Immunologic characteristics: a comparative analysis of TME and cytokine profiles in SC and LM models

To explore whether the SC and LM models exhibit different tumor growth patterns, we examined both models concomitantly in the presence or absence of PBMC transplantation. After transplantation of PBMCs, both models exhibited attenuated tumor growth by gross observation, reduced tumor volume (p<0.05), and a reduced tumor/body weight (BW) ratio (p<0.05) compared with mice that were not subjected to PBMC transplantation (Fig. 2A–D). We conducted standard hematoxylin and eosin (H&E) staining and immunohistochemical staining (IHC) assays to verify the distribution of TILs in both models. Representative H&E and IHC images demonstrate infiltration of CD3⁺ Pan T-cells, CD8⁺ cytotoxic T-cells, and CD 4⁺ helper T-cells following PBMC transplantation, as shown in Fig. 2E and F. The infiltration of CD3⁺ T cells, CD8⁺ T cells, and CD 4⁺ T cells into the tumor was notably more pronounced and concentrated in the LM model. In contrast, Pan-B cells, NK cells, and macrophages were not observed in either model (Supplementary Fig. ^S2). To investigate the cytokines, we measured plasma hIFN-γ and hTGF-β levels. Plasma hIFN-γ levels were significantly increased after PBMC transplantation in both models (p<0.05, Supplementary Fig. ^S3). In the SC model, hIFN-γ shows a continuous increase, whereas in the LM model, it decreases at the day 28. This decrease in the LM model is likely due to advanced cancer progression reaching a terminal stage, where the tumor can no longer grow. While the plasma levels of hTGF-β showed a different tendency than expected (Supplementary Fig. ^S3). In the SC model, there was no difference in hTGF-β levels according to PBMC transplantation until day 7, but after day 14, hTGF-β levels decreased significantly in the PBMC transplantation group (p<0.05). The cancer progression is inhibited, which may lead to a decrease in hTGF-β levels. Conversely, hTGF-β levels in the LM model increased up to day 14, but after day 21, hTGF-β levels decreased in both groups and were significantly higher in those receiving PBMC transplantation.

Open in a separate window

Figure 2

Analysis of tumor growth in a humanized mouse model via transplantation of peripheral blood mononuclear cells on day3. (A) Quantification of tumor growth and tumor weight in the subcutaneous (SC) model (B) and the liver metastasis (LM) model. (C) Representative images of tumors in the subcutaneous (SC) model. (D) LM model. (E) Histological analysis after hematoxylin and eosin (H&E) staining of CD3⁺T cells, CD4⁺ T cells, and CD8⁺ T cells in the SC model and (F) the LM model. Mice of the same sex and age as those in the experimental group were used as controls (n=5 mice per group). Tumor and liver tissues were collected every week after subcutaneous and intrasplenic injections and processed for histologic examination (×200). Error bars indicate the SEM. n.s. not significant; ^*p<0.05; PBMC, peripheral blood mononuclear cell; NSG, NOD SCID gamma mouse; NSG+PBMC, PBMC transplantation with NSG mouse; HLA, human leukocyte antigen; H & E, hematoxylin and eosin.

These data indicate that PBMC transplantation resulted in T-cell proliferation, which inhibited tumor growth, in both the SC and LM models; in particular, central infiltration was observed in the LM model.

Comparative analysis of T cell number changes between the SC and LM models

We used several panels of human leukocyte markers in a flow cytometry assay to analyze the dynamic changes in T cell numbers in each model (Fig. 3). Overall, the total T cell population was increased in both models, and T cell numbers peaked on day 28 in the SC and LM models. T cell distribution, frequencies and proliferations in tumors were quantified using flow cytometry with specific marker panels to elucidate the kinetics of T cell changes in each humanized mouse model. In the SC model, hCD4⁺ T cell numbers increased and hCD4⁺Foxp3⁺ T cell numbers decreased. In contrast, hCD4⁺Foxp3⁺ T cell numbers were increased and hCD4⁺ cell numbers were decreased in the LM model. This indicates that tumor immune evasion and adaptation differ between the SC and LM models. We also detected hCD8⁺ IFN-γ⁺ T cells, which showed an increasing tendency in the LM model, but we did not observe a similar tendency for hCD8⁺IFN-γ⁺ T cells in the SC model. To further investigate the changes in cell populations and the proliferation of human T cells in humanized mice, we evaluated Ki-67 expression. In the SC model, hCD4⁺ki67⁺ T cell numbers peaked at week 21, while hCD3⁺ki67⁺ T cells and hCD8⁺ki67⁺ T cell numbers remained at relatively low levels until week 21, before peaking at week 28. In the LM model, hCD3⁺ki67⁺, hCD4⁺ki67⁺, and hCD8⁺ki67⁺ T cell numbers gradually increased and peaked at 28 weeks. When PD-1 expression in hCD4⁺ T cell and hCD8⁺ T cells was compared between the SC and LM models, PD-1-positive cell numbers in the SC model were found to peak at week 14 and then decrease. In contrast, in the LM model, hCD8⁺PD1⁺ T cell numbers increased and peaked in week 28. We also examined T cells in the blood and found a pattern similar to that observed for tumor infiltration, but the peak was observed a week earlier than the peak of T cell numbers in the tumor (Supplementary Fig. ^S4). Notably, when the expression of PD-1 in hCD4⁺ and hCD8⁺ T cells was compared between the SC and LM models, PD-1 expression increased gradually in the LM model. Overall, in the LM model, regulatory T-cell and cytotoxic T-cell numbers increased, and proliferation and PD-1 expression also increased.

Open in a separate window

Figure 3

The kinetics of human immune cell changes in tumor tissue in the humanized mouse model. (A) Subcutaneous (SC) model. (B) Liver metastasis (LM) model. Tumor immune cell frequencies and absolute numbers in the SC model (n=5) and LM model (n=5) were analyzed every week by flow cytometry using the following phenotypes: human CD45⁺(hCD45⁺), hCD45⁺hCD3⁺, hCD45⁺hCD3⁺hCD4⁺, hCD45⁺hCD3⁺hCD8⁺, hCD3⁺hCD4⁺hCD25⁺Foxp3⁺, hCD3⁺hCD8⁺IFN-γ⁺, hCD3⁺Ki67⁺, hCD4⁺ki67⁺, hCD8⁺ki67⁺, hCD4⁺PD-1⁺(hCD4⁺hCD274⁺), hCD8⁺PD-1⁺(hCD8⁺CD274⁺). Mice of the same sex and age as those in the experimental group were used as controls (n=5 mice per group). Tumor-infiltrating lymphocytes (TILs) from each treatment group were studied by flow cytometry every week after subcutaneous and intrasplenic injections.

Comparison of the immunologic characteristics in the TME changes between SC and LM models following anti-PD-L1 therapy

We examined whether anti-PD-L1 therapy was effective in our humanized SC and LM mouse models. The experimental protocol used is outlined in Fig. 4A and B. In this experiment, we observed a decrease in PD-1 expression on day 14 in the SC model and day 21 in the LM model. Aezolizumab, an anti-PD-L1 inhibitor, was injected into the humanized SC and LM mouse models from day 7 to assess the efficacy of the drug against tumors, followed by administration of the same doses on days 14 and 21. As shown in Fig. 4C and D, atezolizumab reduced gross tumor volume and metastatic burden in both models, and this difference was significant in the LM model (p<0.05). To further explore the profile of tumor-infiltrating lymphocytes (TILs) following treatment with atezolizumab, we performed H&E staining, image analysis by used ZEN Blue Software, and IHC analysis on tumor specimens from both models. First, significant differences in the number of TILs were observed in both models after treatment with atezolizumab, with observed increases in the numbers of pan T cells, helper T cells, cytotoxic T cells, and effector T cells (Fig. 4F,H) by ZEN Blue Software. Notably, in the SC model, we found that most TILs were scattered around the tumor margin, whereas in the LM model, most TILs infiltrated into the center of the tumor (Fig. 4E,G).

Open in a separate window

Figure 4

Schematic and histological analysis of anti-PD-L1 therapy in the humanized mouse models. (A) Subcutaneous model (SC model). (B) Liver metastasis model (LM model). (C) Representative images of tumor growth and quantification of tumor weight in the SC model and (D) the LM model. (E) Histological analysis of hematoxylin and eosin (H&E) staining and CD3+T cells, CD4⁺ T cells, CD8⁺ T cells, and Granzyme B⁺ cells in the SC model and (G) the LM model. (F) Statistical analysis of human immune cells in tumors by ZEN Blue Software image analysis for the SC model and (H) the LM model. Mice of the same sex and age as those in the experimental group were used as controls (n=5 mice per group). Tumor and liver tissues were collected every week after subcutaneous and intrasplenic injections and processed for histologic examination (×200) and ZEN Blue Software image analysis. Error bars indicate the SEM. n.s. not significant; ^*p<0.05; ^**p<0.01; ^***p<0.001; ^****p<0.0001; NSG, NOD SCID gamma mouse; HuNSG, peripheral blood mononuclear cell (PBMC) transplantation with NSG mouse; HuNSG+αPD-L1, anti-PD-L1 therapy was injected into the PBMC transplantation with NSG mouse; H & E, hematoxylin and eosin.

Our data suggest that there are variations in tumor response and the distribution of TILs between the SC and LM models following anti-PD-L1 therapy.

Analysis of T cell and cytokine changes in the SC and LM models after anti-PD-L1 therapy

We analyzed T cell subsets following exposure to anti-PD-L1 therapy in both models by flow cytometry (Fig. 5). hCD8⁺ T cell numbers were significantly increased in the LM model treated with atezolizumab compared with the SC model. Next, we investigated the distribution of memory hCD8⁺ T-cell subsets, specifically naïve T cells (TcNs, hCD8⁺hCD45RA⁺hCCR7⁺), effector T cells (TcEs, hCD8⁺hCD45RA⁺hCCR7^-), central memory T cells (TcCMs, hCD8⁺hCD45RA^-hCCR7⁺), and effector memory T cells (TcEMs, hCD8⁺hCD45RA^-hCCR7^-). We observed that in the LM model, the numbers and sizes of the TcCM and TcEM populations were significantly increased by atezolizumab treatment. However, in the SC model, there were no significant changes observed. Furthermore, to evaluate the role of T cell differentiation and degranulation after anti-PD-L1 therapy, we analyzed individual cytolytic proteins in CD8 T cells that expressed INF-γ⁺ and granzyme B. We observed a significant increase in hCD8⁺ T cell differentiation and degranulation following treatment with atezolizumab in tumors in the LM model compared with the SC model. Furthermore, we observed that treatment with atezolizumab resulted in a significant decrease in hCD4⁺Foxp3⁺ T cell numbers in tumor tissues and blood samples from the LM model compared with those from the SC model.

Open in a separate window

Figure 5

Statistical analysis of flow cytometry of immune cells after anti-PD-L1 therapy in the humanized mouse models on 24 days. (A) The subcutaneous (SC) model and (B) the liver metastasis (LM) model. Tumor immune cell frequencies and absolute numbers in the SC model (n=5) and LM model (n=5) were analyzed by flow cytometry using the following phenotypes: human CD45⁺(hCD45⁺), hCD45⁺hCD3⁺, hCD45⁺hCD3⁺hCD4⁺, hCD45⁺hCD3⁺hCD8⁺, hCD3⁺hCD4⁺hCD25⁺Foxp3⁺, hCD3⁺hCD8⁺IFN-γ⁺, hCD3⁺hCD8⁺Granzyme B⁺, hCD3⁺hCD8⁺hCD107a⁺, TcN (Naïve T cell, hCD8⁺hCD45RA⁺CCR7⁺), TcE (Effector T cell, hCD8⁺hCD45RA⁺CCR7^-), TcCM (Central memory T cell, hCD8⁺hCD45RA^-CCR7⁺), and TcEM (Effector memory T cell, hCD8⁺hCD45RA^-CCR7^-). Error bars indicate the SEM. n.s. not significant; ^*p<0.05; ^**p<0.01. HuNSG, peripheral blood mononuclear cell (PBMC) transplantation with NSG mouse; HuNSG+αPD-L1, anti-PD-L1 therapy was injected into the PBMC transplantation with NSG mouse;

To evaluate the functional activation of T lymphocytes after atezolizumab treatment in our humanized mouse model, we measured the concentrations of several human cytokines in plasma samples from both models. hIFN-γ, hGranzyme B, and hTGF-β but not human tumor necrosis factor (hTNF) showed significant differences in SC model (Fig. 6). hTNF, hIFN-γ, and hGranzyme B levels were increased after atezolizumab treatment, while hTGF-β levels were decreased in LM model. In addition, hTGF-β levels were significantly reduced in the LM model compared to the SC model (p<0.001).

Open in a separate window

Figure 6

Analysis of human cytokine levels after anti-PD-L1 therapy in humanized mouse models. (A) Subcutaneous model. (B) Liver metastasis model. Serum samples were collected from humanized mice after 24 days by cardiac puncture under anesthesia. Cytokine quantification was performed using commercial ELISA kits for human IFN-γ (h IFN-γ), hGranzyme B, hTNF, and hTGF-β. All samples were measured in triplicate. Error bars indicate the SEM. n.s. not significant; ^*p<0.05; ^**p<0.01; ^***p<0.001. HuNSG, peripheral blood mononuclear cell (PBMC) transplantation with NSG mouse; HuNSG+αPD-L1, anti-PD-L1 therapy was injected into the PBMC transplantation with NSG mouse; TNF, tumor necrosis factor.

Mice

NSG mice, which were established by backcrossing a null mutation of the Il2rγ locus onto the NOD.Cg-Prkdc^scid (NOD/SCID) strain, were purchased from The Jackson Laboratory. The mice were sacrificed with isoflurane euthanasia within a month of the end of the study. Experiments involving NSG mice were performed according to the guidelines of the Chonnam National University Medical School Research Institutional Animal Care Committee, and all animal experimental protocols were approved by the Chonnam National University Medical School Research Institutional Animal and Use Committee (Gwangju, Republic of Korea, CNU IACUC-H-2022-31). Our study was confirmed by ARRIVE essential 10 guidelines 2.0³².

Subcutaneous model

To establish the experimental subcutaneous (SC) model, male 5- to 6-week-old NOD-scid IL2Rgamma^null(NSG) mice were injected intravenously (i.v.) with 1×10⁷ human fresh PBMCs, and 3×10⁶ HCT116 human colon carcinoma cells suspended in 100 µL of phosphate-buffered saline (PBS) were injected subcutaneously (s.c.) into the flank of the mice on day 0 of the experiment. Tumor growth was recorded every 3 to 4 days for 4 weeks. Mice were euthanized every week, and SC tumors were removed, weighed, and processed for histologic examination; cells collected from residual tumor tissue, blood, BM, and spleen cells were subjected to flow cytometry analysis.

Liver metastasis model

To establish the experimental liver metastasis (LM) model, HCT116 cells (2×10⁶ cells/mouse) were injected into the spleen of the mice to induce colorectal liver metastasis¹. Splenectomy was performed 10 min after injecting HCT116 cells. The skin and peritoneum of the mice were sutured and used for colorectal liver metastasis studies. In our previous studies aimed at researching the liver metastasis model, we monitored the IVIS 200 Imaging system daily and have been reported^33,34. Based on these studies, we found that after performing splenectomy, the seeding and growth of cancer cells in the liver begin from day 2. After 3 days or 7 days, the mice were injected i.v. with 1×10⁷ human fresh PBMCs. The weight of each mouse was recorded every 3–4 days for 4 weeks. Mice were euthanized weekly, and liver tissue was removed, weighed, and processed for histologic examination. Cells were collected from the liver and blood for flow cytometry analysis. The metastatic burden calculated by this method (Metastatic burden (%)=(tumor area)/(liver area) X 100). The whole liver area, which consists of the left lateral lobe, left and right medial lobes, and the caudate lobe, including both the front and back surfaces, was measured and summed. And the ratio of liver/body weight calculated by this method (Liver/Body weight ratio=Liver weight/Body weight).

Histologic examination and immunohistochemistry

The excised wound tissue was fixed in 10% neutral-buffered formalin and then dissected and embedded in paraffin. Tissue blocks were sectioned at a 3 µm thickness using a microtome and stained with hematoxylin and eosin (H&E). Unstained slides from each tissue block were immunostained with specific antibodies using a Bond-Max Autostainer (Leica Microsystems, Buffalo Grove, IL, USA). Antibodies against the following factors were used: CD7 (1:100 dilution; catalog no. M7255; DAKO, Glostrup, Denmark), CD4 (1:100 dilution; catalog no. NCL-L-CD4_368; Novocastra, Newcastle, UK), CD8 (ready to use; catalog no. PA0183; Novocastra), Granzyme B (1:50 dilution; catalog no. M7235; DAKO), CD79a (1:300 dilution; catalog no. 179 M-95; Cell Marque, Rocklin, CA, USA), CD56 (1:400 dilution; catalog no. M7304; DAKO), CD68 (1:100 dilution; catalog no. MO814; DAKO), and human leukocyte antigens (HLAs) (1:300 dilution; catalog no. ab70328; Abcam, Cambridge, UK). The slides were evaluated by an experienced pathologist (KHL) and photographed using the Nikon DS-Ri2 system (Nikon Japan, Tokyo, Japan). The images were processed using ZEN Blue Software.

Human peripheral blood mononuclear cells

All participants signed an ethical consent form approved by the Institutional Review Board of Chonnam National University Hwasun Hospital (Hwasun, Republic of Korea, CNUHH-2021–045). Human PBMCs were obtained from healthy volunteers and were isolated from blood samples by density-gradient centrifugation with Ficoll-Hypaque (d=1.077 gm/mL, Lymphoprep™, Axis-Shield, Oslo, Norway. 1114545) and washed with PBS. NOD-scid IL2Rgamma^null(NSG) mice were injected i.v. with 1×10⁷ human PBMCs suspended in 100 µL of normal saline.

Cell preparation

The mice were sacrificed with isoflurane euthanasia within a month of the end of the study, which is before developing graft-versus-host disease (GVHD)³⁵. lood was collected from the heart immediately following euthanasia, and serum was frozen in a -80 °C deep freezer for future analyses. Liver, lymph node (LN), spleen, and BM specimens were processed into single-cell suspensions. Tumors were extracted and weighed, and was fixed in 10% formalin (Core fix) for histological analyses. The remaining tumor sections were minced and digested using 4–8 mg of collagenase D (Roche) per tumor in PBS for 1.5–2 h at 37 °C³⁶. The liver was minced into 1-mm pieces and digested using collagenase IV for 30 min at 37 °C. The digested liver extract was filtered through a 70-μm cell strainer and centrifuged at 500×g for 5 min. The resulting cell pellet was resuspended in 10 mL of 35% Percoll containing 100 U/mL heparin and centrifuged at 700×g for 15 min at room temperature. The cell pellet containing the leukocytes was collected and resuspended in 3 mL of red blood cell lysis solution (155 mM NH₄Cl, 10 mM KHCO₃, 1 mM EDTA, 170 mM Tris, pH 7.3). After incubation for 3 min on ice, the cells were washed twice in RPMI 1640 containing 5% FBS³⁷. The BM and spleen cells were prepared by standard methods and were incubated with ammonium chloride-potassium (ACK) lysis buffer (Life Technologies) for 10 min at room temperature with gentle mixing. For mice in which lymph nodes were not visible by the naked eye, small tissue samples from LN sites were processed for flow cytometry analyses to assess the presence of human lymphocytes.

Flow cytometry

For surface staining, cells were washed twice in fluorescence-activated cell sorting (FACS) buffer (1X PBS w/2% FBS, 1 mM EDTA, and 0.1% sodium azide), incubated with Fc blocking reagent (Miltenyi Biotec), and stained with fluorescence-conjugated antibodies in brilliant stain buffer (BD Horizon™, Catalog: 566349). For intracellular staining, cells were labeled with surface markers as previously described and then fixed and permeabilized with the Transcription Factor Buffer Set (BD Pharmingen™, Catalog: 554714). A 5-antibody panel was used for this study. The human T-cell panel (6 colors) consisted of mouse anti-human antibodies against CD45 (clone: HI30, catalog: 564105), CD3 (clone: UCHT1, catalog: 561807), CD4 (clone: RPA-T4, catalog: 561841), CD8 (clone: RPA-T8, catalog: 555367), CD274 (PD-1) (clone: EH12.1, catalog: 563076), and Ki-67 (clone: B56, catalog: 562899). The human B- and NK-cell panel (4 colors) consisted of CD45 (clone: HI30, catalog: 564105), CD3 (clone: UCHT1, catalog: 561807), CD19 (clone: HIB19, catalog: 562440), and CD56 (clone: CMSSB, catalog: 12–0567-42). The human macrophage panel (3 colors) consisted of CD45 (clone: HI30, catalog: 564105), CD68 (clone: Y1/82A, catalog: 562117), and CD163 (clone: GHI/61, catalog: 562670). The human regulatory T-cell panel (5 colors) consisted of CD45 (clone: HI30, catalog: 564105), CD3 (clone: HIT3a, catalog: 564713), CD4 (clone: RPA-T4, catalog: 561841), CD25 (clone: M-A251, catalog: 555431), and Foxp3 (clone: 236A/E7, catalog: 72–5774-40). The human cytotoxic T-cell panel (4 colors) consisted of CD45 (clone: HI30, catalog: 564105), CD3 (clone: UCHT1, catalog: 561807), CD8 (clone: RPA-T8, catalog: 555367), and IFN- γ (clone: B27, catalog: 554702). The human monocyte and dendritic cell (DC) panel (3 colors) consisted of CD45 (clone: HI30, catalog: 564105), CD14 (clone: M5E2, catalog: 561712), and CD11c (clone: -ly6, catalog: 562561), all purchased from BD Bioscience and eBioscience. The human cytotoxic T-cell panel (5 colors) consisted of CD3 (clone: UCHT1, catalog: 561807), CD4 (clone: RPA-T4, catalog: 557871), CD8 (clone: RPA-T8, catalog: 555367), CD45RA (clone: L48, catalog: 337186), and CCR7 (clone: 3D12, catalog: 560548). Human cytolytic proteins in cytotoxic T cells (8 colors) were stained with CD3 (clone: UCHT1, catalog: 561807), CD4 (clone: RPA-T4, catalog: 557871), CD8 (clone: RPA-T8, catalog: 555367), CD45RA (clone: L48, catalog: 337186), CCR7 (clone: 3D12, catalog: 560548), CD107a (clone: H4A3, catalog: 560664), IFN-γ (clone: B27, catalog: 554702), and granzyme B (clone: GB11, catalog: 563389). The cells were further permeabilized using the Fixation/Permeabilization Kit (BD Bioscience) and stained for Ki-67, CD68, Granzyme B, and INF-γ. For staining of Foxp3, cells were permeabilized using the anti-human Foxp3 staining Set-PE (eBioscience™). All cells were analyzed using a BD FACSCanto II flow cytometer (BD Biosciences) (Fig. ^S5, ^S6, ^S7, ^S8). The data were analyzed using FlowJo v10.6.2. software (BD Biosciences).

Measurement of in vivo release of human cytokines

Serum samples were collected from the humanized mice every week for 4 weeks by cardiac puncture under anesthesia. Cytokine quantification was performed using commercial ELISA kits for human IFN-γ (BD Bioscience, San Diego, CA, USA., 555142) and hIL-10 (BD Bioscience, San Diego, CA, USA., 555157), hTGF-β1 (ELISA Duo-set; R&D Systems, Inc., Canada, USA., DY240-05), and Granzyme B (R&D Systems, Inc. Canada, USA., DY2906-05) according to the manufacturer’s instructions. All samples were measured in triplicate, and the absorbance was read at 450 nm in an ELISA reader.

Statistical analysis

Data were analyzed using paired or unpaired t tests, and one-way ANOVA was performed to compare groups; a p value<0.05 was considered to indicate statistical significance. All statistical analyses were performed using SPSS v.23.0 (IBM, Armonk, NY, USA) and/or GraphPad Prism v.9 software (GraphPad Software, San Diego, CA, USA).