Construction and application of a humanized mouse model for prostate cancer in immunotherapy

Mice and cell lines

Four to seven-week-old NOD-PrkdcscidIl2rgtm1/Vst (NPG) SCID mice were purchased from Biocytogen (Beijing, China), while B-NDG-B2m mice were obtained from Vitalstar (Beijing, China). All mice weighed 18 ± 2 g and were housed in the SPF barrier facility of the Fourth Military Medical University (FMMU). The human PCa cell lines LNCaP-FGC, 22RV1, C4-2, DU145, and PC3 were cultured in RPMI-1640 medium supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin. The cells were maintained in a constant-temperature incubator at 37 °C and 5% CO₂, with the culture medium being refreshed every alternate day.

Establishment of Hu-PBMC-CDX models and Hu-PBMC-PDX models

PBMCs were isolated from healthy adult donors (20–45 years old, sex recorded) using Lymphoprep (Tbdscience, Tianjin, China). Donors were screened to exclude individuals with active infections (e.g., HIV, hepatitis B/C), a history of autoimmune or chronic inflammatory diseases, recent (≤ 6 months) use of immunosuppressive medications, or a history of malignancy. PBMCs (1 × 10⁷) were transplanted into NPG or B-NDG-B2m mice via the tail vein. The construction was deemed successful when the proportion of hCD45⁺CD3⁺ T cells in mouse blood exceeded 25% [19, 20]. PBMCs were cryopreserved using a standard protocol: cells were resuspended in cryopreservation solution (Procell, Wuhan, China), frozen in a controlled rate, and thawed rapidly in 37 °C water bath. Post-thaw cells were washed and resuspended in medium, with viability > 90% confirmed by trypan blue exclusion before transplantation.

HLA-class-I alleles (HLA-A, -B, and -C) of 22RV1 cell and PBMCs were analyzed by Organ Transplantation Research Institute of Chinese People's Liberation Army (Supplementary Table 1). Donors with at least partial HLA-class-I allele matching (≥ 50% compatibility) were prioritized for experiments to minimize HLA mismatch-related tumor rejection [21].

5 × 10⁶ 22RV1 cells were injected into the right axilla of humanized mice subcutaneously to establish hu-PBMC-CDX models. One PDX model (B45354) was established by our laboratory from an androgen-independent PCa patient; a high metastatic potential was observed [22]. Tumor volume was calculated as [length × width²]/2.

Both hu-PBMC-CDX and hu-PBMC-PDX models were randomly assigned into 4 groups, followed by intraperitoneally administration (Phosphate-buffered saline, twice per week for three consecutive weeks, n = 5; pembrolizumab (Merck & Co. Inc), 10 mg.kg⁻¹ per injection, twice per week for three consecutive weeks, n = 5; and docetaxel (MedChemExpress, HY-B0011), 10 mg.kg⁻¹ per injection, once per week for three consecutive weeks, n = 5; docetaxel and pembrolizumab, n = 5.

Flow cytometry

Peripheral blood, spleen, and bone marrow-derived single-cell suspensions were prepared and then incubated with BD Pharmingen™ FITC mouse anti-human CD45 (clone HI30) and BD Pharmingen™ PE mouse anti-Human CD3 (clone UCHT1). Additionally, PBMCs were also incubated with BD Horizon™ BV510 mouse anti-Human CD45 (563,204), eBioscience™ PE anti-Human CD8a (2,504,504) and BioLegend™ Pacific Blue Granzyme B (372,218). Cell collection was performed using FC500 (Beckman Coulter, Miami, FL. USA), and data analysis was performed using the FlowJo software (TreeStar, San Carlos, USA).

Immunohistochemistry and immunofluorescence microscopy

Immunohistochemistry was performed using the following antibodies: anti-human CD45 (ab8216), CD4 (ab183685), and CD8 (ab108343) from Abcam. Images were acquired and analyzed using an inverted microscope (OLYMPUS, Tokyo, Japan). Immunofluorescence was conducted with anti-human CD45 (Clone HI30), CD8 (SK1), and CD4 (QA18A35) from BioLegend. Nuclei were counterstained with DAPI (BioLegend).

Statistical analysis

All data are presented as the mean ± standard error (SE) and were analyzed using GraphPad Prism (GraphPad Software, San Diego, USA). For multi-group comparisons, data were analyzed using one-way ANOVA with Tukey’s post hoc test. Normality and homogeneity of variances were verified prior to analysis. P values < 0.05 were considered statistically significant, with adjustments for multiple comparisons performed using Tukey’s HSD method.

Construction of the Hu-PBMC mouse model

To establish a humanized mouse model that can faithfully recapitulate the immune system of cancer patients, we transplanted 1 × 10⁷ PBMCs isolated from healthy donors into adult SCID mice via the tail vein. The purity of these PBMCs was confirmed to be greater than 90%, ensuring that only the highest quality cells were used in this critical procedure (Fig. 1a, b). 3 weeks after the initial transplantation, peripheral blood samples were collected for flow cytometry analysis to evaluate the proportion of human CD45⁺CD3⁺ T cells. Generally, successful reconstitution requires a human T cell percentage greater than 25% in mice. Dynamic monitoring revealed a reconstitution success rate of 60% (6/10) by the third week (Fig. 1c). Furthermore, we performed functional assays on the reconstituted human T cells within the mice, which showed Granzyme B positivity, indicating that the reconstituted human T cells in the humanized mouse model are functional (Supplementary Fig. 1).

Notably, all Hu-PBMC mice developed graft-versus-host reactions (GVHDs) at approximately week 6, which were caused by cytokine secretion from human immune cells, particularly CD8⁺ T cells. Prominent symptoms included weight loss, which commonly commences at 20–30 days post-transplantation and became severe by day 40. GVHDs in humanized mice ultimately result in mortality within 8–10 weeks (Fig. 1d). In addition to this, affected mice may also exhibit a hunching posture that suggests discomfort or pain. Alopecia, or hair loss, is another common sign of GVHDs in humanized mice (data not shown). To further investigate, we conducted a correlation analysis to determine the potential association between the level of peripheral T cell and the severity of GVHD response. Specifically, we measured the percentage of body weight loss on day 40 relative to day 0 and assessed peripheral T cell levels. The analysis revealed a strong positive correlation between peripheral immune reconstitution and GVHD severity (r = 0.860, p = 0.0014) (Fig. 1e). Therefore, selecting a humanized mouse model with a moderate proportion of hCD45⁺CD3⁺ T cells is optimal for experiments to balance immune functionality and minimize GVHD-related complications.

In summary, while GVHD-induced mortality limits the therapeutic window of Hu-PBMC mice, this model remains invaluable for preclinical studies of short-term immunotherapy due to its rapid and straightforward reconstruction process.

The reconstituted immune system is present in various organs

To fully confirm the widespread distribution of the reconstituted immune system beyond its presence in peripheral blood, we analyzed T cell levels across multiple organs. This investigation aimed to deepen our understanding of how the newly formed immune system interacts with and integrates into various bodily systems. 6 weeks post-reconstitution, the mice were humanely sacrificed, and their tissues were systematically analyzed. The results obtained from flow cytometry revealed comparable levels of T cells in the blood and spleen, while lower levels were observed in their bone marrow (Fig. 2a). To further investigate the distribution of immune cells, immunohistochemical staining was performed on spleen and bone marrow samples. The analysis showed substantial infiltration of CD45⁺ immune cells and CD3⁺ T cells in these organs, with a quantitatively higher presence in the spleen compared to the bone marrow (Fig. 2b, c). The distribution of immune cells was then assessed in additional organs, including the lungs, colon, liver, and heart. Notably, the lungs, liver, and colon demonstrated significant infiltration of CD45⁺ immune cells, including CD4⁺ and CD8⁺ T cells, whereas the heart exhibited minimal immune cell infiltration (Fig. 2d, e). Interestingly, this organ-specific variability in immune cell infiltration was observed not only in humanized mice but also in normal mice, suggesting that inherent organ-specific factors influence immune cell distribution. These findings collectively demonstrate that the reconstituted human immune system in humanized mice is robustly established and naturally distributed across multiple organs. This comprehensive integration highlights the utility of the model for investigating systemic immune responses in preclinical studies.

Exploration of factors affecting the humanized mouse model

The successful construction of humanized mice is influenced by multiple factors. To minimize donor variability and improve model consistency, we first compared the reconstitution levels of Hu-PBMCs from different donors. Dynamic monitoring of four donor-derived PBMCs showed varying reconstitution trends, with the highest reconstitution observed in donor 1 and the lowest in donor 4 (Fig. 3a, b). These results indicate that the donor source plays a significant role in the success of humanization. Furthermore, while all of the humanized mice experienced weight loss due to GVHDs 3–4 weeks after PBMC transplantation, each donor-derived mouse exhibited a distinct trend. GVHDs caused by donors 1 and 4 resulted in the most significant and least significant weight loss, respectively (Fig. 3c). The findings indicate that mice with high reconstitution levels are more prone to severe GVHDs. Additionally, the survival monitoring of mice revealed that those derived from donor 4 had the longest median survival of 46 days (Fig. 3d). Therefore, it appears feasible to cryopreserve donors with optimal reconstitution effects for future stable reconstitutions. To this end, we subsequently compared the reconstitution levels of fresh and cryopreserved PBMCs. The results indicate comparable reconstitution efficacy between cryopreserved/resuscitated PBMCs and fresh PBMCs (5/6 vs. 6/6), providing a theoretical basis for subsequent stable construction of humanized mice (Fig. 3e).

Furthermore, we conducted investigation into the impact of gender on reconstitution as subsequent PCa transplantation experiments necessitate the use of male mice. PBMCs (1 × 10⁷) were transplanted into mice of the same strain, with both female and male subjects included. The results of dynamic monitoring showed that male mice exhibited lower rates and levels of reconstitution success (Fig. 3f). To tackle this issue, we escalated the transplantation dosage of PBMCs to 2 × 10⁷ and 3 × 10⁷ cells respectively. The findings demonstrated that increasing the transplantation dose of PBMCs could improve the levels of hCD45⁺ immune cells in male mice (Fig. 3g). However, due to the positive correlation between the degree of peripheral immune reconstitution and the severity of GVHD, we chose to use dosage of PBMCs to 2 × 10⁷ as reconstruction dosage for male mice.

Finally, we investigated the impact of mouse strain on reconstitution. We transplanted the same donor-derived PBMCs into two common SCID mice (NPG and B-NDG-B2m). The levels of peripheral hCD45⁺ cells, body weight, and survival rates were recorded twice a week. As shown in Fig. 3h, reconstituted NPG and B-NDG-B2m mice had comparable reconstruction levels, while B-NDG-B2m mice had a relatively longer study window. Survival monitoring confirmed that the reconstituted NPG mice had a limited lifespan of approximately 40 days, while B-NDG-B2m mice exhibited a relatively prolonged survival time (> 60 days) (Fig. 3i). Furthermore, it was observed that the body weight of reconstituted NPG mice declined rapidly due to GVHDs, whereas the B-NDG-B2m mice remained comparatively stable during this period (Fig. 3j). Further histopathological analysis of skin, liver, and colon tissues harvested at 4–5 weeks post-reconstitution confirmed significantly more severe tissue damage in NPG mice compared to B-NDG-B2m mice (Supplementary Fig. 2). These findings suggest that the successful construction of humanized mice is influenced by multiple factors, including the donor and dosage of PBMCs, as well as the strain and gender of recipient mice, but not by the donor gender.

Establishment and characterization of a Hu-PBMC mouse model for prostate cancer

After successfully establishing a stable humanized mouse model, we selected appropriate human PCa cell lines to generate a xenograft model of PCa, referred to as a CDX. Given that PD-L1 functions as a biomarker for immunotherapy and is closely linked to the response to immunotherapy in other types of cancer, we analyzed the PD-L1 expression in several commonly used PCa cell lines to identify suitable candidates for subsequent investigation. Results showed that the 22RV1 cell line exhibited the highest PD-L1 expression (Fig. 4a and Supplementary Fig. 3), making it the optimal choice for subsequent in vivo studies.

To construct the CDX model, 5 × 10⁶ 22RV1 cells were transplanted into the right flank of mice one week after PBMC transplantation (Fig. 4b). We observed variable tumor growth rates among humanized mouse models reconstituted with PBMCs from different donors, which could be attributed to human leukocyte antigen (HLA) mismatches (Fig. 4c). Further study confirmed that HLA-class-I alleles (HLA-A, -B, and -C), as well as serotypes, matching between PBMCs and 22RV1 cells inhibited the growth of tumor (Fig. 4d). Consequently, mice reconstituted with HLA-matched PBMC donors were selected for experimental analyses. To exclude confounding effects of intrinsic donor immune activity, we implanted HLA-mismatched tumors (PC3/DU145) in mice reconstituted with different donors (one originally HLA-matched to 22RV1, the other mismatched). Tumor growth showed no significant differences between donors, and CD8⁺ T cell infiltration remained equivalent (Supplementary Fig. 4), confirming HLA-matching as the primary driver of tumor suppression.

At week six, all mice were euthanized, and tumor samples were harvested to analyze the TIME. Immunofluorescence revealed sparse infiltration of CD45⁺ immune cells and CD8⁺ T cells in both the central and peripheral tumor regions (Fig. 4e). Flow cytometry corroborated these findings, demonstrating low levels of tumor-infiltrating lymphocytes (Fig. 4f). Additionally, a substantial proportion of CD8⁺ T cells within the tumor express the inhibitory receptor PD-1, highlighting the "cold" tumor phenotype characteristic of the humanized mouse-derived PCa model (Fig. 4g). To further validate the model's ability to replicate the TIME of clinical PCa, immunohistochemical analyses were conducted on tumor specimens from three patients. Similar to the 22RV1-derived CDX models, the three clinical specimens also exhibited certain levels of PD-L1 expression, respectively (Supplementary Fig. 5). Further, the immune cell infiltration patterns observed in these clinical specimens also mirrored the “cold” tumor and microenvironment seen in the mouse model (Fig. 4h). In summary, despite the challenges of achieving complete HLA matching between PBMCs and implanted tumor cells, the co-engrafted humanized mouse model partially recapitulates the TIME of clinical PCa. This model holds significant potential for translational immuno-oncology research and the development of novel therapeutic strategies.

Docetaxel and pembrolizumab synergistically inhibit prostate cancer growth in CDX model

Given the limited clinical efficacy of ICIs monotherapy, we conducted combination therapy in the CDX model to assess whether chemotherapy could potentiate the antitumor activity of ICIs. We opted to evaluate the effectiveness of combination therapy by merging docetaxel, which is the primary clinical treatment for mCRPC, with pembrolizumab (a representative ICI in the clinic) [23]. Firstly, matching PBMCs were transplanted into the mice. 1 week later, 22RV1 cancer cells were inoculated into these humanized mice, which were subsequently treated with docetaxel, pembrolizumab, or a combination of the two drugs. Phosphate-buffered saline was used as the therapeutic control (Fig. 5a). The results demonstrated that docetaxel or pembrolizumab monotherapy did not significantly impede tumor growth; however, its efficacy was augmented when used in combination, as evidenced by reduced tumor weight (Fig. 5b, c). Overall, these results provide valuable insights into potential avenues for improving cancer treatments and highlight the importance of exploring combination therapies as a means of enhancing their effectiveness.

Further mechanistic studies demonstrated that combination therapy significantly increased the infiltration of CD45⁺ immune cells into the tumor microenvironment. Detailed analysis revealed a similar trend for CD4⁺ and CD8⁺ T cell populations, highlighting the role of combination therapy in promoting the recruitment and activation of functional immune cells (Fig. 5d–g). Notably, although docetaxel monotherapy did not increase the overall number of CD8⁺ T cells, further analysis revealed an increased proportion of CD8⁺ Granzyme B⁺ T cells within the tumor. To functionally validate this observation, we utilized two complementary in vitro systems: (1) HLA-matched 22RV1-CD8 ⁺ T cell co-cultures and (2) patient-derived organoid (PDO) co-cultures with CD8⁺ T cells. Both models consistently demonstrated that docetaxel treatment significantly enhanced CD8⁺ T cell cytotoxicity, as evidenced by elevated LDH release in 22RV1 co-cultures and reduced tumor viability in PDOs (Supplementary Fig. 6).

In summary, the combination of docetaxel and pembrolizumab effectively inhibited tumor growth by enhancing immune cell infiltration and activity. These results support the development of combination therapies as a promising strategy to improve the efficacy of cancer treatments.

Docetaxel and pembrolizumab synergistically inhibit prostate cancer growth in PDX model

Considering the limitations of CDX model in simulating the tumor microenvironment, we further transplanted PDX model into the humanized mouse model for combination therapy. The PDX model, previously developed in our laboratory, exhibits multi-organ metastatic potential and was specifically selected for this study due to its characteristics of neuroendocrine prostate cancer (NEPC) [22]. These traits were confirmed by positive expression of synaptophysin (SYP), a neuroendocrine carcinoma marker, consistent with the original tumor tissue, and negative expression of chromogranin A (CGA) in the PDX model (Fig. 6a).

To construct the humanized PDX model, cryopreserved PBMCs with optimal reconstitution capabilities were transplanted into mice approximately 2 weeks after PDX engraftment. Following model establishment, the mice were treated with docetaxel, pembrolizumab, or a combination of both, while phosphate-buffered saline served as the control. While combination therapy significantly reduced tumor volumes (p < 0.05), both docetaxel and pembrolizumab monotherapies showed modest antitumor effects. The difference between pembrolizumab monotherapy and combination therapy was not statistically significant (p > 0.05) (Fig. 6b). To elucidate the mechanisms underlying the observed synergy, we conducted a microenvironment analysis across different treatment groups. As illustrated in Fig. 6c, d, immunofluorescence results demonstrated that pembrolizumab monotherapy modestly increased CD45⁺ immune cell and CD8⁺ T cell infiltration, while combination therapy elicited a more pronounced immune cell recruitment and activation within the tumor microenvironment.

It is worth noting that the incorporation of docetaxel into immunotherapy has shown promising results in augmenting the infiltration of immune cells within tumor microenvironments. This, in turn, can lead to a remarkable shift in the phenotype of patient-derived xenografts (PDX) from an immune-desert (“cold” tumor) to an immune-inflamed (“hot” tumor). This phenotypic shift is critical, as it enhances tumor susceptibility to immune attack by increasing immune cell abundance and activity within the TIME. These findings underscore the potential of docetaxel to augment the efficacy of immunotherapeutic strategies, paving the way for more effective combination treatments in cancer therapy.

Conflict of interest

The authors declare no conflict of interest.

Consent to participate

Written informed consent was obtained from the participants and the patient with PCa, respectively.

Ethics approval

All animal experimental protocols were approved by the Institutional Animal Care and Use Committee of FMMU (Protocol No. 20220705). Fresh peripheral blood and PCa tumor tissue were collected from the Xi’jing Hospital in Xi'an and approved by the Medical Ethics Committee (KY20203128-1).