Introduction

Medulloblastoma (MB) is a brain tumor seen mainly in pediatric patients and which has four molecular subgroups, WNT, SHH, Group 3 and Group 4¹. The Sonic hedgehog subgroup of MB (SHH-MB) constitutes approximately 30% of MB and is characterized by mutations that constitutively activate SHH-signaling. Most SHH-MBs have one of four types of mutations that activate SHH-signaling: activating mutations in SMO, loss of function mutations in the pathway inhibitors PTCH1 and SUFU or loss of TP53 in association with amplification of the pathway effectors GLI2 or MYCN. SHH-MB is further divided into four subtypes based on molecular, biological and clinical parameters^2–6. The work on SHH-MB subtypes has highlighted that particular secondary mutations likely contribution to tumor severity. For example, heterozygous PTEN and KMT2D loss-of-function mutations are enriched in the SHH-1 or beta subtype which has the lowest overall survival rate and highest frequency of metastasis and occurs primarily in infants (<4 years). Our recent work on the chromatin modifier KMT2D using two SHH-MB mouse models showed that when one or two copies of Kmt2d are deleted, not only is there a nearly two-fold increase in tumor penetrance with a Classic tumor cytoarchitecture (highly proliferative), but strikingly a near 100% penetrance of spinal cord metastasis and hindbrain invasion⁷. The transcriptional/chromatin landscape of mouse Kmt2d mutant SHH-MB primary and metastatic tumors was found to be shifted towards decreased expression of neural differentiation genes and an increase in oncogenes seen in more aggressive tumors compared to tumors with intact Kmt2d⁷. Thus, mouse models support the clinical data that loss of KMT2D contributes to more aggressive disease.

The gene encoding the dual specificity phosphatase PTEN (Phosphatase and TENsin homolog deleted on chromosome 10), is one of the most frequently mutated tumor suppressor genes in human cancers⁸, and is mutated in ∼5% of SHH-MB³. PTEN regulates an array of cellular functions, including cell survival, proliferation and metabolism through attenuating phosphoinositide-3 kinase (PI3K-AKT) signaling and genome stability through nuclear functions. Recent work showed that PTEN also can be secreted and act in a cell nonautonomous manner^9–11. PTEN is secreted by tumor cells via binding to TMED10 and then binds to PLXDC2 on tumor associated macrophages (TAMs) and induces inflammatory status through triggering JAK2-STAT signaling, thus contributing to the tumor suppressor function of PTEN. Mouse models of SHH-MB indicate loss of Pten increases tumor incidence, but the underlying cell autonomous and nonautonomous reasons and impact on metastasis are not as clear^{12, 13}.

SHH-medulloblastoma arises in Atoh1-expressing granule cell precursors (GCPs) on the surface of the developing cerebellum. These precursor cells are dependent on SHH-signaling for their sustained proliferation and major expansion during the first two weeks after birth in mice^14–16. In a mouse model of SHH-MB in which a NeuroD2 promoter drives expression of an oncogenic form of SMO (SmoA1) in the GCP-lineage, heterozygous loss of Pten in all cells (null mutation) or conditional knock-out in Nestin-expressing cells, which includes all cerebellar cells¹⁷, leads to increased penetrance and earlier onset of tumor formation with no increase in metastasis¹². Whereas most mouse models of SHH-MB have a classic (highly proliferative) histology, the Pten heterozygous mutant tumors have a nodular character with extensive regions of neural differentiation (NeuN+ cells) and an apparent decrease in cell death and increase in angiogenesis. In a mouse SHH-MB conditional model in which Ptch1 is homozygously deleted at embryonic day 14 (E14) in all GCPs (Atoh1-Cre), additional homozygous loss of Pten leads to rapid tumor formation, and heterozygous loss leads to a slightly more aggressive phenotype than Ptch1 loss alone¹³. Similar to NeuroD2-SmoA1 tumors, in Ptch1 mutant tumors Pten loss causes highly differentiated tumors with decreased cell death. In a mutagenic screen for increased SHH-MB penetrance and metastasis using a Sleeping beauty transposon approach in mice with heterozygous mutations in Ptch1 or Trp53, insertions found in tumors with spinal cord metastasis were associated with mutations in Pten, Akt2, Pik3R and Igf2, implicating increased PI3K signaling in metastasis¹⁸. Consist with this, in a mouse model where Shh is expressed in GCPs, additional expression of an activated form of AKT increases tumor incidence and spinal cord metastasis^{18, 19}. Data from human tumors support that a reduction in PTEN protein enhances tumor progression, since SHH-MB human patients with reduced or no PTEN protein have a worse overall survival¹². However, no correlation was seen between PTEN protein level and histology or metastasis. Thus, several questions remain as to whether Pten loss increases metastasis, what accounts for the faster growth and a more differentiated cellular phenotype in mouse Pten mutant SHH-MBs and why only in some models Pten heterozygous loss is sufficient to greatly enhance tumor progression.

We addressed these questions by comparing tumor progression and spinal cord metastasis in a sporadic mouse model of SHH-MB in which rare GCPs are induced to express an oncogenic form of SMO, SmoM2, at postnatal day 0 (P0) and have deletion of one or both copies of Pten. We found that only homozygous loss of Pten combined with SmoM2 expression enhances tumor progression in this sporadic model, and metastasis is not increased. Furthermore, the differentiated phenotype of the tumors is primarily due to a specific decrease in cell death of non-proliferative cells in the tumors, whereas cell death is as high in regions of high proliferation as in SmoM2 tumors with normal Pten or heterozygous loss of Pten. Loss of Pten in normal GCPs, or those expressing SmoM2, leads to a mild initial increase in the proliferation index, indicating a cell autonomous function of PTEN in reducing proliferation. Single cell RNA-sequencing (scRNA-seq) analysis revealed that proliferating tumor cells that lack Pten not only have the expected increase in mTOR signaling, but also have altered expression of genes involved in neural differentiation and down regulation of inflammatory and interferon alpha response genes. Homozygous loss of Pten in GCPs also leads to a major decrease in infiltration of macrophages into tumors and an altered transcriptome indicating reduced cytotoxicity, revealing a possible cell nonautonomous function of PTEN that normally reduces tumor progression.

Results

PTEN mutations co-occur with mutations that activate SHH-signaling in human SHH-MB

Using the cBioPortal^{20, 21} and a dataset of 127 patient samples considered to be SHH-MB³, we determined whether patients with mutations that activate SHH-signaling have PTEN mutations. 4.3% of the 127 samples had a mutation in PTEN and furthermore 4/5 of the mutations co-occurred with mutations that activate SHH-signaling (Ptch1, SUFU or SMO; Fig. 1A). Thus, although the PTEN mutations were detected in a low percentage of patient tumors, PTEN appears to be predominantly mutated in SHH-MB along with mutations that activate SHH-signaling.

Figure 1.

Homozygous loss of Pten in sporadic SHH-MB accelerates growth and increases penetrance to 100%

We next used our mosaic mutant technique (MASTR)²² to test the impact of lowering Pten in a sporadic mouse model of SHH-MB. The approach involves inducing expression of an oncogenic form of the SHH receptor SMO (SmoM2) and eGFP in ∼20,000 GCPs (∼0.1%) at P0 (Fig. 1B)^22–24. The model allows penetrance and survival time to be determined since less than 75% of mutant mice die by 100 days. Using the model, we asked whether loss of one or both copies of Pten²⁵ in rare GCPs-expressing SmoM2 changes the survival curves of mice by comparing Atoh1-FlpoER/+; R26^{MASTR/LSL-SmoM2} mice (referred to as SmoM2) to Atoh1-FlpoER/+; R26^{MASTR/LSL-SmoM2}; Pten^fl/fl (referred to as SmoM2-Pten^fl/fl) and Atoh1-FlpoER/+; R26^{MASTR/LSL-SmoM2}; Pten^fl/+ (referred to as SmoM2-Pten^fl/+) mice administered tamoxifen at P0 (Fig. 1B). Strikingly, all SmoM2-Pten^fl/flmice succumbed to disease between P20 and P40, whereas SmoM2 and SmoM2-Pten^fl/+mice began dying at ∼P40 and by P100 ∼25% of mice of both genotypes survived (Fig. 1C). Thus, loss of both Pten alleles is required to accelerate tumor progression in a mosaic SmoM2 model of SHH-MB, unlike a SmoA2 model with germline or broad deletion of Pten in the cerebellum where heterozygous deletion is sufficient¹².

Homozygous and heterozygous loss of Pten does not increase metastasis to the spinal cord in SmoM2 SHH-MB

Given the different reported findings as to whether Pten loss increases metastasis in mouse models of SHH-MB, we asked whether in our sporadic SmoM2 mice loss of Pten alters the frequency of spinal cord metastasis when mice show signs of brain tumor disease. Coronal spinal cord sections of each mouse (n=55-138 sections per mouse) were analyzed by cytology for the presence of spinal cord metastasis and tumors were confirmed with immunohistochemical (IHC) staining for GFP, Ki67 and PTEN (Figure 1–figure supplement 1A-C). Quantification of the percentage of mice that had at least 2 sections with more than a few tumor cells showed a similar tumor incidence between the three genotypes (SmoM2: 64%, n=11; SmoM2-Pten^fl/+: 69%, n=13; SmoM2-Pten^fl/fl: 58%, n=12). Furthermore, quantification of the percentage of spinal cord sections per mouse with tumor cells showed no difference between the Pten mutant tumors and SmoM2 (Figure 1–figure supplement 1D). Likewise, measuring the area of the tumors did not detect a difference in metastatic disease when Pten was lowered or deleted (Figure 1–figure supplement 1E, F). Thus, in mouse sporadic SmoM2 SHH-MB, heterozygous or homozygous loss of Pten does not appear to increase spinal cord metastasis.

Pten loss in a SmoM2 sporadic mouse model of SHH-MB induces highly differentiated tumors

Hematoxylin and eosin (H&E) staining of primary tumor sections indicated that homozygous loss of Pten results in tumors with decreased cellularity (Fig1 D-F). IHC staining of tumor sections at end stage confirmed that most tumor cells expressed GFP (Fig 1. G-I) and only SmoM2-Pten^fl/fl tumors lacked PTEN (Fig 1. J-L) and had strong Phospho-AKT (pAKT) (Fig 1. M-O) and pS6 staining (Figure 2–figure supplement 1A-D), thus confirming absence of PTEN and activation of mTOR (mammalian target of rapamycin) signaling only in SmoM2-Pten^fl/fl tumors. Staining with markers of cell proliferation (Ki67) and differentiation revealed that whereas SmoM2 tumors consisted of mostly Ki67+ cells, SmoM2-Pten^fl/fl tumors were highly differentiated expressing the post mitotic marker P27 and neural differentiation marker NeuN with only small areas of proliferative cells (Fig. 2A-H). The SmoM2 tumors had broad expression of the SHH target gene Gli1, whereas the SmoM2-Pten^fl/fltumors expressed Gli1 mRNA only in Ki67+ regions of the tumors (Fig. 2I-L), which was complementary to the neural differentiation marker Synaptophysin (SYP/Syp) that was absent from SmoM2 tumors (Fig. 2K-P). Thus, despite SmoM2 being expressed in the differentiated cells of the SmoM2-Pten^fl/fltumors, SHH-signaling is down regulated specifically in differentiated cells. Interestingly, in 3/9 tumors from SmoM2-Pten^fl/+ mice some sections had patches of cells devoid of PTEN and positive for pAKT, indicating spontaneous loss of the second allele of Pten in rare cells that expand as a coherent group (Figure 2–figure supplement 1E-H). The regions without PTEN in SmoM2-Pten^fl/+ mice had a similar differentiated phenotype (SYP+ and NeuN+) to SmoM2-Pten^fl/fl tumors (Figure 2–figure supplement 1I-K).

Figure 2.

Pten loss protects differentiated tumor cells from cell death

Mouse SHH-MB models that have a highly differentiated cytology are rare and have primarily been reported in two Pten models^{12, 13}. In a Ptch1 mutant model of SHH-MB, it was shown that proliferating tumor cells give rise to NeuN+ postmitotic tumor cells that then die at a high rate²⁶. We therefore asked whether Pten loss (increased mTOR signaling) rescues the post-mitotic tumor cells from death by staining for SYP or NeuN and TUNEL (Fig. 3A-H). SmoM2-Pten^fl/fl tumors had significantly less TUNEL particles in regions of high SYP labeling (SYP Hi) compared to SYP low regions (Fig. 3C-I, Figure 3–figure supplement 1). Moreover, in the regions of low SYP, the density of TUNEL staining was similar to in SmoM2 tumors where SYP is low throughout the tumor. (Fig. 3J). As expected, in PTEN negative patches of SmoM2-Pten^fl/+tumors, TUNEL staining was lower in these SYP Hi regions of differentiated cells (NeuN+) than in SYP low proliferative regions (Ki67+) (Fig. 3J,K and Figure 2–figure supplement 1I-L).

Figure 3.

Loss of Pten leads to rapid cell autonomous changes in SmoM2 GCP behaviors

Given how rapidly the tumors grow in the SmoM2-Pten^fl/flmice and that lethal disease occurs between P20-40, we asked how early a differential growth phenotype can be seen in the three genotypes. In our previous study of the effect of loss of Kmt2d in SHH-MB models, no difference in tumor size was detected at P12 but was by P21⁷. We therefore asked whether SmoM2-Pten^fl/fltumors have a growth advantage over SmoM2 tumors at P12. Staining of sagittal sections of SmoM2 tumors of the three Pten genotypes with EdU (1 hour pulse) and GFP to mark the proliferating GCPs in the external granule cell layer (EGL) showed a massive expansion of the EGL in SmoM2-Pten^fl/flmice compared to SmoM2 mice (Fig. 4 A-C). Quantification of cerebellar area in midline sections revealed a significant increase in SmoM2-Pten^fl/flmice compared to SmoM2 mice and SmoM2-Pten^fl/+ mice (Fig. 4D). The EGL area alone showed a 4-fold average area increase in SmoM2-Pten^fl/fl mice compared to SmoM2 mice and SmoM2-Pten^fl/f+ mice had a slight increase (<2-fold) compared to SmoM2 mice (Fig. 4E). As in end stage tumors, the tumors in SmoM2-Pten^fl/fl mice had a significant decrease in dying cells compared to the other two genotypes at P12 (Fig. 4F-L, Fig. 4–figure supplement 1). In large lesions with extensive differentiation (NeuN+ Ki67-cells) the density of TUNEL particles was decrease in NeuN+ regions compared to proliferative regions. As expected, pAKT staining was upregulated in the EGL of SmoM2-Pten^fl/flmice with no detectable staining in the other two genotypes (Fig. 4M-O).

Figure 4.

One prediction of our findings is that once SmoM2 GCPs lacking Pten differentiate their survival should be enhanced. Therefore, if proliferating GCPs are labeled and then fate mapped for two days, a higher percentage of SmoM2-Pten^fl/fl labeled cells should be differentiated (post-mitotic) compared to in SmoM2 mice. To test this idea, we administered one injection of BrdU into P10 mice and then sacrificed them at P12 and quantified the percentage of BrdU+ cells in the EGL that were Ki67-(Fig. 4P). As predicted, a higher percentage of BrdU+ cells in SmoM2-Pten^fl/fl mice were Ki67-at P12 compared to SmoM2 (Fig. 4Q). This result indicates that by P10 after loss of Pten at P0, either differentiation is promoted in SmoM2 GCPs and/or that the differentiated progeny of GCPs lacking PTEN are spared from cell death. The proliferation rate of GCPs in SmoM2 and SmoM2-Pten^fl/fl mice were similar at P12 (Fig. 4R), as determined by injecting EdU 1 hour before sacrificing the mice (%EdU+ Ki67+/Ki67+ cells), indicating the latter possibility.

Given that at P12 the cellular phenotype of GCPs in SmoM2-Pten^fl/fl mice was similar to in mice at end stage, we analyzed GCPs at P8, eight days after the mice were treated with tamoxifen. Similar to at P12, the area of the proliferating EGL of the cerebellum was significantly greater in both the SmoM2-Pten^fl/fl and the SmoM2-Pten^fl/+mice compared to SmoM2, although the area of the cerebellum was similar across genotypes likely due to the relatively small increase in EGL area (∼2-fold)(Fig. 5A-E). Quantification of the proliferation rate (%GFP+ EdU+/GFP+ cells in the outer EGL (oEGL) of lobule 4/5 (L5) containing proliferating cells was slightly but significantly higher in the SmoM2-Pten^fl/fl and the SmoM2-Pten^fl/+ mice compared to SmoM2 (Fig. 5F). Moreover, the percentage of mutant (GFP+) cells that remained in the outer EGL of all the GFP+ cells in L5 (including those that differentiated) was significantly higher in the two Pten mutant genotypes (Fig. 5G). This result indicates that loss or reduction of Pten initially induces a pro-proliferative or progenitor state in SmoM2-expressing GCPs. By P12 however, SmoM2-Pten^fl/fl GCP-lineage cells accumulate more in the EGL than do SmoM2-Pten^fl/+ or SmoM2 GCPs, in part because there is an accumulation of NeuN+/SYP+ granule cell (GC)-like cells, likely due to a survival advantage of these cells. This survival advantage suggests that the rapid growth of tumors in SmoM2-Pten^fl/fl mice is due to accumulation of GC-like cells and would account for the highly differentiated cytology of end stage tumors.

Figure 5.

Loss of Pten in normal GCPs causes increased proliferation and decreased differentiation

Given that Pten loss initially increases proliferation and a progenitor state of SmoM2 (SHH activated) GCPs, we asked whether Pten loss in normal GCPs has the same effect. Using our mosaic MASTR approach, we used Atoh1-FlpoER to induce GFPcre in scattered GCPs at P0, with or without concomitant deletion of both floxed copies of Pten (Fig. 6A). As in SmoM2 GCPs, we found that at P8 a small but significant increase in the proliferation rate and in the proportion of GFP+ cells that remained in the proliferative outer EGL of Pten mutant GCPs (Fig. 6 B-E). Using Atoh1-Cre to delete Pten in all GCPs in the embryo (E14 onwards), we found that the cerebellar area of Atoh1-Pten conditional mutants (Atoh1/+; Pten^fl/fl) was significantly increased at P5 and later, but not at P1 compared to Atoh1/+; Pten^fl/+ control mice (Fig. 6F-N). Thus, PTEN normally constrains GCP proliferation and possibly differentiation.

Figure 6.

Loss of Pten in SmoM2 SHH-MB tumor cells results in a cell nonautonomous decrease in macrophage infiltration

Given the low density of dying cells in SmoM2-Pten^fl/flSHH-MB, we asked whether the tumors have a change in the density of infiltrating macrophages. We previously showed that tumors in SmoM2 mice have a high infiltration of macrophages ant that they are pro-tumorigenic in a majority of tumors, because reduction of macrophages using PLX5622 treatment leads to a decrease in tumor incidence²⁷. Staining of cerebellar tumor sections from SmoM2-Pten^fl/fl and SmoM2 mice at end stage revealed the expected high infiltration of IBA1+ macrophages in SmoM2 tumors, but also a striking decrease in macrophage density in tumors lacking Pten (Fig. 7A-D). Moreover, co-staining with Ki67 or NeuN showed that in SmoM2-Pten^fl/fl mice the macrophages were preferentially located in the areas with proliferating cells and thus increased cell death (Fig. 7C-F). The decrease in macrophages in the differentiated regions of SmoM2-Pten^fl/fl mice is not due to a lack of blood vessels, based on CD31 staining of tumors (Fig. 7G, H). At P12, in small lesions the macrophages mainly accumulated in the meninges between the lobules but in large lesions they were more concentrated in regions of proliferation in both genotypes and as in end stage were decreased in NeuN+ regions in SmoM2-Pten^fl/fl mice (Fig. 7I-P). Thus, loss of Pten specifically in tumor cells leads to a cell nonautonomous decrease in macrophages in the tumor microenvironment from an early satge.

Figure 7.

scRNA-seq analysis reveals altered neural differentiation and suppression of immune signaling in tumor cells lacking Pten

To gain insight into transcriptional changes that occur in SmoM2 SHH-MB tumor cells that lack Pten and that could account for the cell autonomous changes seen in the tumors, we performed single cell scRNA-seq of whole tumors dissected from SmoM2 and SmoM2-Pten^fl/fl mice (n = 2 per genotype). We chose to analyze SHH-MBs from P16 SmoM2-Pten^fl/fl mice, since this is a few days before mice show signs of disease, and waited until P22 for SmoM2 mice when the tumors are more visible. Marker analysis of sections of the tumors at the two ages confirmed that the cellular phenotypes of large tumors in SmoM2-Pten^fl/fl and SmoM2 mice were similar to tumors at end stage (Fig. 8–figure supplement 1). The SmoM2-Pten^fl/fl tumors had extensive regions of neural differentiation that had reduced cell death (TUNEL particle density) and macrophage density (IBA1+ cells) compared to in proliferative regions (Ki67+) and in SmoM2 tumors at P22 or P16.

Tumor cells from the two genotypes separated well in a PCA plot after filtering out poor-quality cells and integrating the replicates (Fig. 8–figure supplement 2A-C). Clustering of all cells (17,998 SmoM2 and 22,864 SmoM2-Pten^fl/fl) was performed²⁸ to produce twenty clusters with most of the cells, as expected, expressing the GCP lineage marker Pax6 (clusters 0-9 and 13), with nine additional small clusters that included astrocytes, oligodendrocytes, unipolar brush cells, macrophages, endothelial cells and mesenchymal cells (pericytes and fibroblast-like cells) based on marker gene expression (Fig. 8A, B, Fig. 8–figure supplement 2D, E, Table S1). The cells from each sample were distributed in all clusters, although the proportions of cells from each genotype varied between clusters (Fig. 8C, Fig. 8–figure supplement 2F, G, Table S2). Within the Pax6+ clusters 0-9, based on marker gene expression and cell cycle phase (Fig. 8D, E), clusters 3, 4, 8 were proliferative (Ki67+) and enriched for the GCP markers Gli1, Atoh1 and Barhl1, indicating highly proliferative GCP-like tumor cells. Cluster 6 had a mixture of cells in G2/M and G1, indicating cells transioning from a proliferative to differentiated state. Clusters 1, 5, 7 had post mitotic (G1) cells that expressed early GC differentiation markers including Barhl1 and Slc17a6, indicating early GC-like cells. Clusters 0, 2, 9 were also in G1 but expressed the mature GC markers Slc17a7 and Syp (Fig. 8E) and had a low proportion of cells expressing Gfp (Fig. 8E) and a much higher proportion of cells from SmoM2 mice than from SmoM2-Pten^fl/fl (Fig. 8C, F, Fig. 8-figure supplement 2G) indicating mature GC-like cells. Many of the GC-like cells (clusters 0, 2, 9) were likely wild type GCs from the IGL that would disproportionately contaminate the samples from SmoM2 mice as the tumors were smaller than in SmoM2-Pten^fl/fl mice. When only the GC-lineage clusters with extensive Gfp expression were considered (1, 3, 4, 5, 6, 7, 8), the proportions of cells from the two genotypes were fairly evenly distributed in all clusters (Fig. 8G)

Figure 8.

Pseudotime analysis^{29, 30} reflecting the differentiation pathway of cells within clusters 0-9 with an origin in the GCP-like clusters was consistent with our cell type predictions, showing clusters 0, 2, 9 are most mature and that the trajectory proceeds in order through the transitioning cluster 6, then through the early GC-like clusters 1 and 5 to the mature GC cluster 0 and on to cluster 2 or 9 (Fig. 8H). The analysis further indicated that early GC-like cluster 7 has a separate differentiation pathway, possibly indicating that loss of PTEN suppresses full differentiation of the GC-like tumor cells. These results are also consistent with our IHC marker analysis of tumor sections that showed that SmoM2-Pten^fl/fl tumors have an abundance of GC-like cells expressing NeuN and SYP, and further indicated such cels have in an early GC-like state (clusters 1, 5, 7).

We next performed differential expression analyses between SmoM2 and SmoM2-Pten^fl/fl cells in the GCP-like tumor cell clusters (3, 4, 8) and found 1,011 genes were significantly upregulated and 3,710 were downregulated in the cells lacking Pten compared to SmoM2 cells (adjusted p-value≤0.05, Fig. 9A, Table S3). Gene set enrichment analysis (GSEA) of Hallmark terms confirmed mTORC1 signaling is significantly unregulated in SmoM2-Pten^fl/fl tumors compared to SmoM2, as well as terms associated with MYC and E2F targets and unfolded protein response (Fig. 9B, Fig. 9-figure supplement 1A, Table S4). Violin plots of expression of mTORC1 genes supported upregulation of MTOR signaling (Fig. 9C). As for pathways downregulated in SmoM2-Pten^fl/flcells compared to SmoM2, ‘Interferon alpha response’ was greatest, and the list included ‘complement’ and ‘inflammatory response’ (Fig. 9D), possibly contributing to the observed decrease in infiltrating macrophages in such tumors compared to SmoM2. Violin plots of genes involved in Interferon alpha signaling supported their down rgulation in tumor cells lacking Pten (Fig. 9E). GSEA of Gene ontology (GO) terms (biological processes) showed upregulation of terms associated with neural development in SmoM2-Pten^fl/fl tumors compared to SmoM2 (Fig. 9F, Fig. 9-figure supplement 1B, Table S5), indicating loss of PTEN pushes GCP-like cells toward neural differentiation. Violin plots of expression of genes involved in neural regulation supported uprgulation of neurogenesis (Fig. 9G).

Figure 9.

scRNA-seq analysis confirms reduced cell death in differentiated cells of SmoM2-Pten^fl/fl tumors

Since our immunohistochemical analysis of SmoM2-Pten^fl/fltumors showed that cell death was reduced in the NeuN+/SYP+ regions of tumors, we next performed differential expression analyses of only the Pten mutant cells in the early GC-like cluster 1 to the GCP-like clusters 3, 4 and 8 and found 2,738 genes were significantly upregulated and 4,185 were downregulated in cluster 1 cells compared to GCP-like cells (adjusted p-value≤0.05, Fig. 9H, Table S6). GSEA of GO terms (biological processes) confirmed upregulation of terms associated with neural differentiation and also highlighted down regulation of ‘positive regulation of programmed cell death’ in Pten mutant cells in cluster 1 compared to the GCP-like clusters (Fig. 9I, K, Fig. 9-figure supplement 1C, Table S7). Violin plots of genes underlying the terms supported their altered expression in SmoM2-Pten^fl/fl tumors compared to SmoM2 (Fig. 9J, L).

RNA in situ analysis of sections of P22 SmoM2 and P16 SmoM2-Pten^fl/fl tumors was used to confirm some of the genes shown to be differentially expressed between the GCP-like cells in the two genotypes (Fig. 9M-R). Asns (Unfolded protein gene), Vldlr and Cdkn1a (mTORC1 pathway) showed upregulation in GCP-like tumor cells in P16 SmoM2-Pten^fl/fl compared to P22 SmoM2 tumors.

scRNA-seq analysis indicates loss of Pten in SmoM2 SHH-MB tumor cells results in decreased expression of genes associated with cytotoxicity in macrophages

Comparison of the transcriptomes of macrophages in SmoM2-Pten^fl/fland SmoM2 tumors indicated an anti-inflammatory phenotype in tumors lacking Pten because genes related to synaptic signaling and Irf2 were decreased and Cd36, a marker of protumoral M2-type macrophages, was upregulated in SmoM2-Pten^fl/fl tumors (Fig. 9-figure supplement 1D, E, Table S8, S9)^{31, 32}. The observed reduction in macrophage infiltration into SmoM2-Pten^fl/fl tumors compared to SmoM2 tumors and the apparent reduction in cytotoxicity might reflect an immunosuppressive tumor microenviroment that could contribute to the rapid growth of tumors lacking PTEN.

Discussion

Our work delineates cellular and transcriptional changes that lead to greatly accellerated tumor growth when Pten is mosaicly removed in GCPs that have activated SHH-signaling due to expression of SmoM2. We uncovered that in both wild type and SmoM2 GCPs loss of Pten results in an initial increase in mitotic rate and likelihood to remain in the progenitor state. As SmoM2 tumor formation proceeds, tumors lacking Pten accumulate differentiated early GC-like cells that are stalled in their differentiation and are protected from cell death. Thus, the bulk of the Pten mutant tumors are post mitotic cells. Macrophages seem to avoid or do not readily penetrate the differentiated regions of the tumors, perhaps because there are less dying cells. With possible relevance to human tumors, unlike previous models of Pten heterozygous mutant mice that model Cowden Syndrome or hamartoma syndromes³³ or ones that involve embryonic conditional loss of Pten in all GCPs^{12, 13}, we found that heterozygous loss of Pten in sporadic GCPs does not result in more aggressive tumors than activating SHH-siganling alone in rare GCPs. Since our model more closely reflects the sporadic nature of human tumor mutations, our results predict that heterozygous loss of PTEN will primarily have a negative impact on disease outcome when the mutations are germline (PTEN^+/-). If human PTEN mutant SHH-MB have a similar differentiated cellular phenotype to mouse models, then targeting such tumors with drugs to kill proliferating cells will not remove the bulk of the tumor and will require the drugs having access to the rare proliferating GCP-like tumor stem-like cells.

We found that loss or reduction of Pten in SmoM2 mice initially increases expansion of proliferating GCPs through increased proliferation and decreased differentiation such that by P8 (one week after gene mutation at P0) the EGL is expanded in SmoM2-Pten^fl/+ and SmoM2-Pten^fl/fl mice compared to SmoM2. By P12, however the EGL expansion is greatly exhanced only in SmoM2-Pten^fl/flmice, and this is accompanied by an abundance of differentiated cells and a significant reduction in cell death. Our label retention assay involving BrdU injection at P10 revealed that a greater proportion of BrdU⁺ SmoM2-Pten^fl/flGCPs were Ki67^- at P12 than in SmoM2 mice. This result indicates either that by P10 the loss of Pten results in an increase in differentiation of GCPs compared to in SmoM2 mice, which would be consistent with our scRNA-seq analysis, but also likely due to a greater survival of the differentiated cells since the density of TUNEL⁺ particles is reduced in the EGL of SmoM2-Pten^fl/fl mice at P12. Consistent with this, it was shown that NeuN+ cells in SHH-MB tumors preferentially die compared to proliferating or quiescent tumor cells in a Ptch1 mutant model²⁶. This previous result is the opposite to what we observed in late stage SmoM2-Pten^fl/fl tumors where the proliferative comparments of the tumors has greater cell death than the NeuN⁺ or SYP⁺ regions, and a similar density of TUNEL particles to in SmoM2 tumors. Thus, a key reason for the accelerated tumor growth in Pten mutant SHH-MB models is the survival of the differentiated compartment.

In a mouse NeuroD2-SmoA1 model of SHH-MB Pten heterozygosity leads to earlier onset of disease and a differentiated/nodular cellular phenotype when Pten is deleted in the germline or conditionally throughout the brain (Nestin-Cre)¹². Heterozygous loss of Pten in GCPs using an Atoh1-Cre transgene to delete Pten^fl/+ and Ptch1^fl/flapparently also slightly increases the time of onset and produces a differentiated phenotype¹³. In contrast, in our sporadic SHH-MB model (SmoM2-Pten^fl/+) where only rare GCPs lack one copy of Pten (and express SmoM2), the time of onset and penetrance are not accelerated and the tumors have a classic histology. Only in rare SmoM2-Pten^fl/+ clones of cells in a minority of tumors where PTEN protein is lacking, likely due to loss of heterozygosity, did we observe a differentiated phenotype with reduced cell death as seen in SmoM2-Pten^fl/fl mice. The difference in the outcome of our model and the other two is either because of the SHH-activating mutations (SmoM2 vs SmoA1 expression or Ptch1 loss) or more likely because when all GCPs lack a copy of Pten and have activated SHH-signaling that the microenvironemnt provides an advantage for Pten heterozygous cells to form a differentiated tumor. It is also possible that another cell type targeted by the other two Cre transgenes (Netstin and Atoh1) supports tumor growth when Pten loss is heterozygous.

A previous study provided evidence that increased mTOR signaling in SHH-MB promotes spinal cord metastasis by expressing Shh in GCPs along with an activated form of AKT^{18, 19}. However, in our sporadic model and the two other models of Pten loss in mouse SHH-MB^{12, 13}, spinal cord metastasis is not increased. Thus it appears that the way in which the mTOR pathway is activates affects metastasis. Perhaps related to this, in a model of Trp53 heterozygous loss and Rictor homozygous loss, which reduces pAKT, GCPs were found to persist longer than normal in the EGL³⁴, which is the same as what we found when Pten was deleted. Thus, the way the pathway downstream of PTEN is altered disctates the subsequent behavior of the mutant GCPs.

In our previous study we showed that in Kmt2d mutant SHH-MB models, no difference in tumor size was detected at P12 but by P21 tumors lacking one or two copies of Kmt2d were more than twice the size of those with intact Kmt2d and had a classic cytoarchitecture. Thus, there is a delayed positive effect of reducing/ablating this chromatin modifier gene on tumor growth rate compared to loss of Pten where there is a rapid expansion of the EGL. Another major difference between secondary loss of the two tumor suppressors in SHH-MB is that Kmt2d mutant tumors have nearly fully penetrant spinal cord metastasis. Thus, continued mechanistic studies in mouse models of recurrent mutations seen in SHH-MB should provide valuable insights into how the disease is expected to progress and provide ideas for development of targeted therapies based on the cellular phenotypes of tumors not just the signaling pathways altered.

Materials and methods

Animals

The animal care and procedures were performed in accordance with the Memorial Sloan Kettering Cancer Center Institutional Animal Care and Use Committee guidelines (IACUC protocol: 07-01-001). Mice were kept on a 12/12 hour light/dark cycle and in temperature- and humidity-controlled rooms with ad libitum access to standard laboratory mouse chow and water. All transgenic mouse lines were maintained on a mixed genetic background containing 129, C57BL/6J, and Swiss Webster. Both sexes were used for the analysis. The following mouse lines were used in the study: Atoh1-Cre (JAX #011104), R26^LSL-eGFPcre (MASTR, JAX #018903)²², Atoh1-FlpoER (JAX #040091)²³, Pten^{fl/fl 25} and SmoM2 (#005130)²⁴. Genetic recombination was induced with FlpoER by injecting one 100µg/g dose of Tamoxifen (Tm) (Sigma-Aldrich) subcutaneously into the back of P0 mice. Tm was prepared in corn oil (Sigma-Aldrich) at 20 mg/mL and stored at 4°C.

Tissue preparation

For all histological analyses, mice were anesthetized with isoflurane and then perfused transcardially with room temperature (RT) phosphate buffered saline (PBS) with heparin (0.02 mg/mL) and then ice cold 4% paraformaldehyde (PFA, Electron Microscopy Sciences, catalog no: 15714). Brains and spinal cords were dissected and post-fixed in 4% PFA 3-4 days at 4°C and cryopreserved in 30% sucrose in PBS for ∼2 days at 4°C. Brains were embedded in Tissue-Tek OCT compound (Sakura Finetek). Brains were serially cryosectioned sagittally at 12 µm and collected on charged glass slides (Fisherbrand ColorFrost Plus) and stored at -20°C. Frozen spinal cords were coronally sectioned at 20um. Details of reagents are listed in Key Resources Table.

Hematoxylin and eosin (H&E) staining

H&E staining were obtained from Richard-Allan Scientific and includ hematoxylin 2 solution, eosin-Y, bluing reagent, and clarifier 2. Slides were rinsed in PBS for 5 min and incubated in hematoxylin 2 solution for 3 min and then rinsed in deionized water (diH₂O) and placed in staining reagents for 1 min each (diH₂O, clarifier 2, diH₂O, bluing reagent, diH₂O, eosin-Y). Dehydration and defatting were then performed (dH₂O-70% ethanol-95%–95%-100%-100%-xylene-xylene-xylene, 1 min each) and thenslides were mounted with a coverslip and DPX mounting medium (Electron Microscopy Sciences). Details of reagents are listed in Key Resources Table.

Immunofluorescence

Slides were washed 3 time for 5 min in PBS. When necessary, antigen retrieval was then performed by immersing slides in sodium citrate buffer (10 mM sodium citrate with 0.05% Tween-20, pH 6.0) for 20 min at 95°C followed by rinsing in PBS. Slides were then incubated in blocking buffer (5% BSA in 0.4% Triton-X100 in PBS (PBST)) for 1 hour at RT. Slides were then placed in primary antibody solution in blocking buffer at 4°C overnight. Primary antibody information and dilutions are listed in Key Resources Tables. Slides were then rinsed three times in RT 0.1% PBST and incubated in secondary antibody solution in blocking buffer (1:500) at RT for 1 hour. Slides then were incubated in Hoechst (Invitrogen, catalog #H3569 diluted 1:1000 in PBS) for 10 min and rinsed three times in PBS and cover-slipped with Fluoro-Gel (Electron Microscopy Sciences).

TUNEL Staining

TUNEL staining was according to th manufacturers instructions (Roche TUNEL assay kit). Tissue on slides were permeabilized with 0.5% TritonX-100 and then pre-incubated with 1x TdT reaction buffer with 5mM CoCl₂ and 1X PBS for 15 min at RT. Slides were then incubated for 1 hr at 37°C in 1x TdT reaction buffer containing Terminal Transferase (Roche) and Biotin-16-dUTP (Sigma-Aldrich) or Digoxigenin-dUTP (Enzo Life Sciences, Inc). Slides were then incubated with a Streptavidin Alexa Fluor 647 conjugate (Invitrogen catalog #S-32357) or Anti-Digoxigenin-Rhodamine (Sigma-Aldrich catalog), respectively for 1 hr.

BrdU Injection and Staining

To assess cell differentiation, BrdU (Sigma Aldrich) was injected intraperitoneally at 50 µg/g at P10 for analysis at P12. Tissues were collected and processed as described above. For immunostaining, antigen retrival was applied to slides with Rat primary anti-BrdU antibody BU1/75 (ICR1) (Abcam Inc.) to detect BrdU.

EdU (5-ethynyl-2’-deoxyuridine) Injection and Staining

To assess cell proliferation, EdU (Invitrogen) was injected intraperitoneally at 50 µg/g 1 hr before euthanasia. A Click-it EdU assay was used (Invitrogen) with Sulfo-Cyanine5 azide (Lumiprobe Corporation) per the manufacturer’s protocol to stain sections.

RNA in situ hybridization

Sagittal sections (12 um) were processed for in situ hybridization as described³⁵. Except for Gli1 and Syp, antisense riboprobes were labeled with UTP-digoxigenin using a DIG RNA labeling mix (Sigma-Aldrich) with PCR-amplified templates prepared from cDNA synthesized from an early postnatal mouse cerebellum lysate. Signal was detected with alkaline phosphatase-coupled anti-digoxigenin antibodies using BM purple (Sigma-Aldrich) as the substrate. The primers used for PCR amplification were based on the Allen Brain Atlas. Gli1 RNA in situ was performed using an antisense RNA probe made from a plasmid³¹ and for Syp BioMatik generanted a plasmid in Bluescript containing Syp sequences using the primers described in the Key Resources Tables. Primers were flanked in the 3’ with SP6 (antisense) and 5’ with T3 (sense) promoters. Primers for RNA in situ probes are listed in Key Resources Tables.

Microscopy, image processing, and analyses

Images were acquired with a DM6000 Leica fluorescent microscope (Leica Camera, Wetzlar, Germany) or NanoZoomer Digital Pathology microscope (Hamamatsu Photonics, Shizuoka, Japan) and were processed and analyzed using Fiji³⁶ or Photoshop (Adobe Inc., San Jose, CA, USA). Cell counts were manually obtained using the Cell Counter plugin for Fiji or semi-automated using a custom script using the Analyze Particle plugin for Fiji as previously described³⁷.

ScRNA-seq sample Preparation

Atoh1-FlpoER/+; R26^{MASTR/LSL-SmoM2};Pten^+/+(SmoM2) and Atoh1-FlpoER/+; R26^{MASTR/LSL-SmoM2}; Pten^fl/fl (SmoM2 Pten^fl/fl) mice were euthanized at P16 and P22, respectively, and the tumors were dissected from the lateral-posterior cerebellum in ice-cold Hank’s balanced salt solution (HBSS, Gibco). Then the tumors were dissociated in Papain (Worthington) for 30 minutes at 37°C before being washed out in Ovalbumin (Worthington). Tumors were triturated by pipetting up and down using a P1000 and filtered through a 40µm cell strainer (Falcon). Once the cells were dissociated, the dead cells were removed using the kit EasySep Dead Cell Removal (Annexin V) (StemCell). Then cells were centrifuged at 500g at 4°C for 5 minutes. Cells were then washed with PBS. For every sample (2 SmoM2 and 2 SmoM2 Pten^fl/fl), cells were counted using 0.2% Trypan on hemocytometer and diluted before loading to 10X chip, targeting 10,000 viable cells per sample. The scRNA-Seq of tumor cell suspensions was performed on a Chromium instrument (10X genomics) following the user guide manual for 3’ v3.1. Final libraries were sequenced on an Illumina NovaSeq6000 by the IGO core of MSKCC.

scRNA-seq data analysis

The Cell Ranger Single Cell software suite (10x Genomics) was used to align reads and generate feature-barcode matrices. The reference genome used was the Genome Reference Consortium Mouse Build 38 (GRCm38, Gencode annotation mm10). Raw reads were processed using the Cell Ranger count program using default parameters.

Seurat v5.0.2 package was used to generate a UMI (unique molecular identifier) count matrix from the Cell Ranger output³⁸. Genes expressed in less than 10 cells were removed for further analyses. Cells with larger than 18000 UMIs, 0.20 mitoRatio, UMI≤500 and ≤500 genes detected were considered low quality/outliers and discarded from the datasets. Normalization was performed on individual samples using the NormalizeData function with default parameters. The normalized data was scaled by ScaleData function with mitochondrial gene percentage regressed out. Principal component analysis (PCA) was performed on the scaled data by runPCA. Samples were then integrated using IntegrateLayers function with the Harmony Integration method. The first 20 dimensions were used for the FindNeighbors function, and clusters were identified using the FindClusters function with a resolution of 0.5. Data were projected onto the 2D space using the FindUMAP or RunUMAP function with 20 dimensions. Cluster markers and further differential gene expression analyses were performed using normalized counts (NormalizeData) in the RNA assay. Cluster markers were identified using the FindAllMarkers and comparing markers generated to the literature. To refine clustering further, the SubsetData function was used to create a new Seurat object, and the above clustering was reiterated.

Differential gene expression analyses between genotypes or clusters were performed using the FindMarkers function in the Seurat package. Genes with an adjusted P value (Padj) < 0.05 were considered significantly up or downregulated. Results were visualized by Violin plot and Umap plot using Seurat package, and Volcano plot using the EnhancedVolcano package on Bioconductor³⁹. GSEA analysis on GO and Hallmark terms were performed using the enrichGO, clusterProfiler packages on Bioconductor⁴⁰. To build a pseudotime trajectory using the UMAP data, the Seurat object was converted to a monocle 3 (1.3.7) object by as.cell_data_set function^{29, 30}. Cells were then clustered by the cluster cells function. The trajectory GCP-lineage partition was constructed using learn graph function followed by order cells function.

Statstic analysis

All statistical analyses to quantify data were performed using Prism software (GraphPad) and significance was determined as P <= 0.05. Each statistical analysis used is described in the figure legends. Data (shown as scatter graphs) were presented as mean±standard deviation.

Data Availability

Sequencing data have been deposited in GEO (token czkvuymyhnktrgp) under accession code GSE295733. All experimental data generated and analyzed in this study are included in the manuscript and supporting files.

Acknowledgements

We thank members of the Joyner lab for interesting discussions pertaining to SHH-MB and technical support, in particular Reeti Mayur Sanghrajka for insightful suggestions at the beginning of the study. We thank the Memorial Sloan Kettering Cancer Center Flow Cytometry Core, Integrated Genomics Operations Core Facility and the Center for Comparative Medicine and Pathology for technical support.

Additional information

Funding

NCI (R01CA192176), Alexandra L Joyner

MSK Functional Genomics Initiative (GC242211), Alexandra L Joyner

NCI Cancer Center Support Grant (CCSG, P30 CA08748), Alexandra L Joyner Cycle for Survival, Alexandra L Joyner

Ethics

No ethical consideration as no human data were analyzed.

Author contributions

Z.L., S.E.N., Y.L and A.L.J. formulated experiments and analysis approaches. Z.L. and D.N.S. performed animal and histological experiments. Z.L. performed section analysis and quantifications. Z.L., S.E.N. and Y.L. performed scRNA-seq analysis. A.L.J. wrote the draft manuscript and Z.L., Y. L., S.E.N. and A.L.J. prepared Figures and edited the manuscript.

Additional files

Key Resources Table

Supplementary Tables S1-9