Isolation of human CD34+ cells from CB and generation of hu-CB-BRGS mice (Additional file 1) were performed as described previously [19, 20]. Investigators were blinded from donor identities, and the studies were performed in compliance with University of Colorado Institutional Review Board.

The MDA-MB-231 cell line purchased from American Type Culture Collection was used to model TNBC. Cells were authenticated by the Barbara Davis Center for Childhood Diabetes Core and tested for absence of mycoplasma every 3 months. Cells used for in vivo studies were passaged 3 times prior to injection into mice. Cells were harvested during exponential growth and resuspended in a 1:1 mixture of DMEM supplemented with 10% FBS, 1% PenStrep and 1% non essential amino acids and Matrigel (BD Biosciences). Five to ten million cells per mouse were injected subcutaneously into both flanks using a 23-gauge needle [25].

For the colorectal models, the PDX tumors were generated and grown in nude mice prior to implantation in hu-CB-BRGS mice, as previously described ([25, 26] and Additional file 1).

Upon verification of human T-cell chimerism in the hu-CB-BRGS mice, tumors were implanted into both hind flanks of 16–21-week old humanized mice or into non-humanized BRGS controls for MDA-MB-231 and MSI-H. When the average tumor size reached a volume of approximately 150–300 mm³, experimental treatments were begun. Nivolumab, anti-PD-1, was administered intraperitoneally (i.p.) at approximately 30 mg/kg twice per week. For combination therapy studies, OKI-179, a histone deacetylase (HDAC) inhibitor, was administered via oral gavage at 30 mg/kg three times per week and nivolumab was administrated i.p. 30 mg/kg d0–10, then 15 mg/kg after d10 once a week. Control mice were left untreated. Twice weekly, mice were monitored for signs of toxicity and tumors were measured using calipers, shaving mice when necessary, to better visualize tumors (Additional file 1).

Evaluation of human chimerism in the blood is described in Additional file 1.

Euthanasia was performed either at end of study or when total tumor burden approached 3000 mm³ (typically 20–45 d). The detailed timepoints for each mouse in each experiment are provided in Additional file 2. Blood was collected from the heart immediately following euthanasia and sera were frozen for future analyses. Lymph nodes (LNs), spleen, and bone marrow (BM) tissues were processed into single-cell suspensions as described previously [19, 20]. Tumors were extracted, weighed and a small portion was fixed in formalin for histological analyses. The remaining tumor sections were minced and digested for 25 min at 37 °C in Liberase DL (50 μg/ml) containing RPMI serum-free medium. Following incubation, the tissue was strained through a 100 μm filter, and the filter was washed with 5–10 ml of complete RPMI medium containing 10% heat-inactivated FBS, 100 U/ml penicillin, 100 μg/ml streptomycin and non-essential amino acids to stop digestion. The tumor cells were resuspended in fresh DNAase-containing Iscove’s medium with 5% heat-inactivated FCS, Hepes and Glutamax. Cells from the spleen and LNs were counted using a hemocytometer and samples from all tissue preparations were collected for a set time on the flow cytometer to generate a relative cell count.

As described previously, single cell suspensions were incubated with fluorescently-labeled antibodies (Abs, see Additional file 3) to evaluate immune parameters described in Additional file 1.

Upon dissection, a portion of the tumor was fixed in formalin and paraffin-embedded for multispectral imaging on Vectra 3.0 Automated Quantitative Pathology Imaging System (Perkin Elmer, described in Additoinal file 1).

Concentrations of human IgM and IgG in the sera of humanized mice were measured as described previously using a monoclonal mouse anti-human IgM or IgG-based ELISA to avoid detection of the anti-PD-1 drug in the sera [27].

The tumor growth inhibition (TGI) was calculated from the average tumor volume in treated (Vt) and vehicle control (Vvc) groups, according to the following equation: TGI = 100 × (Vt_final - Vt_initial)/(Vvc_final -Vvc_intial). Statistical significance of immune measurements were assessed using Prism software (GraphPad software) with a two-tailed Student’s t-test of equal variance or Welch’s correction when appropriate. P-values less than 0.05 were considered significant.

We have previously reported that BRG(S) mice are efficiently reconstituted with human CD34+ cells from CB shortly after birth [19] and that develop, with time, a significant population of human B and T cells [20, 24] (Fig. 1a). In this study, we found T-cells from hu-CB-BRGS mice expressed high levels of PD-1, levels that increased with age (Fig. 1b).Fig1Fig. 1

Anti-PD-1 therapy inhibits growth of TNBC xenograft in hu-CB-BRGS model. a, PBMCs of humanized mice were analyzed pre-tumor engraftment for expression of mouse (mCD45) and human (hCD45) hematopoietic markers (left panels). The hCD45+ cells were analyzed for expression of T-cell (CD3), B-cell (CD20), and CD8+ T-cell subsets (middle, right panels). A mixture of CB PBMCs and mouse spleen cells served as a positive staining control (top panels). b, Expression of PD-1 on T-cells (hCD45+ CD3+) in PBMCs from a control CB (top) and a 16-week old humanized mouse (bottom). The MFI of PD-1 on CD4+ (black) and most CD8+ (red) T-cells increase with age (right, CD4: P = 0.04; CD8: ns). Chimerism and PD-1 staining results represent > 100 hu-CB-BRGS mice. c, Growth curves of TNBC MDA-MB-231 tumors in nivolumab-treated (blue line) or untreated (black line) hu-CB-BRGS mice (left, n = 6 tumors/3 mice per group) and non-humanized BRGS mice (right, n = 2 tumors per group). P-values: * < 0.05

To characterize immune responses to the TNBC cell line in the hu-CB-BRGS model, we engrafted hu-CB-BRGS mice with MDA-MB-231 to examine hCD45+ cells 11 and 21 days post-treatment. Prior to tumor implantation, mice with similar levels of human hematopoietic and T-cell chimerism in the blood were stratified into treatment groups to minimize engraftment effects (Additional file 4: Figure S3a). Again, we observed tumor growth inhibition in the anti-PD-1-treated group compared to untreated controls (Fig. 2a). We next examined both peripheral immune tissues (LN, spleen, BM) as well as tumor-infiltrating leukocytes (TILs) in untreated and anti-PD-1-treated hu-CB-BRGS mice by flow cytometry to evaluate human T, B and myeloid subsets, which have been shown to differ from the human chimerism in the blood and to change over time [20].Fig2Fig. 2

Human T-cell chimerism in lymph organs and tumors of TNBC mice model a, Growth of MDA-MB-231 tumors in hu-CB-BRGS mice treated (blue line) or untreated (black line) with nivolumab is shown. b, Representative flow plots illustrate gating strategy used to determine human T- and B-cell chimerism. Mouse (mCD45) and human (hCD45) Abs were used to detect species-specific hematopoietic cells (top panels) and hCD3 and hCD20 Abs identified T and B cells, respectively, among the hCD45+ gated cells (bottom panels). A mixture of CB (hPBMC) and mouse spleen cells served as positive staining controls and representative flow plots are from TILs of control or anti-PD-1 treated hu-CB-BRGS mice. 10 tumors were analyzed on d11 and d21 in the control group; 4 tumors on d11 and 6 tumors on d21 were analyzed in the anti-PD-1-treated group. Human immune cell (c), T-cell (d) and CD4 T-cell (e) chimerism in lymph organs and tumors of hu-CB-BRGS mice. CD4+ T-cells are gated on hCD45 + hCD3+. CD8 T-cells are calculated as CD4-negative T-cells. Left panels show percentages in lymph organs, middle panels show frequencies and right panels show relative cell numbers in TILs. Each point represents data from the indicated organ from an individual hu-CB-BRGS mouse that was either untreated (−, black) or treated with nivolumab (+, blue) for 11 or 21 days. P-values: * < 0.05, **** < 0.0001

Although human B-cells (CD19+, CD20+) were well represented in the LNs and spleens of humanized mice, there were very few in the TILs, and treatment did not affect their representation (Fig. 4a). However, the presence of human IgM and IgG in the sera suggested B-cells are functional in these mice (Fig. 4a).Fig4Fig. 4

Few B-cells and many myeloid human cells in TILs of TNBC-engrafted hu-CB-BRGS mice. a, No change observed in human B-cell frequencies (left, middle panels) or function (hIgM and IgG concentrations in serum, right panel) in tumor-bearing, humanized mice with nivolumab treatment. b, Decreased human myeloid cell frequencies in TNBC tumors treated with nivolumab. The upper plots depict the flow cytometry gating strategy for identification of the CD4 + CD3- myeloid population among hCD45+ cells from hPBMC (staining control), and indicated organs of a humanized mouse engrafted with TNBC cell line and data from each mouse is shown in upper right graphs. The lower plots show flow cytometry gating (left) for myeloid populations. The myeloid cells are also identified as hCD45+, CD3-, CD19-, (CD14+, CD33+, CD11b + or CD11c+) and the frequency of the HLA-DR^lo/− population among these cells is shown. Each dot represents data from the indicated organ from an individual humanized mouse that was either untreated (−, black) or treated with nivolumab (+, blue) for 11 or 21 days, as indicated. P-values: * < 0.05

In addition to altering immune cells, immunotherapy drugs may also influence the cancer cells or tumor microenvironment (TME), changes that may be vital in the drug’s mechanism of action. Thus, we determined the expression of hPD-L1, HLA-A,B,C (MHC Class I) and HLA-DR (MHC Class II) on TNBC tumor cells using flow cytometry. To distinguish the human tumor cells, we gated out the human and mouse CD45+ cells and analyzed the non-hematopoietic cells expressing HLA-A,B,C and EpCam (Fig. 5a). HLA-DR, but not HLA-A,B,C, expression was higher on the anti-PD-1 treated tumors at day 21 than controls, suggesting nivolumab treatment may improve antigen recognition (Fig. 5b and c). The human hematopoietic cells also expressed these molecules (Fig. 5a). No differences were observed in PD-L1 expression on the tumor cells over time or with treatment; however, there was a significant increase at day 11 followed by a decrease in PD-L1 expression at day 21 on the human CD45+ cells in the TILs from treated mice (Fig. 5d).Fig5Fig. 5

Human PD-L1, MHC class I and class II expression on MDA-MB-231 tumors excised from hu-CB-BRGS. a, Representative flow cytometric analyses for expression of HLA-A,B,C (MHC class I), HLA-DR (MHC class II) and hPD-L1 in digested tumor cell suspensions. The tumor cells are defined as mCD45-, hCD45-, HLA-A,B,C⁺, Epcam⁺. The mouse (mCD45) and human (hCD45) hematopoietic cells serve as negative and positive control, respectively. The MFI of HLA-A,B,C (b), HLA-DR (c) and PD-L1 (d) staining is plotted for each tumor. Expression of HLA molecules is normalized to the positive control stained at the same time. Each dot represents data from a tumor from a humanized mouse that was either untreated (−, black) or treated with nivolumab (+, blue) for 11 or 21 days, as indicated. P-values: * < 0.05

HDAC inhibitors (HDACi) have been shown to increase tumor immunogeneicity and improve anti-tumor immune responses in several cancers, including CRC models [36–41]. Thus, we queried whether a Class I selective HDACi drug (OKI-179) would augment the anti-tumor response in combination with nivolumab in the TNBC hu-CB-BRGS model. Interestingly, the combination of OKI-179 and nivolumab significantly inhibited tumor growth (TGI = 77%) more than nivolumab alone (TGI = 45%, Fig. 6a) as also seen in the individual tumor growth curves (Additional file 4: Figure S6a). At day 10, the dose of nivolumab was reduced to allow for detection of combination effects. Immunological phenotyping did not demonstrate enhanced human or T-cell numbers in the TILs (Fig. 6b), although samples were collected at a later timepoint (d30) to allow detection of tumor growth differences and thus may have affected immune comparisons. However, more T-cells were activated (HLA-DR+) in the TILs of mice that received dual therapy and the TILs secreted more IFNγ relative to T-cells from the spleen (Fig. 6b). Also, the percentage of Tregs significantly decreased in the combo-treated mice (Fig. 6b). As above, there was no change in Tim3 expression. PD-1 expression was downregulated with nivolumab treatment, but upregulated in the TILs of OKI-179 single-drug treatment (Fig. 6c). There were also more myeloid cells, including those with HLA-DR^lo expression, in the tumor relative to the BM and spleen (Fig. 6d and data not shown). Analysis of the tumors cells showed no change in PD-L1 expression among any treatment groups, suggesting HDAC inhibition does not affect the expression of this inhibitory molecule (Additional file 4: Figure S6b).Fig6Fig. 6

Combination therapy of HDAC inhibitor and nivolumab slows growth of TNBC xenograft in hu-CB-BRGS mice. a, Tumor growth curves of MDA-MB-231 TNBC xenografts implanted into hu-CB-BRGS mice that were treated with either control (black), nivolumab (blue), OKI-179 HDACi alone (green), or combination (red). Tumor weights from end of study mice at d30 harvest (control-4 mice/8 tumors; nivolumab-3 mice/6 tumors, OKI-179-1 mouse/2 tumors, combo-2 mice/4 tumors). b, T-cell (hCD45 + CD3+) number (top left), frequency of HLA-DR+ of CD3+ (top right), IFNγ+ of CD8+ (bottom left) and FoxP3+ of CD3+ (bottom right) in TILs or lymph organs of TNBC-implanted humanized mice. c, MFI of Tim3 (left) or PD-1 (right) on T-cells from TILs. d, Frequency of myeloid cells among the hCD45+ population in increased in TILs relative to spleens. Each point represents data from the indicated organ from an individual humanized mouse that was either untreated (−,black), treated with nivolumab (P, blue), OKI-179 (H, green) or both (C, red). P-values: * < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001

After recapitulating anti-tumor activity and characterizing immune responses of a TNBC cell line in the hu-CB-BRGS mice, we next sought to determine whether more heterogenous models such as PDXs could be used to evaluate targeted agents in CRC models. To mimic clinical results, we selected MSI-H and MSS CRC models and hypothesized they would be responsive or resistant, respectively, to anti-PD-1 therapy. As above, hu-CB-BRGS mice were bled and stratified into control or treatment arms according to chimerism prior to tumor implantation (Fig. 7a). We established CRC MSS and MSI-H PDX humanized mice, with tumor takes of 94 and 89% respectively. Tumor-implanted animals were then treated with nivolumab using the protocol developed for the TNBC-implanted mice. Significant tumor growth inhibition was observed in treated mice with MSI-H CRC hu-CB-BRGS compared to the control mice (Fig. 7b, left). Nivolumab treatment of the same xenograft implanted into non-humanized BRGS mice did not affect tumor growth (Fig. 7b), whereas overall, TGI was 80% in hu-CB-BRGS versus 23% in non-humanized mice (Fig. 7c).Fig7Fig. 7

Nivolumab inhibits growth of a MSI-H CRC PDX that correlates with increased human T cells in the tumor compared to the untreated MSI-H CRC PDX or a less-responsive MSS CRC-PDX in hu-CB-BRGS mice. a, Human hematopoietic (hCD45+) chimerism in the blood prior to CRC MSI-H PDX implantation. b, Tumor growth curves of MDA-C099–203 CRC-PDX (MSI-H) implanted into humanized (left) or non-humanized (right) BRGS mice. Hu-CB-BRGS control group: n = 6 mice/10 tumors; nivolumab group: n = 8 mice/15 tumors. BRGS control group: n = 2 mice/3 tumors; nivolumab group: 3 mice/6 tumors. c, Tumor growth inhibition comparison in the hu-CB-BRGS MSI-H, BRGS MSI-H and hu-CB-BRGS MSS mice. d, Human hematopoietic (hCD45+) chimerism in the spleen of CRC-MSI-H bearing hu-CB-BRGS mice. e-i, Immune measurements, as assessed using flow cytometry, of TILs or spleen from CRC MSI-H implanted humanized mice treated (+) or not (−) with nivolumab. e, The number of human (hCD45+), T (hCD3+) and CD8 (CD3 + CD8+) T-cells in the TILs. f, The numbers of IFNγ-producing T cells in the TILs (gate: hCD45 + CD3+). The MFI of inhibitory receptors PD-1 (g) and Tim3 (h) on human T cells (hCD45 + CD3+). Human PBMCs served as control. i, The MFIs of PD-L1 expression on the CRC MSI-H PDX tumors excised from the humanized mice were determined on the hCD45-mCD45-Epcam+HLA-A,B,C+ population in the TILs. j, Tumor growth curves of CRC172 MSS PDX in hu-CB-BRGS mice (n = 4 mice/8 tumors per group). k, The number of hCD45+ cells in the TILs of either the MSS CRC172 PDX or the MSI-H MDA-C099–203 PDX grown in the hu-CB-BRGS. Open circles in MSS nivolumab treated mice represent tumors from the nivolumab-treated hu-CB-BRGS mouse harvested d10 after start of treatment, with the control mice, while closed blue circles are from the 3 treated hu-CB-BRGS MSS mice harvested on d30. All MSI-H bearing hu-CB-BRGS mice were harvested d23–24 after start of nivolumab. Each point represents data from the indicated organ from an individual hu-CB-BRGS mouse that was either untreated (−, black) or treated with nivolumab (+, blue), as indicated. P-values: * < 0.05, ** < 0.01, *** < 0.001, **** < 0.0001