOS in RA detection at an early stage. The complement C5 component loss in the NOG mice used in this investigation results in decreased activation of the remaining mouse immune system [29]. A completely functional human immune system composed of T, B, NK, and dendritic cells, as well as monocytes/macrophages and granulocytes [30]. The Human Blood Lymphocytes SCID (Hu-PBL-SCID) approach involves injecting human peripheral blood mononuclear cells, resulting in fast proliferation of lymphocytes. Hu-PBL-SCID has reduced myeloid and B cell growth, as well as underlying xenogeneic graft-versus-host disease (GVHD) lasting 4–8 weeks [31]. Total RNA-seq study with CFA injected paw and KEGG analysis emphasizes the validity of huPBMC-AIA model. One of the most promising avenues for future research lies in extending the application of T cell imaging probes to other autoimmune conditions. The adaptability of our imaging approach suggests its potential utility in diseases such as rheumatoid arthritis, systemic lupus erythematosus, and multiple sclerosis, among others. Investigating T cell behavior in diverse autoimmune contexts can uncover commonalities and unique aspects of immune responses, facilitating a more comprehensive understanding of autoimmunity.

ICOS imaging study presented here does have limitations. The flow cytometry data had indicated the increased expression of ICOS in bone marrow, while with NIR-I imaging dye Cy7, we failed to capture this signal in vivo. To further improve our work, we will employ NIR-II dye (1000–1700 nm) for deeper tissue imaging. Another limitation was that CFA injection could only induce acute inflammation, and the foot swelling appeared too fast. Thus, the hypothesis to acquire NIRF images before RA symptoms showing up could not be realized in this model. In the future, we should test collagen antibody-induced arthritis (CAIA) model, which may provide a chronic inflammation microenvironment.

In this proof-of-concept work, we successfully identified ICOS as a sensitive and specific biomarker for RA detection, and NIRF imaging with Cy7-ICOS mAb was a promising approach in early RA diagnosis. Among the complex microenvironment of RA, one single biomarker imaging could make the diagnosis easier and simple, which may provide a better choice for RA patients management. Our findings suggest ICOS imaging techniques are warrant further evaluation in the clinic.