Loss of Dok-3 in Non-tumor Cells Induces Malignant Transformation of Benign Epithelial Tumor Cells of the Intestine

Benign tumors become malignant when they acquire the ability to breach epithelial basement membrane and invade into neighboring tissues, which is the first essential step of the invasion-metastasis cascade followed by intravasation, extravasation, formation of micrometastases, and colonization of distant organs. Metastasis is estimated to be responsible for around 90% of cancer-associated mortality, yet the underlying mechanisms of the invasion-metastasis cascade remain poorly understood (1).

It is widely accepted that accumulation of genetic alterations in driver genes, including oncogenes and tumor suppressor genes, causes tumorigenesis and tumor malignant progression, a multistep carcinogenic process initially documented in colorectal cancer (2). Emerging evidence suggests that the tumor cell–intrinsic gene mutations promote carcinogenesis not only by affecting tumor cells but also by shaping the tumor microenvironment (TME), which comprises activated and/or recruited non-tumor stromal cells such as fibroblasts, endothelial cells, lymphocytes, macrophages, and other types of inflammatory cells (3, 4). For example, the loss of the tumor suppressor Smad4 in tumor epithelial cells mutated in Apc promotes the invasiveness of intestinal tumors through recruitment of matrix metalloproteinase (MMP)-expressing CCR1⁺ bone marrow–derived cells to the invasion front (5).

Recent studies have further shown that even stromal cells can induce development and malignant progression of tumors by modulating neighboring epithelial and/or tumor cells. For example, glutathione peroxidase 4 (Gpx4) deficiency in myeloid cells induces tumorigenesis in various tissues and malignant progression of colorectal tumors through reactive oxygen species–mediated epithelial and/or tumor mutagenesis, including mutations in the driver genes: Apc, kras, and Tp53 (6), indicating stromal cell–dependent mechanisms underlying the epithelial and/or tumor cell–intrinsic mutations that cause tumor development and malignant progression. Similarly, Tgfbr2 deficiency in fibroblasts has been shown to induce prostate intraepithelial neoplasia and forestomach squamous cell carcinoma (7). Moreover, the loss of Smad4 or Lkb1 in T cells resulted in gastrointestinal carcinomas or polyps, respectively (8, 9). However, the involvement of epithelial mutagenesis in the above examples of stromal cell–driven tumor development remains unclear.

The Dok family of adaptor proteins consists of seven members that can be classified into three subgroups based on their structural similarities and expression patterns (10). Dok-1, -2, and -3 compose a subgroup, and are key negative regulators of hematopoietic growth and survival signaling. These Dok family proteins cooperatively inhibit macrophage proliferation and indeed Dok-1/Dok-2/Dok-3 triple knockout mice on a C57BL/6 and 129/Sv mixed background develop histiocytic sarcoma, an aggressive malignancy of macrophages (11). In addition, on a 129S1/SvImj genetic background, single, double, or triple deficiency in Dok1/2/3 genes causes lung cancer in mice (12), together indicating tumor suppressor function. However, the role of Dok-1/-2/-3 in malignant conversion of benign tumors remains unknown. Therefore, we employed the intestinal benign tumor model ApcMin/+ mice (13, 14) in which malignant progression has been well studied by adding mutations to tumor epithelial cells mutated in Apc (15).

In the current study, by using ApcMin/+ mice lacking Dok-1/-2 or Dok-3, we show that Dok-1/-2 and Dok-3 play distinctive roles in growth and malignant progression of intestinal tumors. Intriguingly, the loss of Dok-3 induced malignant progression in ApcMin/+ mice without significant stimulation of gene mutations in tumors. In addition, transplantation of Dok-3–deficient bone marrow cells also induced malignant progression in ApcMin/+ mice, indicating a novel tumor cell–extrinsic mechanism underlying the malignant conversion of benign tumors.

All experiments involving animals were approved by the Animal Ethics Committee of The Institute of Medical Science, The University of Tokyo (A21-21; Tokyo, Japan). ApcMin/+ mice (C57BL/6J-Apc^Min/J: stock no. 002020) and Csf1^op mice (B6;C3Fe a/a-Csf1^op/J: stock no. 000231) were obtained from Jackson Laboratory. Rag1 knockout (stock no. 002216), Cd4/8 knockout (stock no. 002664), Tcrd knockout (stock no. 002120), and Igh6 knockout (stock no. 002399) mice were obtained from Dr. Hiroshi Kiyono (The University of Tokyo, Tokyo, Japan) through the Jackson Laboratory. Dok-3 knockout mice on a C57BL/6 and 129/Sv mixed background were provided by Dr. Brian Seed (16), which had been backcrossed with C57BL/6 at least 12 times. Dok-1/-2 knockout mice on a C57BL/6J background were described previously (17). Mice from 6 to 7 months old were used in this study.

For histopathological analysis, tissues were fixed with 4% paraformaldehyde, paraffin embedded, and serially sectioned at 4 μm thickness. The paraffin sections were stained with hematoxylin and eosin (H&E). H&E-stained sections were analyzed using an optical microscope CX-33 (Olympus). Histologic assessment was reviewed in consultation with a pathologist. For IHC, paraffin sections prepared at 4 μm thickness were deparaffinized, rehydrated, and subjected to antigen retrieval by heating the sections at 90°C in HistoVT One (Nacalai Tesque). After inactivating endogenous peroxidases with 3% hydrogen peroxide and blocking in 2.5% normal goat or horse serum, the sections were incubated with a primary antibody against Ki67 (NeoMarkers), E-Cadherin, or CD3ε (Cell Signaling Technology). ImmPRESS Reagent Anti-Rabbit IgG and ImmPACT DAB (Vector Laboratories) were used for detection. After counterstaining with hematoxylin, the slides were mounted. For immunofluorescence, tissues were fixed in 4% paraformaldehyde, optical cutting temperature-embedded, snap-frozen in liquid nitrogen, and sectioned at 6 μm thickness. The frozen sections were incubated with the primary antibodies specific for CD45, CD11b, Gr1, B220 (BioLegend), Cytokeratin, or α-smooth muscle actin (SMA; Abcam), followed by Alexa Fluor 488–conjugated or Alexa Fluor 594–conjugated goat secondary antibodies (Molecular Probes/Thermo Fisher Scientific). After counterstaining with 4ʹ,6-diamidino-2-phenylindole (DAPI), the slides were mounted. All bright-field and fluorescence images were captured using a BZ-9000 microscope (Keyence). The mean Ki67-labeling indices were calculated as the number of Ki67-positive cells per total number of tumor cells by counting at ×120 in five independent microscopic fields per tumor (n = 3).

Nucleated bone marrow cells were collected from 8-weeks-old wild-type (WT) and Dok-3 knockout mice and intravenously injected into lethally irradiated ApcMin/+ mice (abbreviated as Apc mice). Irradiation was performed using an IBL-437C instrument (¹³⁷Cs, CIS Bio-International) at 9.5 Gy.

Total RNA was extracted from cells, tumors (n = 3 for Apc mice or ApcMin/+;Dok-3 knockout mice, abbreviated as Apc/Dok3 mice), or organoids (n = 3; generated as described previously; ref. 18) with ISOGEN (Nippon Gene) according to the manufacturer's instruction. cDNA was synthesized with PrimeScript RT reagent kit (Takara), and qRT-PCR analysis was carried out on a CFX Connect Real-Time PCR Detection System (Bio-Rad) using TB Green Premix Ex Taq II (Tli RNaseH Plus; Takara) and the following primer sets: Dok3 (F: TTTGGCAAGAAATGCTGGCG, R: CCCTGTGGACCTGTCGCCTG) and Gapdh (F: ATGGTGAAGGTCGGTGTGAACG, R: CGCTCCTGGAAGATGGTGATGG), Il1b (F: CTTGTGCAAGTGTCTGAAGCAG, R: AGGTCAAAGGTTTGGAAGCAGC), Il6 (F: ACCACTTCACAAGTCGGAGGC, R: CTGCAAGTGCATCATCGTTGTT), Il7a (F: ACCTCAACCGTTCCACGTC, R: GCTTTCCCTCCGCATTGAC), Cxcl1 (F: TCCAGAGCTTGAAGGTGTTGCC, R: AACCAAGGGAGCTTCAGGGTCA), Cxcl2 (F: CATCCAGAGCTTGAGTGTGACG, R: GGCTTCAGGGTCAAGGCAAACT), Cxcl10 (F: ATCCTGCTGGGTCTGAGTGG, R: AGGATAGGCTCGCAGGGATG), Ccl11 (F: TCCATCCCAACTTCCTGCTGCT, R: CTCTTTGCCCAACCTGGTCTTG), Tnfa (F: CCACGTCGTAGCAAACCACC, R: GACAAGGTACAACCCATCGGC), Cox-2 (F: TCCCGTAGCAGATGACTGCC, R: CCTTGGGGGTCAGGGATGAA). Semiquantitative RT-PCR analysis was performed with GoTaq (Promega) and the following primer sets: Dok1 (F: TTGGAGATGCTGGAGAATTCGC, R: AGTCAGTTCTGAGGATATCCTG), Dok2 (F: AGTGACTGGATACAGGCCATC, R: AGCAATGACCTTTTCTAAGGC), Dok3 (F: TTTGGCAAGAAATGCTGGCG, R: TCCATGGGAACAAAGCCCCT), Gapdh (F: GCACCACCAACTGCTTAGCC, R: CCATCACGCCACAGCTTTCC).

Mononuclear cells from size-matched pooled intestinal tumors (containing five tumors ≥2 mm in diameter) of 6-month-old Apc mice or Apc/Dok3 mice (n = 6 for each genotype) were stained with the following fluorochrome-conjugated antibodies (BioLegend): allophycocyanin (APC)-conjugated anti-CD3ε (145-2C11); APC/Fire 750-conjugated anti-CD11c (N418); Alexa Fluor 488–conjugated anti-Foxp3 (MF-14); Alexa Fluor 700–conjugated anti-F4/80 (BM8); Brilliant Violet (BV421)-conjugated anti-CD4 (RM4-5); Brilliant Violet (BV510)-conjugated anti-SiglecF (E50-2440) and anti-B220 (RA3-6B2); Brilliant Violet (BV605)-conjugated anti-CD11b (M1/70); FITC-conjugated anti-Ly6C (HK1.4) and anti-Gr1 (RB6-8C5); PE-conjugated anti-Ly6G (1A8), anti-TCRγ/δ (GL3), and anti-CD45 (30-F11); phycoerythrin-cyanine7 (PE-Cy7)-conjugated anti-CD8a (53-6.7). Antibody-stained cells were acquired with FACSAria (BD Biosciences) and analyzed with FACSDiva software Version 8.0.2 (BD Bioscience).

Size-matched pooled ileal tumors (containing 9–12 tumors ≥2 mm in diameter) and mouse tails were used for genomic DNA extraction (n = 3 mice for Apc or Apc/Dok3). Genomic DNA was extracted using lysis buffer (10 mg/L Proteinase K, 0.1 mol/L Tris-HCl pH8.0, 0.4% SDS, 5 mmol/L EDTA, 0.2 mol/L NaCl) followed by phenol-chloroform extraction and isopropanol precipitation. Extracted genomic DNA was processed at Macrogen using Truseq DNA PCR-Free library preparation (Illumina). For whole-genome sequencing (WGS), tumor DNA samples were sequenced to an average depth of 100 × coverage, and matched non-tumor (tail) DNA samples were sequenced to an average depth of 30 × coverage, with 150 bp paired-end reads on Illumina Novaseq 6000 instruments. The Illumina platform generates raw images and base calling through an integrated primary analysis software called RTA (Real-Time Analysis). The BCL/cBCL (base calls) binary was converted into FASTQ using Illumina package bcl2fastq2-v2.20.0. The demultiplexing option (–barcode-mismatches) was set to perfect match (value: 0).

Candidate somatic mutations were detected through the Genomon pipeline (https://github.com/Genomon-Project/genomon-docs/tree/v2.0). GRCm38/mm10 was used as the mouse reference genome. Candidate mutations in a tumor sample were identified with the following procedure: A candidate with (i) the Fisher exact P ≤ 0.01; (ii) 0.1 < strand ratio < 0.9; (iii) ≥5 variant reads in the tumor sample; (iv) the variant allele frequency (VAF) in the tumor sample ≥0.02; and (v) the VAF in the matched normal sample <0.1 was adopted. ANNOVAR (https://annovar.openbioinformatics.org/en/latest/) was used to annotate the detected variants.

Data were analyzed by Student t test, Fisher exact test, or the χ² test and were expressed as mean ± SD (Student t test) or percentage (Fisher exact and χ² tests). P < 0.05 was considered statistically significant.

Sequencing FASTQ data files from WGS have been deposited in the Japanese Genotype-phenotype Archive, which is hosted by the DDBJ, under accession number DRA015135.

To investigate the role of Dok-1, -2, and -3 in the development and malignant conversion of intestinal tumors, we crossed mice lacking Dok-1 and Dok-2, the closest homologs with redundant functions among Dok family adaptors (10), or Dok-3 knockout mice with ApcMin/+ mice (abbreviated as Apc mice), a mouse model of intestinal tumorigenesis bearing a heterozygous mutation in the tumor suppressor gene Apc (13, 14). At 6 months of age, the total number of tumors (1.0 mm ≤ Φ) in the small intestines of Apc mice lacking Dok-1/-2 (ApcMin/+;Dok-1/-2 knockout, abbreviated as Apc/Dok1/Dok2 mice) was significantly greater than that of Apc mice (Supplementary Fig. S1A). In particular, the number of relatively large (3.0 mm ≤ Φ), but not small (1.0 mm ≤ Φ < 3.0 mm), diameter tumors was significantly increased in Apc/Dok1/Dok2 mice compared with that in Apc mice (Supplementary Fig. S1B), indicating that Dok-1/-2 deficiency enhances intestinal tumor growth. In contrast, neither the size nor the total number of tumors in Apc mice was affected by Dok-3 deficiency, comparing Apc and ApcMin/+;Dok-3 knockout mice (abbreviated as Apc/Dok3 mice; Fig. 1A and B). However, Apc/Dok3 mice, but not Apc/Dok1/Dok2 mice, developed a significantly increased number of invasive tumors; namely those invading the submucosa, the muscularis propria, or reaching the serosal surface (abbreviated as Sm, MP, or Se in the figures, respectively; Fig. 1C–E; Supplementary Fig. S2A–S2C). Whereas most tumors in Apc mice were noninvasive (87%) and exhibited no deep invasion beyond the smooth muscle layer (muscularis propria), 56% of the tumors in Apc/Dok3 mice were invasive and 11% of the tumors penetrated beyond the muscularis propria and reached the serosal surface (Fig. 1D and E). Of note, depth of invasion indicates the deepest invasion area of each invasive tumor determined by H&E-stained serial sections obtained from whole small intestines (Supplementary Fig. S3; refs. 5, 19, 20). In addition, stromal reaction associated with these tumors supports their invasiveness rather than pseudoinvasiveness (Supplementary Fig. S4; ref. 21). Also, consistent with the previous observations showing that the invasion front of human colorectal adenocarcinomas displays low proliferative activity (22), the rate of proliferating Apc/Dok3 tumor cells in the invasion front was significantly lower than that in the tumor center (Supplementary Fig. S5A–S5C). Importantly, marked tumor invasion was also observed in colon tumors in Apc/Dok3 mice (Supplementary Fig. S6); 90% of colorectal tumors in Apc/Dok3 mice, but not in Apc mice, invaded the muscularis propria or beyond (Supplementary Fig. S6C). These results indicate that Dok-3 deficiency affects tumor malignant progression but not tumorigenesis in Apc mice.

To assess how Dok-1/-2 deficiency enhances tumor growth but Dok-3 deficiency induces malignant conversion of benign tumors, we first evaluated expression levels of the Dok family genes in tumor epithelial cells of Apc mice. Interestingly, although we observed mRNA expression of Dok-1 and Dok-2 in tumor epithelial cells, we failed to detect Dok-3 mRNA in the epithelial cells, but did find it in leukocytes (Fig. 2A; Supplementary Fig. S7). In addition, even in tumor organoid cells from Apc mice, the Dok-3 mRNA expression was undetectable (Fig. 2A; Supplementary Fig. S7). Given the growth inhibitory function of Dok-1/2 (12, 17, 23), these findings suggest that the loss of Dok-1/2 enhances tumor cell growth in a tumor cell–intrinsic manner, while Dok-3 may induce malignant conversion of benign tumors in a tumor cell–extrinsic manner. Thus, we transplanted Dok-3–deficient bone marrow cells, including hematopoietic stem cells that can systemically restore blood cells, into irradiated Apc mice to investigate the impact of Dok-3 loss in the bone marrow–derived stromal cells on tumor invasion. Because it has been reported that irradiation required for the transplantation induces intestinal tumor invasion in Apc mice (24), the abdominal region containing intestines of Apc mice was covered with a tungsten shield to protect the region from irradiation. As expected, most tumors in Apc mice transplanted with bone marrow cells from WT mice (Apc: BMT WT) remained noninvasive (82%) and exhibited no deep invasion beyond the muscularis propria (Fig. 2B–D), similar to those observed in Apc mice (Fig. 1C–E). In contrast, Apc mice transplanted with Dok-3–deficient bone marrow cells (Apc: BMT Dok-3 knockout) developed invasive tumors in higher proportion (60%) than Apc: BMT WT, including some reaching the serosal surface (7%; Fig. 2B–D), similar in pattern to that observed in Apc/Dok3 mice compared with Apc mice (Fig. 1C–E). These findings indicate that bone marrow–derived stromal cells lacking Dok-3 have the potential to induce invasion of Apc tumors in mice, demonstrating a previously unrecognized, tumor cell–extrinsic mechanism in malignant progression.

Accumulation of driver mutations in genes including Apc, Braf, Kras, Smad4, Tp53, or Pik3ca is known to cause invasive tumor development in mouse intestine (15, 25–27), and these mutations in humans are also associated with colorectal malignancies (28, 29). Moreover, benign tumors in Apc mutant mice acquire invasive ability due to additional driver mutations (15), and myeloid cells lacking the Gpx4 gene induce increased mutation frequencies in tumors as reflected in their whole-exome sequences, causing malignant progression of azoxymethane (AOM)-induced colonic benign tumors (6). These findings raise the possibility that Dok-3 deficiency enhances gene mutation frequency in tumor cells bearing the ApcMin mutation to generate additional driver mutations required for malignant conversion. Thus, we performed WGS using tumors from Apc mice possessing or lacking Dok-3. The WGS data clearly demonstrate that the total number of mutations in tumors from Apc mice was not affected by Dok-3 deficiency (Fig. 3A). Furthermore, the distribution of mutations across the genome, their substitution patterns, and the number of insertion-deletion (indel) mutations were comparable between tumors from Apc mice and those from Apc/Dok3 mice (Fig. 3A and B). Indeed, somatic mutations in 95 genes significantly mutated in colorectal cancer, including six known driver genes (APC, BRAF, KRAS, SMAD4, TP53, and PIK3CA; refs. 28, 29; Supplementary Table S1), were not observed in tumors from Apc mice, irrespective of the presence or absence of Dok-3. These findings together indicate that Dok-3 deficiency that drives malignant conversion of benign tumors had no apparent effect on mutagenesis in tumors, suggesting a mutagenesis-independent progression of malignancy induced by non-tumor cells.

Because hematopoietic reconstitution of Apc mice with Dok-3 knockout bone marrow cells induced a level of invasive tumors comparable with Apc/Dok3 mice (Figs. 1C–E and 2B–D), we first examined the impact of lymphocyte depletion in Apc/Dok3 mice by crossing those mice with DNA recombinase Rag1-deficient mice, which lack mature T and B lymphocytes (30). Interestingly, the rate of invasive tumor occurrence in Apc/Dok3 mice lacking Rag1 (26%) was significantly lower than that in Apc/Dok3 mice (56%; Fig. 4A and B). Furthermore, the rate of tumors invading the muscularis propria or beyond in Apc/Dok3 mice, 18% or 11%, was reduced to 2% or 0%, respectively, in Apc/Dok3 mice lacking Rag1 (Fig. 4C). Therefore, to identify lymphocyte populations essential for tumor invasion in Apc/Dok3 mice, we crossed Apc/Dok3 mice with CD4 and/or CD8-deficient mice to deplete CD4⁺ and/or CD8⁺ T cells. Note that the development of B cells and myeloid cells is unaltered by CD4 and CD8 deficiency in mice (31, 32). The rate of invasive tumors in Apc/Dok3 mice lacking CD4⁺ T cells (37%) or CD8⁺ T cells (35%) or CD4⁺ and CD8⁺ T cells (13%) was significantly lower than that in Apc/Dok3 mice (56%; Fig. 4D and E), and the rate of invasive tumors in Apc/Dok3 mice lacking both CD4⁺ and CD8⁺ T cells was indistinguishable to that in Apc mice (13% vs. 13%; Fig. 4E). Furthermore, 11% of the tumors reached the serosal surface in Apc/Dok3 mice, whereas such invasive tumors were not observed in either Apc/Dok3 mice lacking CD4/CD8 or Apc mice (Fig. 4F). In contrast, Apc/Dok3 mice lacking Igh6 or Tcrd, which lack B lymphocytes or γδT cells, respectively, show similar levels of malignant progression to Apc/Dok3 mice (Fig. 4G–L). In addition, aside from a suppression of small tumor (1.0 mm ≤ Φ < 2.0 mm) numbers observed in Apc/Dok3/Cd4/Cd8 mice compared with Apc/Dok3 mice, tumorigenesis in Apc/Dok3 mice was not significantly affected by depletion of any of the tested lymphocyte populations (Fig. 4M). Together, these findings demonstrate that CD4⁺ and CD8⁺ T cells, but not B and γδT cells, have an essential role in malignant progression of benign tumors in Apc/Dok3 mice, although the underlying mechanisms are yet unclear.

Inflammatory responses play critical roles in the development and malignant progression of tumors. Given that immune cells infiltrating the microenvironment, a hallmark of inflammation, is known to affect tumor cell malignancy (33), we examined the impact of Dok-3 deficiency in Apc mice on inflammatory status of tumors. Although CD4⁺ and CD8⁺ T cells are essential for malignant progression in Apc/Dok3 mice (Fig. 4), leukocytes, including these T cells, myeloid cells, granulocytes, and B cells did not markedly accumulate in tumors of Apc/Dok3 mice beyond the levels observed in Apc mice (Fig 5A and B). In addition, flow cytometric analysis showed that the frequencies of individual immune cell populations among CD45⁺ leukocytes in tumors of Apc/Dok3 mice were similar to those in Apc mice aside from monocyte and CD4⁺ T-cell populations, which were slightly higher and lower in Apc tumors lacking Dok-3, respectively (Fig. 5C and D). Furthermore, expression levels of inflammatory cytokines, chemokines and COX-2 in tumors of Apc/Dok3 mice were not significantly increased in comparison with those observed in tumors of Apc mice (Fig. 5E). These findings demonstrate that generation of an inflammatory microenvironment is unlikely to be a primary driver of malignant progression in Apc/Dok3 mice.

In the current study, we show that Dok-3 deficiency induces malignant progression in the intestinal benign tumor model, Apc mice (Fig. 1C–E). Interestingly, the loss of Dok-1/-2, the homologs of Dok-3, did not significantly affect malignant progression but rather enhanced tumor growth in Apc mice (Supplementary Fig. S1 and S2), indicating distinct roles of Dok-3 and Dok-1/-2. Although the underlying mechanism remains unclear, distinct binding partners of Dok-1/-2 and Dok-3 adaptors might cause these differences. For instance, p120 rasGAP selectively binds Dok-1/2 (10); in contrast, Grb2 preferentially binds Dok-3 (34), though any involvement of Grb2 in tumor invasion remains unclear.

Although we cannot completely exclude the possibility that the loss of Dok-3 in intestinal epithelial cells by itself can induce malignant progression, the expression of Dok3 mRNA was not detectable in either quantitative or semiquantitative RT-PCR analyses (Fig. 2A; Supplementary Fig. S7). Therefore, it is unlikely that the loss of residual Dok-3 expression causes malignant progression in Apc mice. Indeed, transplantation of Dok-3–deficient bone marrow cells induced tumor invasion in Apc mice (Fig. 2B–D) to an extent similar to that observed in Apc/Dok3 mice (Fig. 1D and E), indicating that Dok-3–deficient non-tumor cells have the potential to induce malignant conversion of benign tumors, namely a previously unknown, stromal cell–induced malignant progression in a spontaneous tumor model. Although the underlying mechanisms of Dok-3 loss–induced malignant conversion await further investigation, our examination of Apc/Dok3 invasive tumors found no signs of epithelial–mesenchymal transition (EMT), which is involved in a wide range of malignant progression (35). In detail, Dok-3 loss had no impact on the expression of EMT markers in Apc tumors, including EMT-inducing transcription factors (Supplementary Fig. S5E). Furthermore, tumor cells in the center and invasion front of tumors retained expression of epithelial cell marker E-Cadherin in Apc/Dok3 mice (Supplementary Fig. S5A–S5D), whose loss is a hallmark of EMT. Tumor invasion without inducing EMT is also reported in Apc mutant mice lacking the negative regulator of Notch signaling, Aes, in tumor epithelium (19). However, the impact of Dok-3 loss on Notch signaling remains to be investigated.

As mentioned above, the loss of Gpx4 in myeloid cells induced malignant progression in the AOM-induced colonic benign tumor model (6). However, the rate of invasive colonic tumors in myeloid Gpx4-deficient AOM mice (∼10%) was about 10-fold lower than that in Apc/Dok3 mice (100%; Supplementary Fig. S6B). Furthermore, although Gpx4 deficiency in myeloid cells significantly increased exome mutagenesis in tumors (6), which eventually led to malignant progression, our WGS analysis demonstrates that the loss of Dok-3 did not enhance genomic mutations in tumors of Apc mice, in terms of both levels and patterns of mutations (Fig. 3). Thus, these findings suggest a novel tumor cell mutagenesis-independent mechanism underlying malignant progression induced by the loss of Dok-3 in stromal cells.

Interestingly, depletion of CD4⁺ or CD8⁺ T cells partially but significantly inhibited the tumor invasion observed in Apc/Dok3 mice (Fig. 4D–F). In addition, Apc/Dok3 mice lacking both CD4⁺ and CD8⁺ T cells showed no enhancement of malignant progression in comparison with Apc mice (13% vs. 13%: the rate of invasive tumors in Fig. 4E), indicating an essential and cooperative role of CD4⁺ and CD8⁺ T cells in the Dok-3 loss–induced tumor invasion in Apc mice. In contrast, depletion of B or γδT lymphocytes had no impact on tumor invasion in Apc/Dok3 mice (Fig. 4G–L). However, the mechanism by which CD4⁺ and CD8⁺ T cells mediate malignant progression in Apc/Dok3 mice has not yet been elucidated. Given that tumors, including invasion fronts, in Apc/Dok3 mice did not show any increased infiltration of CD4⁺ and CD8⁺ T cells in comparison with Apc mice (Fig. 5A, B, and D), these lymphocytes might exert distal effects, for instance, via exosomes, because immune cell–derived exosomes are known to regulate cancer progression and metastasis (36). Indeed, CD8⁺ T cell–derived exosomes are reported to promote invasion of melanoma and lung cancer cells by increasing the expression of MMP9 via Fas signaling in the malignant cells (37). Furthermore, given the possible distal effects of T cells, comparison between Dok-3 expression in peripheral CD4⁺ and CD8⁺ T cells from adenocarcinoma-bearing patients and that from non–adenocarcinoma-bearing patients with familial adenomatous polyposis might help to gain insight into the clinical relevance of Dok-3 loss–induced malignant progression, even though such datasets appear unavailable at this point.

It has been shown that Th2-type CD4⁺ T cells indirectly promote invasion and subsequent metastasis of malignant tumors (mammary adenocarcinomas) by directly regulating tumor-associated macrophages (38), which are the most abundant immune cells in the TME and promote cancer initiation and malignant progression (39). Although it is unknown whether macrophages are involved in malignant conversion of benign tumors in ApcMin/+ mice lacking Dok-3, it might be possible that Dok-3–deficient macrophages play a role in the malignant progression by cooperating with CD4⁺ and CD8⁺ T cells. However, we could not investigate the involvement of macrophages in tumor invasion in Apc/Dok3 mice, because we failed in our attempt to breed mice lacking Dok-3 and Csf1, the latter of which is critical for development of macrophages (40). Namely, when crossing between ApcMin/+;Dok-3 knockout mice bearing a heterozygous mutation for Csf1 (Csf1^op/+; ref. 41) and Dok-3 knockout;Csf1^op/+ mice, Csf1^op homozygous mutant mice (Csf1^op/^op) lacking Dok-3 were absent in the expected numbers at weaning regardless of Apc mutation status (Supplementary Table S2) save for one ApcMin/+;Dok-3 knockout mouse, which died at 1 month of age. These results suggest severe lethality of Dok-3 knockout;Csf1^op/^op mice for reasons not understood.

Tumor-intrinsic accumulation of genetic alterations is reported to drive malignant progression by shaping the inflammatory microenvironment. For instance, the loss of Smad4 in tumor epithelial cells mutated in Apc enhances expression of the inflammatory chemokine CCL9 to promote malignant progression of the Apc mutant tumors by recruiting MMP-expressing CCR1⁺ bone marrow–derived cells as mentioned earlier (5). However, enhanced gene mutations or inflammatory responses in tumors were not detected in Apc/Dok3 mice (Figs. 3 and 5), suggesting an as yet unidentified mechanism underlying the malignant progression.

Together, our findings uncover previously unrecognized tumor cell–extrinsic cues that can induce malignant conversion of benign tumors irrespective of exacerbating gene mutations or inflammatory responses in tumors. However, the current study does not sufficiently address mechanisms, including how T cells and potentially other stromal cells promote tumor invasion, nor the clinical relevance of the proposed concept. Thus, further studies are required for better understanding stromal cell–driven malignant progression and developing new therapeutic strategies targeting such malignant progression.

S. Arimura reports grants from Japan Society for the Promotion of Science (JSPS) during the conduct of the study. Y. Yamanashi reports grants from JSPS and The Research Foundation for Microbial Diseases of Osaka University during the conduct of the study. No disclosures were reported by the other authors.

S. Arimura: Conceptualization, resources, funding acquisition, investigation, methodology, writing-review and editing. A. Inoue-Yamauchi: Resources, investigation, writing-original draft, writing-review and editing. K. Katayama: Investigation. T. Kanno: Investigation. H. Jozawa: Investigation. S. Imoto: Supervision. Y. Yamanashi: Conceptualization, resources, supervision, funding acquisition, methodology, writing-original draft, project administration, writing-review and editing.

We thank R. F. Whittier and H.M.A. Mohamed for critical reading of the article and thoughtful discussions and M. Nojima for statistical assistance. We also thank A. Takahashi and T. Yagami for technical assistance. This work was financially supported by JSPS KAKENHI grant number 18K19470 (to Y. Yamanashi), JSPS KAKENHI grant number 15H05574 (to S. Arimura) and The Research Foundation for Microbial Diseases of Osaka University.