Longitudinal proteomics analysis in the immediate microenvironment of islet allografts during progression of rejection

Abstract

The applicability and benefits of pancreatic islet transplantation are limited due to various issues including the need to avoid immune-mediated rejection. Here, we used our experimental platform of allogeneic islet transplant in the anterior chamber of the eye (ACE-platform) to longitudinally monitor the progress of rejection in mice and obtain aqueous humor samples representative of the microenvironment of the graft for accurately-timed proteomic analyses. LC-MS/MS-based proteomics performed on such mass-limited samples (~5 μL) identified a total of 1296 proteins. Various analyses revealed distinct protein patterns associated with the mounting of the inflammatory and immune responses and their evolution with the progression of the rejection. Pathway analyses indicated predominant changes in cytotoxic functions, cell movement, and innate and adaptive immune responses. Network prediction analyses revealed transition from humoral to cellular immune response and exacerbation of pro-inflammatory signaling. One of the proteins identified by this localized proteomics as a candidate biomarker of islet rejection, Cystatin 3, was further validated by ELISA in the aqueous humor. This study provides (1) experimental evidence demonstrating the feasibility of longitudinal localized proteomics using small aqueous humor samples and (2) proof-of-concept for the discovery of biomarkers of impending immune attack from the immediate local microenvironment of ACE-transplanted islets.

Significance:

The combination of the ACE-platform and longitudinal localized proteomics offers a powerful approach to biomarker discovery during the various stages of immune reactions mounted against transplanted tissues including pancreatic islets. It also supports proteomics-assisted drug discovery and development efforts aimed at preventing rejection through efficacy assessment of new agents by noninvasive and longitudinal graft monitoring.

1. Introduction

Type 1 diabetes (T1D) is a life-threating condition because of the insulin deficiency resulting from the autoimmune destruction of the insulin-producing β-cells in pancreatic islets. Insulin deficiency leads to hyperglycemia, and chronic hyperglycemia in turn leads to serious health complications. Notably, many diabetic complications still occur despite lifelong insulin therapy. Alarmingly, T1D global incidence is increasing by 2–3% per year, with an overall annualized incidence of 22.9 cases per 100,000 people from 2001 to 2015 in the US alone [1]. While the specific mechanisms responsible for T1D are not known, certain HLA alleles and some environmental influences are considered risk factors and, notably, T1D is currently the only autoimmune condition without an approved immunotherapy [2-5]. Moreover, there are no predictive biomarkers that can reliably distinguish at-risk subjects who will progress toward symptoms from those who will not. This makes attempts at prevention very challenging and therapeutic interventions at the time of diagnosis significantly less efficacious. Transplantation of isolated pancreatic islets has emerged as a promising form of biological insulin replacement therapy. Since it can significantly improve the quality-of-life of patients with brittle T1D, it is now approved as “standard-of-care” in many countries [6-8]. However, the long-term efficacy of transplanted islets can be limited by various reasons that lead to graft failure with immune-mediated rejection being a major contributor [9]. Hence, the elucidation of changes induced in the protein networks by the onset of the inflammatory and immune responses during the progression of islet allograft rejection and the identification of early signs and biomarkers of rejection are of particular interest since they would allow timely therapeutic interventions to preserve the islet graft.

In the present study, we aimed to demonstrate the proof-of-concept that (1) longitudinal proteomics can be performed using small samples of the immediate microenvironment of islet allografts and (2) this significantly facilitates the identification of biomarkers of the different phases of rejection since they could be locally enriched in these samples. We were particularly interested in early biomarkers of the mounting of immune attack and degradation of the graft before irreversible damage has occurred. We used our established experimental platform where allogeneic pancreatic islets are transplanted in the anterior chamber of the eye (ACE) [10]. We have previously shown that by allowing longitudinal visual monitoring of the grafts, the ACE-platform uniquely enables the early detection of immune reactions against ACE-transplanted allogeneic islets noninvasively and longitudinally with single cell resolution [11,12]. We have also recently shown that the aqueous humor is representative of the immediate local microenvironment of ACE-transplanted islets and that omics-type studies in the aqueous humor might facilitate the identification of locally enriched islet- and immune-related features that could be of great value as biomarkers [13].

Mass spectrometry-based proteomics has evolved in recent years to allow the detection of extremely low abundance proteins and make possible longitudinal monitoring in small-volume biological samples. Limitations due to low sample volume have been particularly hindering proteomic applications in the field of diabetes and pancreatic islets biology [14]. The ability to detect and identify potential peptide/protein biomarkers in clinical samples depends in part on the achieved degree of proteome coverage. Notably, relatively low abundance disease-specific biomarker proteins can be masked by the high and medium abundance proteins present in biological fluids, especially in the blood plasma [15]. The immunoaffinity-based depletion of these proteins followed by peptide-level fractionation may reduce the sample complexity for LC-MS analysis, and this has broadened the proteome coverage by reducing the sample complexity and dynamic range [16]. For instance, we and others have shown that immunoaffinity depletion of high-abundance plasma proteins coupled with off-line fractionation of protein digests provides a high coverage range of the proteome [17-20]. Using such advanced approaches, we have been able to routinely identify and quantify 1000–2500 plasma proteins [21-25], and identify putative serum biomarkers of new-onset T1D [26-29].

Here, we performed proteomic analyses in small mass-limited aqueous humor samples obtained longitudinally from the same mice during the progression of rejection of their ACE-transplanted allogeneic islets (starting before the transplant). We identified and quantified 1296 proteins, of which 178 were represented across all samples. The analyses revealed specific peptide/protein signatures and corresponding immune network interactions characteristic for each stage of the allograft rejection. These studies also allowed the identification of a putative early biomarker of the mounted inflammatory and immune response.

2. Materials and methods

2.1. Animals and animal care

All animal studies were approved by the University of Miami's Institutional Animal Care and Use Committee (IACUC) and the animal procedures followed the guidelines established by the Committee on Care and Use of Laboratory Animals, Institute of Laboratory Animal Resources (National Research Council, Washington DC). Mice were obtained from the Jackson Laboratory (Bar Harbor, ME, USA) and were housed under the supervision of the University of Miami's Department of Veterinary Resources (DVR) in micro-isolated cages in Virus Antibody Free (VAF) rooms with free access to autoclaved food and water.

2.2. Pancreatic islet isolation and transplantation

Pancreatic islets were obtained by enzymatic digestion of pancreata from 18 to 22 weeks old male DBA/2 donor mice, followed by purification on density gradients using protocols standardized at the Diabetes Research Institute (DRI) Pre-Clinical Cell Processing and Translational Models Core. After overnight culture, isolated islets were implanted in fully anaesthetized mice in the ACE as previously described in detail [10]. Recipient C57BL/6 mice of both genders were transplanted in the ACE with 20–40 islet equivalents (IEQ) in one eye.

2.3. Sample collection

Under general anesthesia, aqueous humor samples (4–6 μL) were collected from mice with ACE-transplanted islets by direct aspiration using glass micropipettes (40–60 μm tip diameter). Four samples were collected from each mouse: one at baseline prior to transplant and three post-transplant. After transplant, samples corresponding to the Pre-Onset phase of rejection were collected 7 days after transplant; at rejection Onset samples were collected during actual destruction of the islet grafts as judged by the longitudinal imaging [11,30]; and Post-Onset samples were collected 14–21 days after the onset of rejection. Parallel blood samples were obtained through the tail vein from the same mice immediately after the aqueous humor collection. Blood was collected into purple-top EDTA Microvette CB 300 (Sarstedt AG & Co.; Germany) and centrifuged immediately to separate the plasma. Plasma was transferred into acid-cleaned small 0.5 mL non-stick surface microcentrifuge tubes (VWR; Radnor, PA, USA). Both aqueous humor and plasma samples were frozen at −80 °C immediately after collection for further analysis. Cytokine measurements in aqueous humor were done by Bio-Plex assay (Bio-Rad, USA) and were acquired by a Luminex instrument using xPonent software and analyzed using Milliplex Analyst software (Luminex, USA).

2.4. Proteomics analysis

Samples were diluted 10 folds in 0.2% N-dodecyl β-D-maltoside in 50 mM Tris pH 8 containing 10 mM dithiothreitol and incubating for 30 min at 37 °C. Reduced cysteine residues were alkylated by adding 400 mM iodoacetamide to a final concentration of 40 mM and incubating for 45 min at 25 °C in dark. Proteins were digested for 2 h at 25 °C using 1:20 (enzyme/protein ratio) of sequencing-grade endoproteinase Lys-C, followed by overnight incubation with 1:20 sequencing-grade trypsin at 25 °C. Digestion was quenched by adding pure formic acid to a final concertation of 1% and loaded into a C18 trap column. Chromatography was done on a Waters NanoAquity UPLC system with a custom packed C18 column (70 cm × 75 μm i.d., Phenomenex Jupiter, 3 μm particle size, 300 Å pore size) coupled to a Q-Exactive mass spectrometer (Thermo Fisher Scientific). Peptides were eluted with the following gradient of water (solvent A) and acetonitrile (solvent B) both containing 0.1% formic acid: 1–8% B in 2 min, 8–12% B in 18 min, 12–30% B in 55 min, 30–45% B in 22 min, 45–95% B in 3 min, hold for 5 min in 95% B and 99–1% B in 10 min. Analytes were analyzed online by nanoelectrospray ionization. Full mass scans were collected from 300 to 1800 m/z at a resolution of 35,000 at 400 m/z. Tandem mass spectra were collected in data-dependent acquisition of the top 12 most intense parent ions using high-energy collision induced dissociation (HCD) fragmentation (2.0 m/z isolation width, 30% normalized collision energy; 17,500 resolution at 400 m/z), before being dynamically excluded for 30 s.

LC-MS/MS data were analyzed with MaxQuant software (v. 1.6.5.0) [31]. Peptides were identified by searching against the mouse reference proteome databased from Uniprot Knowledge Base (downloaded August 14, 2018). Searching parameters included cleaving by trypsin in both peptide termini but allowing up to 2 undigested sites per peptide. Acetylation of protein N-terminal and methionine oxidation were considered as variable modifications, whereas cysteine carbamidomethylation was set as a fixed modification. Mass error tolerance was set as the default parameter of the software. Quantitative information was extracted using the intensity-based absolute quantification (iBAQ) method [32].

2.5. Determination of cystatin 3 by ELISA

Cystatin 3 was detected and quantified using a corresponding ELISA kit (Mouse/Rat Cystatin 3 Quantikine ELISA kit, #MSCTC0, sensitivity 12.9 pg/mL, intra-assay CV 3.16%; R&D Systems, Minneapolis, MN, USA). The mouse parallel aqueous humor and plasma samples from each time point were diluted 1/200 for detection. All ELISA measurements were conducted as per manufacturer's instructions, with the standard curve yielding good linearity (r² = 0.988).

2.6. Statistical and bioinformatics analyses

Standard statistical analyses including t-tests and ANOVA were performed in GraphPad Prism v8.2 (GraphPad; La Jolla, CA, USA, RRID:SCR_002798). Data were curated for proteins identified in all four longitudinal time-points and then used for the visualization of associated interaction networks, canonical pathways, upstream regulators, functions & diseases enrichment, and biomarker prediction using Ingenuity Pathway Analysis (IPA) software (Qiagen Bioinformatics; Redwood City, CA, USA; https://www.qiagenbioinformatics.com/products/ingenuity-pathway-analysis; RRID:SCR_008653). MetaboAnalyst 4.0 (https://www.metaboanalyst.ca; RRID:SCR_015539) was used for Principal Component Analysis and hierarchical clustering by proteins using the default settings for data processing and median-normalization and auto-scaled (mean-centered and divided by the standard deviation of each protein). The input for these analyses consisted of matrices of relative abundance values of proteins identified by their gene names.

3. Results

3.1. Longitudinal in vivo monitoring of islet allograft rejection

Allogeneic DBA/2 donor islets were transplanted in the ACE of C57BL/6 recipient mice, and aqueous humor samples were collected before (Pre-TX) and after the transplant. After transplant, samples were collected at three time-points spanning the pre-onset, onset, and post-onset phases of rejection; they are designated accordingly (Fig. 1A). As evidenced in the longitudinal images, repeated sampling of the aqueous humor in the same eye neither affected the eye nor influenced the kinetics of rejection of the islet allografts (Fig. 1B). Median time to rejection was 18 days post-transplant.

Fig. 1.

Islet transplantation in the anterior chamber of the eye (ACE) allows longitudinal direct evaluation of allogeneic islet grafts in parallel with sampling of their local microenvironment during the progression of allo-rejection. (A) Representative sequential photographs of a C57BL/6 (B6) mouse eye before and after transplantation of DBA/2 islets in the ACE and during the progression of immune-mediated rejection. The islet allografts (white masses on top of iris) were clearly visible after transplantation and before rejection onset (Pre-onset; POD7). At onset of rejection (Onset), the individual islets were severely diminished in size, and smaller islets became barely visible. During the Post-onset phase (2–3 weeks after rejection onset), all islets had disappeared. Note that the cornea remained clear throughout the repeated sampling showing only a localized scar at the incision site from the transplant procedure. (B) Kaplan-Meier survival curve showing the kinetics of destruction of the islet allografts in the ACE of the recipient mice (n = 5). Photos shown in A and data in B were with repeated collection of aqueous humor samples (representative of the immediate microenvironment of the islet allografts) from the same eye(s) before the transplant and at the various stages of rejection after the transplant, demonstrating no negative impact of the repeated sampling on the eye or the rejection kinetics.

3.2. Longitudinal localized proteomics during the progression of rejection

Because of the small size of the longitudinally collected aqueous humor samples, we first verified that reliable proteomics results can be obtained with them using our nano-UPLC platform-based LC-MS/MS method. Aqueous humor samples pooled from multiple mice were divided into volumes of 10, 15, and 30 μL and then analyzed separately for protein content to check for consistency of the results with decreasing sample volumes. Relative abundances obtained from the lower volume samples correlated well with those obtained from the largest volume (30 μL) (Supplementary Data Fig. S1). Therefore, we proceeded with the analysis of the longitudinal aqueous humor samples collected from the same mice before the transplant (Pre-TX) and during the progression of rejection of their islet allografts. The localized proteomics performed in these longitudinal samples identified a total of 1296 peptide/proteins. Of these, 178 proteins were represented across all four time points (Pre-TX, Pre-Onset, Onset, and Post-Onset). A comprehensive list of these proteins is provided in the Supplemental Data (Table S1A,B). Our analysis revealed two distinct subsets of proteins with strongly increasing or decreasing abundances during the rejection progression. One set of 12 proteins had significant and more than 20-fold increase in the Post-Onset phase (relative to Pre-TX) with some showing > 100-fold change (Lecithin-cholesterol acyltransferase [Lcat], Spectrin alpha erythrocytic 1 [Spta1], and Complement C9 [C9]) and even > 1000-fold increase (Serine protease inhibitor family F member 1 [Serpinf1]) (Fig. 2A). Another set of 18 proteins showed significant and > 20-fold reduction in abundance with some showing > 100 × (e.g., Serine protease inhibitor clade A member 1d [Serpina1d], Alpha 2-HS glycoprotein [Ahsg], Apolipoprotein A1 [ApoA1], Kininogen 1 [Kng1]) and even > 1000 × decrease (Hemopexin [Hpx], Beta-globin [Hbbt1], and Transferrin [Tf]) (Fig. 2B). The numbers and distribution of proteins with increased abundances from 4 to 20-fold and > 20-fold in each rejection phase (i.e., Pre-Onset, Onset, and Post-Onset) are shown as Venn diagrams (Fig. 2C).

Fig. 2.

Proteins with significant and > 20-fold change in the immediate microenvironment of the islet allografts during progression of rejection as revealed by localized proteomics in longitudinal aqueous humor samples. (A,B) Soluble proteins showing > 20-fold (A) increase or (B) decrease in abundance relative to baseline before the transplant (Pre-TX) in longitudinal aqueous humor samples collected during the progression of rejection of ACE-transplanted allogeneic islets. Protein levels were measured by LC-MS/MS in the individual samples collected at the various phases of rejection (Pre-Onset, Onset, and Post-Onset) from recipient mice (n = 5) and normalized to the corresponding protein levels in baseline samples obtained from each mouse before the islet transplant (Pre-TX). (C) Venn diagrams showing the overlap in the proteins identified in the Pre-Onset, Onset, and Post-Onset phases of rejection that showed (left) a ≥ 4–20-fold increase and (right) > 20-fold increase in the Post-Onset phase compared to Pre-TX.

3.3. Changes in the local proteome during the rejection progression

The protein patterns identified in the longitudinal local samples during the various phases of the islet allograft rejection were evaluated in comparison to the baseline condition (i.e., before the transplant) using clustering analysis tools. Principal component analysis (PCA) applied to the complete dataset revealed a clear dissociation of the Pre-TX phase from the Onset and Post-Onset phases, but not from the Pre-Onset phase (Fig. 3A). Consistent with this, clustering of proteins presented as heatmaps showed clear divergence in the groups of proteins that are characteristically either increased or decreased post-transplant compared to pre-transplant with accentuated differences in the Onset and Post-Onset phases of rejection (Fig. 3B).

Fig. 3.

Longitudinal changes in the local proteome as revealed by clustering/enrichment analyses during the progression of islet allograft rejection. (A) Principal component analysis (PCA) showing changes in protein clustering across the Pre-Onset, Onset, and Post-Onset phases of islet allograft rejection as compared to pretransplant (Pre-TX). Plots were generated by MetaboAnalyst 4.0. The x and y axes represent principal components 1 and 2 (PC1, PC2), respectively; numbers in parentheses indicate the contribution of each component to explain the total variance. (B) Hierarchical clustering analysis of protein levels/abundances assessed by LC-MS/MS and presented as heatmaps highlighting the top 25 differentially expressed proteins during the Pre-Onset, Onset, and Post-Onset phases of rejection versus Pre-TX. Protein levels were normalized by the median of all values for each individual protein and auto-scaled (mean-centered and divided by the standard deviation). Columns represent individual samples from each mouse (M1–M5; n = 5) and rows indicate each identified protein annotated by the corresponding gene name. Red and blue colors indicate increased or decreased protein levels, respectively. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

To gain a deeper insight into the functions, processes, and pathways of potential physiological relevance to the immune-mediated islet allograft rejection, we exploited the analytical capabilities of the Ingenuity Pathway Analysis (IPA) software using the “Comparison Analysis” module to analyze in parallel the complete longitudinal proteomics datasets for “Canonical Pathways”, “Upstream Regulators”, and “Diseases & Functions” that are impacted during the progression of rejection. The analyses showed significant involvement of immune and inflammatory pathways/processes consistent with the advance of the islet allograft rejection. These results are presented in heatmaps to highlight trends across the Pre-Onset, Onset, and Post-Onset phases of the rejection (Fig. 4). Canonical pathway analysis showed gradual activation of the acute phase response signaling and the complement system as the most significantly affected pathways (Fig. 4A). Predicted upstream regulators analysis showed predominance of pro-inflammatory proteins such as interleukin-5 (IL-5), prolactin (PRL), or IL-4 in the Onset and Post-Onset phases (Fig. 4B). Further, analysis of disease and functions identified a large number of immune-relevant events, most of which were increasingly activated in the Onset and Post-Onset phases of rejection, and many of which are predominantly involved in cytotoxic functions, cell movement, and innate and adaptive immune responses (Fig. 4C).

Fig. 4.

Comparative enrichment analyses in longitudinal samples of the local proteome of islet allografts reveal a pattern consistent with the initiation and progression of immune inflammatory responses during the progression of the islet allograft rejection. (A–C) Proteomics datasets corresponding to the Pre-Onset, Onset, and Post-Onset phases of rejection (normalized to Pre-TX) analyzed for predicted involvement in various immune inflammatory processes/pathways using the “Comparison Analysis” module of IPA software and displayed as heatmaps. (A) Canonical pathways analysis showing activation of the acute phase immune response consistent with the initiation and progression of the islet allograft rejection. Blue colors indicate activation, white colors inactivation, and colour intensity the magnitude of the p-values. (B) Upstream regulator analysis revealing major pro-inflammatory agents with significant impact on immune responses (e.g., PRL, IL-5, and IL-4) during the progression of the islet allograft rejection. (C) Analysis of diseases and functions that were predicted by IPA to be most affected during the progression of rejection. Proteins associated with cytotoxicity of T-lymphocytes and cellular immune response including cell movement were among the most enriched features at the Onset and Post-Onset phases of rejection. In B and C, red colors indicate activation, green inactivation, and colour intensity the magnitude of the z-value. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3.4. Dynamics of the immune and inflammatory response networks

Given that the above analyses revealed several specific immune and inflammatory events as part of the longitudinal changes in the local proteome during the progression of rejection, we also explored potential logical, cause-effect network connections to better understand the functional implications yielded by the comparative enrichment analyses. Toward this, we subjected the longitudinal proteomic datasets to “Network Prediction Analysis” in IPA (Fig. 5). The networks predicted as the most relevant based on the significant alteration in the relative abundance of the proteins analyzed for the Pre-Onset phase were clearly dominated by a central node involving humoral immune responses in connection with IL-21 receptor (IL-21R), consistent with a prelude of an adaptive immune response preceded by proliferation and/or differentiation events (Fig. 5A). Analysis of the Onset phase data predicted a network with higher degree of complexity with two primary nodes centered on the Tumor Growth Factor beta (TGF-β) and Major Histocompatibility Complex (MHC) class II as major players, among other immune-relevant nodes and connections with lower predicted scores that are shown in faded colors (Fig. 5B). For the Post-Onset phase, three nodes were predicted as most affected, and all three signaling networks have known involvement in inflammatory and cellular immune responses predominantly by T-lymphocytes through activation of the T-Cell Receptor (TCR) and Nuclear Factor Kappa B (NF-κB) (Fig. 5C). High resolution versions for closer examination of the networks shown in Fig. 5 are provided in the Supplemental Data (Fig. S3).

Fig. 5.

Network prediction analyses reveal transition from humoral to cellular immune response and exacerbation of pro-inflammatory signaling during the progression of rejection. Illustrative sequence of immune and inflammatory pathway networks recruited during the progression of islet allograft rejection as predicted by IPA software based on the longitudinal proteomics datasets acquired from the local microenvironment during the (A) Pre-Onset, (B) Onset, and (C) Post-Onset phases of the islet allograft rejection as compared to pre-transplant (Pre-TX). Abbreviations correspond to the gene names of the proteins involved. Whereas, the network predicted most relevant among immune and inflammatory diseases/processes involved based on the relative abundance of the proteins analyzed at (A) Pre-Onset was dominated by a central node involving humoral immune response and evident initiation of a cell proliferation/differentiation response driven by IL-21. The predicted networks during the ensuing (B) Onset and (C) Post-Onset phases showed robust recruitment of pro-inflammatory immune responses driven by activation of the NF-κB pathway, increased MHC class II (i.e., antigen presentation), and increased TCR expression (i.e., T-cell proliferation/activation). In the Post-Onset phase, activation of TGF-β was also predicted, likely, consistent with active resolution of the allo-immune responses to prevent immune overshoot and bystander damage to surrounding tissues. Network elements are represented by symbols of various shapes and colors; those shown in full colour had the highest relevance among immune-relevant nodes and connections, some less relevant are shown in faded colors in the background. Marker shapes are as follows – horizontal-oval: transcription regulator; vertical-oval: transmembrane receptor; diamond: enzyme; square: cytokine; vertical-rectangle: G-protein coupled receptor; broken-lined vertical-rectangle: ion channel; and horizontal-diamond: peptidase. Marker colors – orange: predicted activation and blue: predicted inhibition. Connecting line colors – orange: activation; blue: inhibition; yellow: findings are inconsistent with the state of the downstream molecule; and gray: not predicted. High resolution versions of these networks are provided in the Supplemental Data (Fig. S3). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

3.5. Possible early biomarker(s) of islet allograft rejection

Next, we explored the feasibility of identifying biomarker proteins in the locally enriched proteome that could be useful in predicting rejection of the islet allografts before irreversible damage to the graft occurred. Using the “Biomarker Prediction” module of IPA software, we focused the analyses on the datasets obtained in the Pre-Onset and Onset phases of rejection to identify proteins with evident positive trends when compared to baseline (Pre-TX). Evaluation for diseases and functions related to immune and inflammatory responses yielded a predicted network with three major protein candidates: Cystatin 3 (Cst3), Fatty acid-binding protein 5 (Fabp5), and Beta-2-Microglobulin (B2m) (Fig. 6).

Fig. 6.

Candidate biomarkers of allograft rejection predicted by longitudinal localized proteomics. The biomarker candidates cystatin 3 (Cst3), fatty acid-binding protein 5 (Fabp5), and beta-2-microglobulin (B2m) were predicted by the “Biomarker prediction” module in IPA software based on the Pre-Onset and Onset phase datasets to identify proteins with evident positive trends when compared to baseline (Pre-TX). The predicted biomarkers are shown circled in red as part of an illustrative pathway network displaying immune and inflammatory processes likely to be affected during the early phase of rejection. Dashed lines represent functional relationships supported by experimental observations; octagons and crosses denote functions and associated disease(s), respectively. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Out of these three proteins identified as candidates for early biomarkers of islet allograft rejection, we focused on Cystatin 3 (Cst3) because of the change in its relative abundance during the progression of rejection (Fig. 7A; also see Fig. 2A) and for other considerations discussed further below. We sought to further validate the changes observed by mass-spectrometry in the aqueous humor samples by ELISA-based quantifications in parallel aqueous humor and plasma samples collected from the same mice during the progression of islet allograft rejection. Consistent with the proteomics data, these measurements showed significantly elevated Cst3 levels in the aqueous humor samples during the rejection progression compared to the corresponding baselines (Pre-TX) (Fig. 7B). Notably, Cst3 levels did not change in the parallel plasma samples (Fig. 7C).

Fig. 7.

Validation of Cystatin 3 (Cts3) as a putative early biomarker of islet allograft rejection identified by localized proteomics. (A) Cts3 protein abundance levels as assessed by LC-MS/MS proteomics in longitudinal samples of the immediate microenvironment (aqueous humor) of ACE-transplanted islets during the progression of their rejection. Longitudinal samples were collected from each mouse (n = 5) prior to transplant and at the various phases of rejection. (B,C) Cts3 protein levels measured by ELISA in parallel (B) aqueous humor and (C) plasma samples collected simultaneously from the same mice (n = 5) during the different phases of rejection of ACE-transplanted allogeneic islets. Data are shown as box and whiskers plots indicating the medians (horizontal lines in each box) and the upper and lower quartiles with all data points shown as open round symbols at each time point. Asterisks indicate significant differences vs Pre-TX (one-way ANOVA followed by Tukey's multiple comparison test; *p < .05, **p < .01).

4. Discussion

We have previously established that the ACE-platform provides unique opportunities to study the pancreatic islet immunobiology in unprecedented detail [11,12,30,33-36]. More recently, we showed that having access to the immediate microenvironment of ACE-transplanted islets, as represented in the aqueous humor, facilitates the identification of locally enriched islet- and immune-related metabolites that could provide informative metabolic signatures of the islet functional and inflammatory status [13]. In the present study, we expanded on this work to (1) further demonstrate the feasibility of longitudinal proteomic analyses during the progression of rejection of ACE-islet allografts, (2) highlight the potential of localized proteomics in the elucidation of relevant mechanistic details, and (3) demonstrate the feasibility of identifying locally-enriched early protein-based biomarkers of islet rejection. It is hoped that changes identified this way in the local proteome, combined with those in the associated metabolome, lipidome, and transcriptome, could provide a comprehensive biomarker signature to reliably predict rejection before irreversible damage to the graft occurs. Moreover, these localized multi-omics studies enabled by the ACE-platform could provide not only comprehensive insight into the mechanisms of islet allograft rejection, but also new therapeutic targets in support of drug discovery and development efforts aimed at preventing rejection and promoting long-term acceptance of allografts of pancreatic islets as well as other cellular transplants (e.g., stem cells and stem cell-derived islet organoids) [37,38]. While changes here reflect those occurring in the eye as a transplant site, we [11,12] and others [39-44] have consistently shown that both allo- and auto-immune responses against ACE-transplanted islets mirror those against islets during T1D pathogenesis within the pancreas as well as at extrapancreatic sites after transplant (e.g., kidney and ear). Moreover, the viability of syngeneic islet grafts and tolerated allografts in the ACE remains unaltered indicating that the immune attack is responsible for the allogeneic islet destruction in the ACE model [11,30,36].

Recent advances in proteomics have enabled untargeted studies in aqueous humor samples obtained exclusively from humans mainly due to the relatively larger sample volumes (50–200 μL) that can be obtained from the human eye [45-49]. Here, we demonstrated the feasibility and reliability of untargeted proteomics of aqueous humor from murine models that facilitate experimental manipulation for deeper mechanistic insights but yield much smaller sample volumes (~5 μL) (Supplementary Data Fig. S1). Specifically, our untargeted proteomics analyses of such mass-limited sample volumes enabled the current longitudinal studies of islet allograft rejection using aqueous humor samples from the same mice. This study identified > 1200 proteins. Adopting a similar LC-MS-based proteomics approach and much larger starting sample sizes, Adav and coworkers [49] identified ~1000 proteins, Chowdhury and coworkers [45] identified 676 proteins, and Murthy and coworkers [46] identified 763 proteins. Compared to these state-of-the-art studies, our proteomic data utilizing much smaller sample amounts, provided improved coverage of the aqueous humor proteome, which included that of the ACE-transplanted islets. These first of their kind studies revealed several proteins that had strongly (> 20×) and significantly altered relative abundances during the progression of the islet allograft rejection (Fig. 2). Notably, those showing the highest fold-increase included Serpin Family F Member 1 (Serpinf1; also known as Pigment Epithelium-Derived Factor, PEDF), Complement component 9 (C9), Destrin (Dstn), Histidine-rich glycoprotein (Hrg), and the Serine protease inhibitor A3N (Serpina3n) – all with known immune functions consistent with graft rejection. Serpinf1 is highly expressed in dendritic cells (DCs), which are primary professional antigen presenting cells (APCs). It is also expressed in pancreatic islets; hence, its post-transplant increase could be associated with the early graft-infiltration by DCs during the Pre-Onset phase as well as with its release from dying islet cells at Onset of rejection during their active destruction. Further, PEDF is highly abundant in the retina, where it protects photoreceptors through neurotrophic, neuroprotective, and antioxidant effects [50]. PEDF has been reported to act as a potent anti-inflammatory factor with a mechanism of action mainly involving the attenuation of macrophage function [51]. Hence, its significant increase during the Onset and Post-Onset phases of the islet allograft rejection likely implies a physiological reaction to limit the severity of the inflammatory response and prevent bystander damage to adjacent tissues within the eye. C9 is an essential part of the complement system of innate immunity [52]; Dstn is a constituent of the actin cytoskeleton that is vital for cell movement and phagocytosis [53]; Hrg is implicated in innate immunity by regulating macrophage phagocytosis [54]; and Serpina3n is implicated in inflammation and is transcriptionally regulated by the pro-inflammatory cytokine IL-6 [55,56]. Consistent with the latter, we observed significantly increased IL-6 expression in the aqueous humor of rejecting mice at onset compared to naïve non-transplanted mice (see Supplementary Data Fig. S2) [30]. The Purine nucleoside phosphorylase (Pnp), whose deficiency is associated with severe immunodeficiency mainly through compromised T-lymphocyte function both in humans and mice [57], was also among the proteins that increased dramatically during the rejection progression. Notably, while some of the increased proteins such as Afamin (Afm) and Hrg (Fig. 2A) are abundant in plasma, their significant increase in the immediate microenvironment of the islet allografts during the progression of rejection shows an association with rejection and warrants their further investigation as possible biomarkers or contributing factors in immune-mediated islet allograft rejection.

On the other hand, proteins found here to show strong and significant decrease during rejection (Fig. 2B) have known anti-inflammatory functions. For example, Serine protease inhibitor A3K (Serpina3k) has anti-inflammatory effects by altering the expression levels of Tumor Necrosis Factor-alpha (TNF-α) and Vascular Endothelial Growth Factor (VEGF) [58]. Consistent with this, we have previously shown in the same allogeneic model that TNF-α is significantly elevated at the rejection onset of ACE-transplanted islets [30]. Clusterin (Clu), which was reduced by > 20-fold, prevents inflammation by inhibiting metalloproteinase 9 (MMP9) [59]. Hence, the identification of these immune regulatory and anti-inflammatory proteins through longitudinal localized proteomics could create new opportunities for identifying novel draggable targets to reduce inflammation and promote immune regulation and islet allograft survival by highlighting therapeutic intervention possibilities that could be especially important during the early phases of the immune response. Indeed, anti-inflammatory therapies are routinely used in conjunction with immunosuppression in transplant applications and have been shown to prevent or delay the onset of hyperglycemia in T1D rodent models and patients [7,60,61].

To further investigate the relevance of the longitudinal changes we observed in the local proteome during the progression of the islet allograft rejection, we performed various in silico analyses across the different phases of rejection. For example, analysis of canonical pathways and upstream regulators revealed activation of the Prolactin (Prl) pathway and acute phase response signaling beginning in the Pre-Onset phase (Fig. 4A,B). Both of these are consistent with a rapid early innate immune response triggered by the complement system and pro-inflammatory cytokines. Interestingly, the expression patterns of several pro-inflammatory cytokines identified in the analysis of upstream regulators here (e.g., IL-1a, IL-5, and IL-6) were corroborated by protein measurements in aqueous humor samples obtained at rejection onset in the same model (see Supplementary Data Fig. S2) [30]. Consequently, several of the key proteins, pathways, and functions predicted by the current proteomic analyses during the different phases of rejection provided important insight into a sequence of cellular and molecular immune mechanisms during the rejection progression, many of which have been previously characterized, where early involvement of the innate immune system initiates the response against the islet allografts and contributes to the recruitment of adaptive immunity later in the rejection process (Fig. 5) [62,63].

Moreover, molecular network analysis of the Pre-Onset phase data predicted a central role of immunoglobulins consistent with a strong initial humoral immune response in conjunction with IL-21 with a known role in B- and T-lymphocyte expansion [64]. Changes in local vascular permeability, as we observed previously in the ACE model [12], might contribute to this immune response. This analysis was also consistent with recent experimental findings in skin allografts in a mouse model where blocking IL-21 delayed rejection [65]. Similar analysis of the Onset and Post-Onset datasets highlighted network nodes centered around MHC class II, T cell receptor (TCR), and NF-κB, all of which are essential for antigen presentation and activation of T-lymphocytes that primarily carry out the target-cell cytotoxicity during rejection [62,66,67]. Interestingly, the analysis also revealed TGF-β as a central node during both the Onset and Post-Onset phases of rejection (Fig. 5B,C). While TGF-β likely contributes to immune regulation during the resolution of immune reactions in the Post-Onset phase [68,69], its involvement during the Onset phase may either contribute in conjunction with IL-17 and IL-21 to the activation of effector Th17 T-lymphocytes [30,68] or possibly act in concert with other immune regulatory proteins such as PEDF (see above) as a mechanism to preempt overshoot of the alloimmune response and avoid bystander damage to surrounding tissues; which we have not observed here (Fig. 1) or in our prior studies [11,30].

Extensive research efforts have examined the potential utility of β-cell-derived extracellular vesicles (EVs), antigens and microRNAs (e.g., miRNA-375 and GAD65), and unmethylated insulin DNA as biomarkers of β-cell death by autoimmune destruction during T1D development or shortly after islet transplant in the liver due to initial hypoxia and instant blood mediated inflammatory reaction (IBMIR) [70-73]. However, there are still no reliable biomarkers yet for early detection of immune-mediated islet allograft rejection to prompt timely and efficient therapeutic intervention to preserve long-term function. Encouraged by the above findings on the longitudinal changes in the local proteome, we explored the possibility of identifying locally enriched protein-based early biomarkers that could reliably predict islet destruction before significant and irreversible damage to the graft has occurred. Analysis of the longitudinal proteomic datasets during the rejection progression identified three biomarker candidates: B2m, Fbap5, and Cst3 (Fig. 6). Interestingly, most of the actions of these proteins are associated with cell movement/chemotaxis, recruitment and expansion of both CD4⁺ and CD8⁺ T-lymphocytes, and MHC class II/antigen presentation [74-76]. We focused on Cst3, on one hand, because of the progressive change in its observed relative abundance and, on the other, because of its ubiquitous expression in cells, tissues, and biological fluids [77] – important practical considerations for the translational and clinical applicability of putative biomarkers. In fact, Cst3 has been clinically used as a biomarker of kidney function due to its small molecular weight and ease of detection to assess glomerular filtration rate (GFR) [78]. Recent studies have also uncovered regulatory elements in the Cst3 gene promoter region leading to increased Cst3 in association with pro-inflammatory conditions [79,80]. Cst3 is highly expressed in APCs [81], further supporting its involvement in the islet allograft rejection in our studies [82]. Cst3 is also expressed in β-cells and is linked to β-cell function and diabetes. Notably, Cst3 gene expression levels were shown to increase upon β-cell stress/damage by streptozotocin (STZ) treatment [83]. This line of evidence, combined with our current findings showing significantly increased Cst3 in the local proteome during the rejection progression (Fig. 7A), highlighted Cst3 as a possible locally-enriched early biomarker of islet allograft rejection. This notion was further supported by the ELISA measurements in independent longitudinal samples of aqueous humor collected during the rejection progression of ACE-transplanted allogeneic islets (Fig. 7B). Notably, Cst3 levels did not change in parallel plasma samples (Fig. 7C), further highlighting the usefulness of localized proteomics in the identification of islet–/immune-related protein-based biomarkers that could be diluted and hence overlooked in the general circulation.

5. Conclusions

The current study provided novel insight into the longitudinal changes in the local proteome of islet allografts during progression of their rejection. In silico analyses of proteomic datasets obtained in longitudinal samples of the immediate microenvironment of the islet allografts during rejection progression identified Cst3 as a putative early biomarker of rejection; which was further validated by ELISA protein measurements in independent mouse recipients that rejected their ACE-islet allografts. These studies demonstrated for the first time the feasibility of untargeted localized proteomics in small samples in the microliter range. They also established the proof-of-concept of initial discovery of protein-based islet- and/or immune-related biomarkers enriched in the local microenvironment of islet grafts using the combination of the ACE-platform and proteomics. Notably, selected locally-enriched biomarkers could be further investigated in the general circulation for their potential in clinical application where blood samples are routinely used for diagnostic and prognostic purposes. Furthermore, this concept can be expanded to multi-omics approaches to facilitate the identification of locally-enriched biomarkers and the construction of integrated biomarker signature(s) of the functional and immune status of transplanted islets and other tissue allografts.