Reconstitution of human immune cells in NIKO mice

Neonate NIKO mice were engrafted with human CD34+ cells via intrahepatic injection. Starting from one-week post-partum, peripheral blood, thymus, lung, liver, spleen and bone marrow were harvested to analyze the reconstitution of human immune cells in these organs. We observed successful human immune cell reconstitution across all organs of the NIKO mouse, especially the vestigial thymus. Most tissues analyzed saw increased relative percentages of human CD45+ versus combined human and mouse CD45+ cells over time, peaking at 16.5 to 28.5 weeks post-partum. The degree of reconstitution varied amongst organs, but at the peak, human CD45+ immune cells comprised greater than 50% of all CD45+ cells in all organs. The thymus was the exception, which displayed near complete reconstitution of human immune cells from 6 weeks post-partum. As the mice aged, the relative percentage of human CD45+ cells declined in all organs from its peak (Fig. 1A, Additional file 1: Fig. S1), possibly due to reduced proliferation of these cells or increased proliferation of mouse immune (myeloid) cells.

Fig. 1

Human immune cell reconstitution of NIKO mice. A Timeline for engrafting human CD34+ cells and harvesting of blood and organs for analysis. Percentage reconstitution of human immune cells in blood, bone marrow, spleen, lung, liver and thymus of NIKO mice over a course of 52 weeks. mean ± SD are indicated; n = 5. B Proportion of various human immune subsets in the blood, bone marrow, spleen, lung and liver over time. For the thymus, immune cells largely comprised of B and T cells, proportion of subsets as shown. Bars show mean ± SD; n = 5

We immunophenotyped the human CD45+ cells to track changes in the human immune cell population over 52 weeks and observed drastic changes in the composition of leukocytes in all organs. One week after human CD34+ cell engraftment, all tissues (excluding the thymus) were dominated by a combination of HLA-DR+ myeloid cells, residual CD34+ progenitor cells, and CD66b+ granulocytes (Fig. 1B, Additional file 1: Fig. S1). B cells were the next primary cell type to appear during week 3 and dominated the hematopoietic compartment up to week 16.5, while the myeloid population correspondingly decreased. This was especially true in the blood, spleen and bone marrow, where B cells comprised the majority of immune cells from week 3 to 16.5. The liver and lung appeared to support the myeloid cell population the longest, comprising 23.73% and 22.59% of human CD45+ cells in the liver (up to week 9) and lung (up to week 16.5), respectively. T cells began to appear in the blood, spleen, lung and liver at week 9, and in the bone marrow at week 16.5. They quickly surpassed all other cell types, making up the majority of the human CD45+ cells by week 28.5 in the spleen (78.87%), bone marrow (89.44%), lung (93.42%), liver (96.67%) and blood (97.06%) (Fig. 1B, Additional file 1: Fig. S1).

The thymus showed a skew towards the adaptive immune cell compartment from week 3, being dominated by B cells (41.02%), single positive CD4 (13.94%) T cells, as well as the expected CD3-positive, CD4/CD8 double positive (DP) (13.97%) and double negative (DN) (18.13%) thymocytes. The frequency of B cells steadily decreased as the mice matured. This was accompanied by an increase in DP T cells, peaking at week 16.5 (74.05%) and decreasing with age, thereafter, denoting appropriate thymopoiesis of human T cells (Fig. 1B, Additional file 1: Fig. S1).

Overall, human CD34+ cells, when transplanted into the liver of neonate NIKO mice, can reconstitute various human immune cell types, the composition of which drastically changes as the mice age. While there was a good mixture of T, B and myeloid cells up to week 16.5, the mice exhibited a T cell-dominant phenotype from week 28.5 onwards. This phenomenon was consistent across all organs except for the thymus, which led us to focus our study on T cells.

Population shifts of T cells in aging humanized mice

One of the effects that aging has on the human T cell population is the decreased proportion of naïve T cells in the blood, with a corresponding increase of memory and senescent T cells [22]. In humanized mice, we found that in the blood, lung and spleen, there was a significant decrease in the proportion of naïve CD4+ and CD8+ T cells starting after weeks 9 and 16.5, respectively, to week 52 (Fig. 2A-C, Additional file 2: Fig. S2A-C). This decrease was accompanied by a concomitant increase in memory subsets (Fig. 2A-C). However, this phenomenon was not apparent in the liver, likely due to the inherently low proportion of naïve cells even in younger mice (Fig. 2D, Additional file 2: Fig. S2D), which could be attributed to the activation of naïve cells through antigen presentation by liver-resident cells [23]. The declining naïve and increasing memory T cell compartments in aging humans were hence mirrored in the aging humanized mice.

Fig. 2

Changes in T cell population as humanized mice aged. A to D Proportion of naïve and memory CD4+ and CD8+ T cell subsets in blood (A), lung (B), spleen (C) and liver (D). Bars show mean ± SD; n = 5. E and F Percentage of CD8+ T cells (E) and CD4.+ T cells (F) that express senescent makers, CD57 and KLRG1, over time. Kruskal–Wallis test was used for significance tests. mean ± SD and significant p values are indicated; n = 5

As cellular senescence is one hallmark of aging, we investigated whether the T cells also displayed increasing surface markers of cellular senescence with age. Defining senescent-like T cells as KLRG1+CD57+ cells [24,25,26], we observed that in blood, lung, spleen and liver, there was a progressive increase in the proportion of senescent-like CD8+ T cells over time from week 9 to 52. At week 52, the proportion of senescent-like cells was significantly higher than week 9 with 12.11%, 17.44%, 22.67% and 26.50% in the spleen, liver, blood, and lung, respectively (Fig. 2E). A similar phenomenon was seen, albeit to a lesser extent, for the CD4+ T cells in the blood, lung and liver, but not the spleen (Fig. 2F). As such, our humanized mouse model recapitulated T cell population changes in aging humans.

Population and functional insights through single cell RNA sequencing

To better understand the population shifts between young and aging immune cells, we performed single cell RNA-sequencing on splenocytes from humanized mice, as the spleen has been assessed to be a good indicator of systemic immune functions [27]. After quality control and filtering, a total of 10,881 (4,088 young and 6,793 aging) immune cells were analyzed. The data was visualized by t-distributed stochastic neighbor embedding (tSNE), and the cells were grouped into nine clusters. Combining information on the top ten features with specific marker genes (CD3, CD4, CD8, CD19 and NCAM1), the clusters potentially represent (to a large extent) CD8+ T cells (Cluster 1 and subset of 2), CD4+ T cells (Clusters 2 and 3), B cells (Clusters 4, 7 and 9), Granulocytes (Cluster 5), Regulatory T (Treg) cells (Cluster 6), and Natural Killer (NK) cells (Cluster 8) (Fig. 3A, Additional file 4: Fig. S4). Consistent with our earlier findings, aging splenocytes largely comprised of T cells, with a much smaller B cell population compared to young splenocytes. A closer look at the tSNE plot revealed that T cells from aging mice within T cell clusters 1, 2 and 3 were situated lower than those from younger mice along the vertical tSNE plane, indicating a distinct shift in CD8+ and CD4+ T cell transcriptome with age (Fig. 3A, Additional file 4: Fig. S4). Unexpectedly, KLRG1 expression in these T cells did not appear to correlate with age (Additional file 4: Fig. S4).

Fig. 3

Single cell RNA sequencing analysis of splenocytes. A tSNE plot showing nine graph-based clusters that potentially represent different human immune subsets found in splenocytes harvested from aging (55-week-old) and young (14-week-old) humanized mice. Table shows top ten expressed genes for each cluster with likely cell type in parenthesis. B Gene set enrichment analysis of genes that were up- or down-regulated by at least twofold. Top ten enriched GOs are shown for CD8+ cells in Cluster 1, CD4+ cells in Cluster 2, CD4+ cells in Cluster 3 and CD4+ cells in Cluster 6. p values are indicated

Since this study focuses on T cells (CD3+), we compared the differentially expressed genes between aging and young CD3+ cells within CD8+ Cluster 1, CD4+ Cluster 2, CD4+ Cluster 3, and CD4+ Cluster 6 (Tregs). Genes that were significantly up- and down-regulated by more than two-fold were subjected to Gene Set Enrichment Analysis (GSEA) to yield Gene Ontologies (GO) that provide insights into affected molecular functions and biological processes. Looking at CD8+ T cells of Cluster 1, genes that were significantly downregulated in aging cells are associated with immune responses (adaptive and innate), response to stimulus, immune system process and defense response, while those that were significantly upregulated are related to developmental processes such as morphogenesis (Fig. 3B, far left). Aging CD4+ T cells in Cluster 2 also expressed lower levels of genes that correspond to immune processes, as well as mitosis and cell proliferation. These cells instead showed increased expression of genes involved in antigen presentation and T cell stimulation (Fig. 3B, left). For their counterparts in Cluster 3, downregulated genes corresponded to oxidoreductase activity, regulation of B cell activation, and IgM immunoglobulin complex, while upregulated genes were enriched in amino acid metabolism and nucleoside phosphorylase activity (Fig. 3B, right). Finally, for Tregs in Cluster 6, the downregulated genes were associated with activation of immune cells and regulation of immune system, while the upregulated genes were associated with MHC Class II processes and lymphocyte co-stimulation (Fig. 3B, far right). Taken together, single cell RNA-sequencing of aging and young splenocytes revealed immune population shifts as well as changes in molecular functions and biological processes that could explain age-related morbidity, such as increased tumor incidence and progression.

T cell dysfunctions in aging humanized mice

Given the reduced proportion of naïve and increased proportion of senescent-like T cells in aging humanized mice, as well as the GOs affected by transcriptomic changes, we next investigated the functional aspects of the aging T cells. We performed in vitro stimulation, using PMA and ionomycin, and intracellular staining (ICS) of splenocytes from week 18 and week 60 humanized mice to compare the intrinsic capability of T cells to produce cytokines. The proportion of IFNγ-producing CD4+ and CD8+ T cells at week 60 was lower compared to week 18, even though it was not significant. Strikingly, there was a significant reduction in the proportion of Granzyme B positive CD8+ T cells at week 60 compared to week 18, regardless of in vitro stimulation. Similarly, the proportion of Granzyme B positive week 60 CD4+ T cells was about half that of their week 18 counterparts, although the difference was not significant (Fig. 4A).

Fig. 4

In vitro analysis of T cell function (A) Intracellular staining of splenocytes for interferon-gamma and granzyme B, 4 h after culture with or without PMA/ionomycin stimulation. CD4 and CD8 cells were distinguished by FACS. Bars show mean ± SD and p values are indicated; n = 5. B and C Analysis of cytokines in the supernatant 72 h after culturing splenocytes in the absence (B) or presence (C) of anti-CD3/28 stimulation, using LEGENDplex™ T helper cytokine panel (left) or LEGENDplex™ CD8/NK panel (right) to probe for T helper or CD8/NK related cytokines, respectively. Bars show mean ± SD and p values are indicated; n = 5. Mann–Whitney test was used for all significance tests

In addition to looking at intrinsic cytokine production capability of the cells, we simulated T cell receptor-dependent activation using anti-CD3 and anti-CD28 antibodies in vitro for 72 h. The cytokine profile of the culture supernatant was measured using BioLegend’s LEGENDplex™, and the results supported the ICS data, where week 60 splenocytes produced significantly less IFNγ and Granzyme B compared to week 18 splenocytes. The bead-based immunoassay also showed that while week 18 T cells were functionally capable of producing a wide range of pro-inflammatory and cytotoxic cytokines such as IL-5, IL-13, IL-6, IFNγ, TNF-α, IL-17A, IL-22, GranzymeA, GranzymeB, Perforin, and Granulysin in response to physiological activation, week 60 splenocytes were largely deficient in this regard (Fig. 4B, C). Since our results demonstrated that aging humanized mice were dominated by T cells throughout multiple organs, and that these T cells were functionally compromised compared to their younger counterparts, we theorized that aging mice would have impaired tumor-killing capacity.

Disparity in tumor dynamics between aging and young humanized mice

We examined the effects of immune aging on tumor dynamics in our humanized mouse model. Human hepatocellular carcinoma cells (PLC/PRF/5) were subcutaneously engrafted into the flanks of aging (week 60) and young (week 18) humanized mice, which were then monitored for tumor growth over 4 weeks (Fig. 5A). Tumors harvested from aging mice were comparatively bigger than those from young mice (288.7mm3 vs 129.1mm3) at 4 weeks (Fig. 5B), an observation that aligns with the higher incidence of cancer in the aging human population. Looking at tumor volume over time, aging mice had consistently larger tumors that were approximately up to three times the size of the tumors from their younger counterparts (Week 1, 27.2 mm3 vs 16.8 mm3; Week 2, 72.2 mm3 vs 34.4 mm3; Week 3, 185.9 mm3 vs 61.5 mm3) (Fig. 5B).

Fig. 5

Tumor study in aging and young humanized mice. A Schematic diagram showing reconstitution of NIKO with human immune cells and subsequent engraftment of liver cancer cell line PLC/PRF/5 in young (Week 18) and aging (Week 60) humanized mice. B Pictures of tumors harvested 4 weeks post engraftment. Scale bar, 1 cm. Change in tumor volume over time. Bars show mean ± SD and p values are indicated; n = 5. C Analysis of TILs for percentage of hCD45 and CD4/CD8 ratio. Bars show mean ± SD and p values are indicated; n = 5. D Analysis of TILs for percentage of cells expressing exhaustion markers. Bars show mean ± SD and p values are indicated; n = 5. E KLRG1 expression on CD3+ T cells in blood and spleen of non-tumor engrafted humanized mice. Bars show mean ± SD and p values are indicated; n = 5. F LEGENDplex™ analysis of serum from tumor engrafted mice using Inflammation panel 1, 4 weeks post engraftment. Bars show mean ± SD and p values are indicated; n = 5. Mann–Whitney test was used for all significance tests

To understand the interplay between T cells and tumor development, we isolated the TILs, which largely comprised of CD3+ T cells, for flow cytometric analysis. Strikingly, the percentage of hCD45 in young mice (69.7%) was significantly higher than aging mice (30.4%) (Fig. 5C, left). Aging mice also presented with a lower proportion of CD4+ and a higher proportion of CD8+ T cells compared to young mice, and thus a lower CD4/CD8 ratio, although this was not significant (Fig. 5C, right). As T cell exhaustion is a common phenomenon related to reduced anti-tumor response [28], we looked at exhaustion markers. Inhibitory receptors programmed cell death protein 1 (PD-1) and T-cell immunoglobin domain and mucin domain protein 3 (TIM3) were highly expressed in similar proportions of aging and young effector memory T cells, reflecting a typical exhausted phenotype of TILs, and suggesting that these pathways are not responsible for the difference in tumor development. The exhaustion markers cytotoxic T-lymphocyte-associated protein 4 (CTLA-4, 3.05% vs 1.47%) and T cell immunoreceptor with Ig and ITIM domains (TIGIT, 2.96% vs 0.54%) were expressed in a significantly higher proportion of aging CD4+ effector memory T cells (Fig. 5D, left), while KLRG1 (11.88% vs 3.95%) was expressed in a significantly higher proportion of aging CD8+ effector memory T cells (Fig. 5D, right). CTLA-4, TIGIT and KLRG1 may thus contribute to the reduced anti-tumor activity in aging mice. In fact, KLRG1 expression on CD3+ T cells was already elevated in the blood and spleen of age-matched aging humanized mice in the absence of tumor (Fig. 5E). This T cell exhaustion phenotype was corroborated by the significantly lower concentrations of various T-cell associated inflammatory cytokines, such as IFN-γ, TNF-α, IL-6 and IL-17A, in the plasma of tumor-bearing aging mice (Fig. 5F). Taken together, these data suggest that the exhaustion of aging T cells could account for the more aggressive tumor dynamics in aging mice.