Expansion of cancer stem cell pool initiates lung cancer recurrence before angiogenesis

Significance

Latent tumor cells are the crucial reason of tumor recurrence and the death of cancer patients. Preventing latent tumor relapse can prolong patients’ survival and have a long time surviving with latent tumor cells. Here, we describe a lung cancer suspensive tumor model in mouse and find that a high level of cancer stem cells undergoing asymmetric cell division in latent tumor is the key issue to reactivate a suspensive tumor. The results clearly delineate the state of latent tumor in vivo. A high level of serum IGF-1 can induce the suspensive-to-progressive tumor transition though promoting CSCs symmetric division, which illuminate a key checkpoint of cancer relapse before angiogenesis, highlighting a potential therapeutic target for preventing tumor recurrence.

Abstract

Angiogenesis is essential in the early stage of solid tumor recurrence, but how a suspensive tumor is reactivated before angiogenesis is mostly unknown. Herein, we stumble across an interesting phenomenon that s.c. xenografting human lung cancer tissues can awaken the s.c. suspensive tumor in nude mice. We further found that a high level of insulin-like growth factor 1 (IGF1) was mainly responsible for triggering the transition from suspensive tumor to progressive tumor in this model. The s.c. suspensive tumor is characterized with growth arrest, avascularity, and a steady-state level of proliferating and apoptotic cells. Intriguingly, CD133⁺ lung cancer stem cells (LCSCs) are highly enriched in suspensive tumor compared with progressive tumor. Mechanistically, high IGF1 initiates LCSCs self-renewal from asymmetry to symmetry via the activation of a PI3K/Akt/β-catenin axis. Next, the expansion of LCSC pool promotes angiogenesis by increasing the production of CXCL1 and PlGF in CD133⁺ LCSCs, which results in lung cancer recurrence. Clinically, a high level of serum IGF1 in lung cancer patients after orthotopic lung cancer resection as an unfavorable factor is strongly correlated with the high rate of recurrence and indicates an adverse progression-free survival. Vice versa, blocking IGF1 or CXCL1/PlGF with neutralizing antibodies can prevent the reactivation of a suspensive tumor induced by IGF1 stimulation in the mouse model. Collectively, the expansion of LCSC pool before angiogenesis induced by IGF1 is a key checkpoint during the initiation of cancer relapse, and targeting serum IGF1 may be a promising treatment for preventing recurrence in lung cancer patients.

Cancer recurrence, a major cause of cancer death, can be preceded by an interlude, termed tumor dormancy, that can last years or even decades without clinical symptoms (1). The incidence of cancer recurrence after aggressive cancer surgery is also high (2). Recent studies have indicated that minimal indolent tumors were the origins of metastasis and recurrence (3–5). However, the molecular mechanisms underlying tumor dormancy and rebirth are far from clear.

Currently, there are two distinct theories to explain tumor dormancy: Cellular dormancy describes that cells enter a quiescence status where growth is arrested in G₀/G₁ phase of cell cycle, and cells are completely inactive and asymptomatic (6); population dormancy presents that a cluster of tumor cells maintain a balance between proliferation and apoptosis without expansion (7). These dormant switches are two hurdles that must be overcome for tumor recurrence and metastasis initiation. When tumor cells disseminate into a new site, the interaction between the tumor cells and their microenvironment determines whether the cells enter proliferation or dormancy (8). Arrested cells can come back as recurrent tumor when the microenvironment changes after a long-time latency. Angiogenesis, immune response, and cytokine network may account for the transition between indolent and aggressive tumor (9–11). Therefore, research progress on the mechanisms about tumor microenvironment regulating relapse will provide promising targets for the treatment of recurrence and metastasis.

Cancer stem cells (CSCs) are generally considered to be responsible for tumorigenesis and cancer metastasis (12, 13). Importantly, the existence of CSCs may account for the relapse of cancer, especially after radiotherapy and chemotherapy. However, it is still debatable if CSCs differ from indolent tumor cells. CSCs are deemed as latent cells due to their low proliferation rate, and resistance to chemotherapy and radiotherapy, which is similar to the characteristics of dormant cells. However, it could be argued that although dormant cells can repopulate a whole system, they will not be considered as stem cells unless they possess self-renewal ability (14). CSCs have a unique biological process for self-renewal that one CSC may produce one daughter CSC via asymmetric cell division or two daughter CSCs via symmetric cell division, which ensures the CSC population to be maintained or expanded for long-term clonal growth (15). However, whether the imbalance of self-renewal in CSCs is responsible for cancer relapse has not been investigated.

In this study, we demonstrated that lung cancer stem cells (LCSCs) maintained a dynamic equilibrium of tumor cell proliferation and death via asymmetric cell division in indolent tumor. However, the division of LCSCs could be switched to symmetric cell division by a high level of insulin-like growth factor 1 (IGF1), which led to the rapid expansion of LCSC population and angiogenesis, resulting in lung cancer recurrence.

Results

Xenografting Human Lung Cancer Tissues Induces the Tumor Recurrence in Mouse.

Patient-derived xenograft (PDX) maintains the histopathological and molecular characteristics of the parental tumor, offering an exciting tool for studying targeted therapies (16, 17). Interestingly, we happened upon an interesting phenomenon that s.c. xenografting certain human lung cancer tissues in nude mice could induce the recurrence of suspensive xenograft tumor derived from nonsmall-cell lung cancer (NSCLC) primary cells LSC1 (Fig. 1 A and B). Synthetically, one-third (4/12) of lung cancer PDXs could initiate the suspensive tumor recurrence (Fig. 1C). To investigate the mechanisms of tumor relapse, we established s.c. xenograft suspensive tumors in nude mice by limiting dilution of inoculated NSCLC primary cells LSC1 (squamous cell carcinoma), LAC1 (adenocarcinoma), and cell line A549 (adenocarcinoma) (SI Appendix, Fig. S1A). When 1 × 10⁶ LSC1 or LAC1 cells and 1 × 10⁵ A549 cells were inoculated, progressive tumors (volume > 50 mm³) were detected in all tested mice within 4 wk (SI Appendix, Fig. S1B). However, when 1 × 10⁵ LSC1, 1.5 × 10⁵ LAC1, and 1 × 10⁴ A549 cells were inoculated, most of the xenograft tumors (∼70%) could maintain a prolonged latent period without tumor volume changes (volume < 50 mm³) (SI Appendix, Fig. S1B). Therefore, the tumor volume threshold between suspensive and progressive tumor was set to 50 mm³ (SI Appendix, Fig. S1C). Luciferase-labeled tumor cells for bioluminescent tracking indicated that most tumor cells died within 1 wk after injection, but the number of surviving cells barely changed until they reached a steady-state suspensive tumor (SI Appendix, Fig. S1D). Hence, these suspensive tumors were applied to investigate the triggers of tumor recurrence.

Fig. 1.

IGF1 triggers NSCLC recurrence. (A) Suspensive tumors were established by injecting a limited number of lung cancer cells into the right dorsal flank of nude mice and lung cancer tissues from different patients were s.c. transplanted into the left dorsal flank of suspensive tumor-bearing mice. An interesting discovery shows that lung cancer patient-derived xenografts (PDXs) from patient 1 but not patient 2 can induce suspensive tumor recurrence within 2 wk, suggesting only some human lung PDXs can awake s.c. suspensive tumor from a distant site in nude mice. (B) Preliminary study showed that 4/12 lung cancer PDXs could induce suspensive tumor recurrence. (C) The xenograft tumor volumes were measured every 2 wk. (D) Five of 14 human NSCLC PDXs could induce suspensive tumor recurrence but not normal lung tissues and PBS. (E) The expression of several key factors and their interactors were increased in recurrent tumor (RT) compared with suspensive tumor (ST) analyzed with transcriptome sequencing. (F) Tumor recurrence could be triggered by Matrigel (BD Biosciences, n = 9 mice) transplantation, but not inactivated Matrigel (n = 6 mice) and PBS control (n = 6 mice). (G) Eleven cytokines were divided into three groups to test their activation ability of tumor recurrence according to their functions in regulating cell proliferation, angiogenesis, and stemness. (H) Tumor recurrence could be triggered by group 1 factors that including IGF1, LIF, and SCF, but not the other two groups. n = 6 mice for each group. (I) Further tests showed that only IGF1 could induce suspensive tumor recurrence. n = 6 mice per group. (J) ELISA was used to test the human IGF1 levels in two groups of mice serum: mice bearing suspensive tumors (n = 9 mice) and mice bearing recurrent tumors induced by NSCLC PDX (n = 6 mice). ***P < 0.001 (two-tailed unpaired t test). (K) Immunohistochemistry staining showed the expression changes of phosphorylated IGF1 receptor (p-IGF-1R) in suspensive tumors 1 d after IGF1 stimulation. (Scale bar: 100 μm.) OC, outer capsule; TC, tumor cells. In D, F, H, and I, the threshold of tumor volume between suspensive tumors and recurrent tumors was indicated by the red dotted line in Right, and the ratio of tumor recurrence was also indicated.

Next, to further confirm the intriguing finding, NSCLC patient-derived cancer tissues and normal lung tissues xenograft (∼100 mm³) were s.c. transplanted into the left dorsal flank of nude mice, respectively. Note that these mice have already borne a suspensive tumor at the right side of mice before xenograft (SI Appendix, Fig. S2A). Intriguingly, about one-third (5/14) of human NSCLC PDXs rather than normal lung tissues could effectively trigger the transition from suspensive tumor to progressive tumor within 2 wk (Fig. 1D). Moreover, transcriptome sequencing of suspensive tumor and recurrent tumor indicated the high expression of several active factors and their receptors (Fig. 1E). These results implied that certain active stimuli deposited in NSCLC microenvironment might rouse suspensive tumor after xenografting.

IGF1 Triggers the Transition of Suspensive Tumor and Progressive Tumor.

Matrigel is a solubilized basement membrane matrix extracted from the Engelbreth-Holm-Swarm mouse sarcoma and is applied to angiogenesis and propagation of human tumors in immunosupressed mice (18, 19). Now that NSCLC PDXs could induce tumor recurrence, Matrigel (200 µL per mouse), heat-inactivated Matrigel, and phosphate buffer saline (PBS) were s.c. transplanted into the left dorsal flank of nude mice, respectively, to detect their activation effects on suspensive tumor. Activated Matrigel, but not heat-inactivated Matrigel, was more efficient to induce the recurrence of suspensive tumor (Fig. 1F). In addition, PBS (Fig. 1 D and F) and other inflammation stimulations (SI Appendix, Table S1) could not awaken the suspensive tumor.

It has been shown that Matrigel contained various biologically active proteins, such as IGF1, TGFβ, EGF, bFGF, and PDGF, that could increase the frequency of tumor xenograft (20). Hence, to identify the stimuli, we selected 11 cytokines that were enriched in Matrigel (SI Appendix, Table S2) (20, 21). These factors were divided into three groups according to their different bioactivities, such as promoting cell proliferation, stimulating angiogenesis, and regulating cell stemness, and administrated through tail i.v. injection with a high dose for per factor (Fig. 1G and SI Appendix, Table S2). Results showed that only group 1 factors, including IGF1, LIF, and SCF, could effectively induce the transition from suspensive tumor to progressive tumor (Fig. 1H). Separated tests further showed that IGF1 (2 μg per mouse) was only responsible for the initiation of the tumor relapse (Fig. 1I). In addition, the human IGF1 levels in mice serum were also compared by ELISA between mice with and without the human NSCLC PDX-triggered recurrence. Results showed that the average serum IGF1 levels of mice with recurrent tumors were significantly higher than that in mice with suspensive tumors (Fig. 1J).

Moreover, we also tested the responses of suspensive tumors to IGF1 under different dosages (50 ng, 400 ng, 2 μg, and 10 μg per mouse) and found that tumor recurrence could only be induced at the high level of IGF1 (≥2 μg per mouse) (SI Appendix, Fig. S2B). Western blotting showed that IGF1 receptor (IGF1R) in suspensive tumors was activated by the high dose of IGF1 (≥2 μg per mouse) 1 d after stimulation (SI Appendix, Fig. S2C). Furthermore, immunohistochemistry staining also confirmed the activation of IGF1 receptor (phosphorylated IGF1R, p-IGF1R) in suspensive tumor 1 d after IGF1 stimulation (2 μg per mouse) (Fig. 1K). Taken together, these findings suggest that a high level of IGF1 in serum is a hazardous factor of tumor recurrence for NSCLC patients.

A Dynamic Balance Between Proliferation and Apoptosis in Suspensive Tumor.

To further investigate the mechanism of tumor recurrence, we followed up the xenograft tumors for 14 wk after LSC1 cells inoculation. Intriguingly, hematoxylin and eosin (H&E) staining showed that indolent tumor constructed a characteristic hollow microsphere at 10 wk after LSC1 cells injection (Fig. 2A). This microsphere structure consisted of three parts: cavity inside, surviving tumor cells around, and capsule part outside (SI Appendix, Fig. S3A). Immunofluorescent (IF) staining with human pan-cytokeratin (pCK) antibody confirmed the several layers of surviving tumor cells (Fig. 2B and SI Appendix, Fig. S3A).

Fig. 2.

A high proportion of LCSCs maintains tumor latency. (A) Representative H&E staining images showed the formation of suspensive tumor at the indicated times after injection of 10⁵ LSC1 cells. (B) IF staining of surviving LSC1 cells in suspensive tumor stained with anti-pCK (red). Cell nuclei were counterstained with DAPI (blue). (C) Representative images of BrdU label retaining cells in suspensive tumor (ST) and progressive tumor (PT). The percentages of BrdU⁺ cells were summarized in Right. (D) Representative IF images of suspensive and progressive tumors stained with Ki67 (Left, red) or TUNEL assay (Right, red). The percentages of Ki67- or TUNEL-positive cells were summarized in Right. (E) Western blotting was applied to compare the expressions of total p38, phosphorylated p38 (p-p38), total ERK1/2, and phosphorylated ERK1/2 (p-ERK1/2) between suspensive tumors and progressive tumors. β-Tubulin was used as a loading control. Relative expressions of p38/p-ERK1/2 in suspensive tumors and progressive tumors were analyzed by ImageJ software. (F) Flow cytometry was used to analyze the percentages of CD133⁺ cells in suspensive tumors and progressive tumors. n = 4 xenograft tumors per group. (G) Representative IF images of suspensive and progressive tumors stained with anti-CD133 (red). Tumor cells were stained with human pCK (green), and cell nuclei were counterstained with DAPI (blue). The percentages of CD133⁺ cells in suspensive and progressive tumors were summarized in Right. (H) The stem sphere formation assay was performed to analyze to the stemness of suspensive tumors and progressive tumors that have been digested into single-cell suspension with Liberase (Roche). (I) Western blotting indicated that the stem cell markers, Nanog, Oct-4, and Sox-2, were up-regulated in suspensive tumors. β-Tubulin was used as a loading control. Relative expressions of these proteins were analyzed by ImageJ software. (J) The percentages of EdU⁺ tumor cells in suspensive and progressive tumors were summarized. In all images, two-tailed unpaired t test. Error bars represent mean ± SD; ***P < 0.001. (Scale bars: A and C, 100 μm; B, 200 μm; D, G, and J, 50 μm.)

To detect the proliferative activity of surviving tumor cells in suspensive tumor, thymidine analog BrdU was injected in vivo 1 d before the execution of nude mice. IF staining showed that all of the BrdU⁺ tumor cells were located in the periphery of surviving tumor mass, and the percentage of BrdU⁺ tumor cells in suspensive tumors was less than that in progressive tumors (Fig. 2C). Moreover, Ki67-staining showed that 4% and 22% of tumor cells were Ki67-positive in suspensive and progressive tumors, respectively (Fig. 2D), which suggested that cell proliferation activity was sustained in suspensive tumors, although it was much lower than that in progressive tumors. Furthermore, TUNEL assay in situ found that the apoptotic index in suspensive and progressive tumors was 4% and 2%, respectively (Fig. 2D). Since the levels of proliferation and apoptosis in suspensive tumors were equal, it was reasonable to believe that suspensive tumor was maintained in a stable structure with a dynamic balance between proliferation and apoptosis. Previous study has demonstrated that the ratio of phospho-p38 (p-p38) to phospho-ERK1/2 (p-ERK1/2) could represent cell survival status (22). When the ratio of p-p38/p-ERK is up-regulated, cells tend toward quiescence. Contrarily, cells are considered in proliferative status. Western blotting showed that the ratio of p-p38/p-ERK was markedly increased in suspensive tumors than that in progressive tumors (Fig. 2E), which indicated the slower cell proliferation rate in suspensive tumors for maintaining a dynamic balance between growth and death.

High Proportion of LCSCs in Suspensive Tumor.

Since CSCs have been considered as the origin of cancer recurrence (23), we hypothesized that the suspensive tumor might contain more CSCs compared with the progressive tumor. To confirm this hypothesis, CD133 as a LCSC marker was used to detect the proportions of LCSCs in suspensive and progressive tumors (24). First, flow cytometry tests (Fig. 2F) and IF staining (Fig. 2G) showed the enrichment of CD133⁺ LCSCs in suspensive tumor than that in progressive tumor. Furthermore, the stem cell microsphere formation frequency of suspensive tumor-derived single-cell suspension was dramatically higher than that from progressive tumor (Fig. 2H). In addition, Western blotting showed the higher expressions of stemness-associated proteins, Oct4, Nanog, and Sox2 (Fig. 2I), and multidrug resistance-associated proteins, ABCB1 and ABCG2 (SI Appendix, Fig. S3B), in suspensive tumors than that in progressive tumors. Given that the property of cosegregation of immortal DNA in CSCs, we explored the levels of label-retaining cells in suspensive and progressive tumors using EdU pulse–chase assay. Results showed that higher percentage of EdU⁺ tumor cells in suspensive tumor than that in progressive tumor were detected by IF staining, which further suggested the enrichment of LCSCs in suspensive tumor (Fig. 2J). Taken together, all these findings suggest the high proportion of LCSCs in suspensive tumor to maintain tumor dormancy.

IGF1 Promotes the Symmetric Cell Division of LCSCs.

Previous studies have demonstrated that CSCs had a property of cosegregation of immortal DNA, which was necessary for their self-renewal ability (25, 26). To investigate the dynamic switching of the LCSCs cell division between asymmetry and symmetry regulated by IGF1, we next administrated EdU pulse–chase assay in vivo. Double strands of parental cell DNA were prelabeled with EdU in vivo for 2 wk and then continued to culture cells for 4 wk without EdU (Fig. 3A). LCSCs could retain immortal DNA via cosegregation, but EdU-labeled DNA in non-LCSCs would be diluted after several times of passages. Moreover, IGF1 stimulation and BrdU incorporation were administrated at the same time 1 d before IF staining analysis. Symmetric cell division is described as that two daughter cells are EdU and BrdU positive, and the division that two daughter cells are BrdU positive, but only one daughter cell is EdU positive, was considered as asymmetric cell division (Fig. 3B, Left) (27). Results showed that the frequency of symmetric cell division in LCSCs in suspensive tumor was significantly increased (73 ± 2.6%) 1 d after IGF1 stimulation (Fig. 3B, Right). Intriguingly, when the suspensive tumor transformed into progressive tumors (2 wk after IGF1 stimulation), the frequency of symmetric cell division in LCSCs was decreased into a low level (8 ± 2.1%) (Fig. 3B). However, the percentage of BrdU⁺ tumor cells had no significant changes 1 d after IGF1 treatment, which suggested that the recurrence triggered by IGF1 was a result of boosting the symmetric cell division of LCSCs rather than stimulating cell proliferation (SI Appendix, Fig. S4A). Moreover, the increased expressions of three pluripotency transcription factors, Oct4, Nanog, and Sox2, in suspensive tumors 1 d after IGF1 stimulation were detected by Western blotting (SI Appendix, Fig. S4B).

Fig. 3.

IGF1 promotes tumor recurrence by inducing symmetric cell division of LCSCs. (A) EdU pulse–chase experiment in vivo. (B) Representative IF images of asymmetric and symmetric cell division in vivo. The percentages of EdU asymmetry (Asy) or symmetry (Sym) in LSC1 and LAC1 cells after IGF1 (500 ng/mL) or PBS treatment were summarized in Right. (C) Representative images of the paired-cell assay of CD133⁺ LCSCs (LSC1 cells) in vitro. (Right) Quantification of EdU asymmetry or symmetry in CD133⁺ LCSCs maintaining in 10% FBS-containing medium supplemented with IGF1 or PBS for 1 d. (D) Western blotting was used to detect the activation of IGF1R/PI3K pathway in CD133⁺ LCSCs from LSC1 and LAC1 cells under PBS and IGF1 stimulations for 2 d. β-Tubulin was used as a loading control. (E) Western blotting detecting the phosphorylation levels of β-catenin and the protein expressions of β-catenin downstream effectors (c-Myc and CD44) and its regulators (Snail and NUMB) in CD133⁺ LCSCs from LSC1 and LAC1 cells under PBS and IGF1 stimulations for 2 d. β-Tubulin was used as a loading control. Representative IF staining for β-catenin (F) and NUMB (G) in the paired-cell assay of CD133⁺ LCSCs (LSC1 cells). (Right) Quantification of EdU asymmetry or symmetry in CD133⁺ LCSCs maintaining in 10% FBS-containing medium supplemented with IGF1 or PBS for 1 d. (H) Summary of in vivo and in vitro EdU pulse–chase and paired-cell assays illustrate that IGF1 induces symmetric cell division of LCSCs. In all images, two-tailed unpaired t test. Data represent mean ± SD; **P < 0.01, ***P < 0.001. (Scale bars: 10 μm.)

To further confirm the functions of IGF1 in promoting LCSCs symmetric cell division, EdU pulse–chase labeling was also performed in vitro. Fluorescence-activated cell sorting was used to separate CD133⁺ LCSCs that were then cultured with differentiation medium supplemented (containing 10% FBS) with IGF1 (500 ng/mL) or PBS treatments for 1 d, respectively. IF staining showed that, compared with PBS treatments, the frequency of symmetric cell division for CD133⁺ LCSCs was significantly increased under the stimulation of IGF1 (Fig. 3C). Moreover, the expression levels of the embryonic stem cell transcription factors, such as Oct4, Nanog, and Sox2, were significantly increased in CD133⁺ LCSCs 1 d after IGF1 stimulation (SI Appendix, Fig. S4C). This evidence suggests that rapid expansion of LCSC pools induced by IGF1, although promoting the symmetric cell division of LCSCs, is necessary for the lung cancer recurrence.

IGF1 Stimulates LCSCs Symmetric Cell Division via PI3K/Akt/β-Catenin Axis.

Previous studies have demonstrated that IGF1 could regulate cell proliferation, apoptosis, differentiation, stemness, and metabolism through the activation of PI3K and MAPK signaling pathways (28, 29). To further investigate the mechanism of IGF1 in promoting LCSC symmetric cell division, we sorted CD133⁺ LCSCs from LSC1 and LAC1 cells and treated them with IGF1 (500 ng/mL) or PBS for 1 d. Western blotting was used to test the phosphorylation levels of several key mediators involved in the PI3K (Fig. 3D) and MAPK (SI Appendix, Fig. S4D) signaling pathways. The results found that the expressions of phosphorylated IGF1R, PI3K, Akt, and GSK-3β were increased in LCSCs treated with IGF1, compared with PBS-treated control cells (Fig. 3D). However, for MAPK pathway, the expressions of phosphorylated MEK1/2, ERK1/2, p90RSK, and ELK1 showed no obvious changes between IGF1- and PBS-treated cells (SI Appendix, Fig. S4D). To further confirm this finding, the PI3K inhibitor LY294002 (50 μM) and MEK1/2 inhibitor U0126 (10 μM) were used to investigate the correlation between the activation of pathways and the frequency of symmetric cell division in LCSCs under IGF1 or PBS treatments. Results showed that LY294002 but not U0126 could effectively inhibit the symmetric cell division of LCSCs (SI Appendix, Fig. S4E). In summary, IGF1 boosts the symmetric cell division of LCSCs via PI3K/Akt/GSK-3β axis.

β-Catenin, a key downstream modulator of Wnt signaling, is implicated in the early embryonic development and tumorigenesis (30, 31). GSK-3β can destabilize β-catenin by phosphorylating it at Ser33, Ser37, and Thr41 residues (32). However, phosphorylated at Ser552 and Ser675 sites by Akt can stabilize β-catenin protein and promote its nuclear localization (33). Therefore, the phosphorylation level of β-catenin in LCSCs treated with IGF1 or PBS was characterized by Western blot analysis. Results showed that the expressions of total β-catenin and phosphorylated β-catenin at Ser552 and Ser675 residues were increased, while phosphorylated β-catenin at Ser33/Ser37/Thr41 residues were decreased after IGF1 treatment (Fig. 3E), which suggested that IGF1 promoted the symmetric cell division of LCSCs by stabilizing β-catenin via PI3K/Akt/GSK-3β pathway. In addition, the expressions of two key downstream targets of β-catenin, c-Myc and CD44, that were involved in cell stemness regulation, were also increased after IGF1 stimulation (Fig. 3E).

Snail and NUMB are two crucial regulators for CSCs self-renewal. The former is able to interact with β-catenin and regulates its target genes expression (34), while the latter can promote β-catenin degradation via ubiquitylation (35). Western blotting showed the increased expression of Snail and the down-regulation of NUMB in LCSCs after IGF1 treatment (Fig. 3E), which further confirmed that β-catenin was a vital performer for IGF1-induced symmetric cell division of LCSCs. Moreover, when LCSCs undergo symmetric cell division, increased nuclear localization of β-catenin (Fig. 3F) and down-regulation of NUMB (Fig. 3G) were detected in both daughter cells. Collectively, these in vivo and in vitro results suggest that IGF1 boosts the symmetric cell division of LCSCs via PI3K/Akt/β-catenin axis (Fig. 3H).

IGF1 Induces Angiogenesis by Stimulating the Production of CXCL1/PlGF in LCSCs.

The angiogenic switch is a checkpoint for the transition from dormant avascular nodule to outgrowing vascularized tumor (36). In this study, angiogenic factors in suspensive tumor and recurrent tumor were detected by angiogenesis antibody array (AAH-ANG-1-4; Raybiotech) that can analyze 20 proteins associated with angiogenesis. Growth-regulated oncogene (GRO, including chemokine CXCL1, CXCL2, and CXCL3) and placental growth factor (PlGF) were up-regulated in recurrent tumor induced by IGF1 stimulation (Fig. 4A). Reverse transcription-PCR (RT-PCR) showed the significantly increased expressions of CXCL1 and PlGF in suspensive tumor 1 d after IGF1 stimulation (Fig. 4B).

Fig. 4.

IGF1 up-regulates the angiogenic factors CXCL1 and PlGF in LCSCs. (A) Angiogenic factors in suspensive tumor and recurrent tumor were detected by angiogenesis antibody array (Raybiotech, AAH-ANG-1-4). Growth-regulated oncogene (GRO) includes chemokine CXCL1, CXCL2, and CXCL3). (B) Reverse transcription PCR (RT-PCR) detecting the expressions of CXCL1/2/3 and PlGF in suspensive tumors at different times after IGF1 stimulation. (C) Real-time quantitative PCR showed the dramatically increased expressions of CXCL1 and PlGF in CD133⁺ and CD133⁻ cells that treated with IGF1 for 1 d in vitro, compared with PBS control. Data represent mean ± SD; ***P < 0.001. (D) Representative IF staining images of the new vessels (CD34⁺) in the suspensive tumors and recurrent tumors that have been stimulated by IGF1 (2 μg per mouse) for 1 wk. (Scale bars: 100 μm.) The number of capillaries per visual field in suspensive tumor (ST) and recurrent tumors (RT) were summarized in Right. Data represent mean ± SD; ***P < 0.001. (E) Schematic diagram showed that IGF1 could stimulate the secretion of CXCL1 and PlGF from CD133-positive LCSCs, which further results in angiogenesis.

To further analyze the relations between the increased symmetric cell division of LCSCs induced by IGF1 and angiogenesis, we separated CD133⁺ and CD133⁻ LCSCs from LSC1 and LAC1 cells and tested the expressions of these angiogenic factors before and after IGF1 stimulation by real-time quantitative PCR. Results showed the dramatically increased productions of CXCL1 and PlGF in CD133⁺ LCSCs treated with IGF1 for 1 d compared with PBS control (Fig. 4C). However, there were no significant changes of these angiogenic factors in CD133⁻ cells after IGF1 stimulation (Fig. 4C). Most importantly, the recruitment of activated fibroblasts (SI Appendix, Fig. S5A) and angiogenesis (Fig. 4D) were found in recurrent tumors 1 wk after IGF1 stimulation. RT-PCR also showed the higher expression of CXCL1 and PlGF in CD133⁺ LCSCs than that in CD133⁻ non-LCSCs (SI Appendix, Fig. S5B). These data illustrate that the increasing cell stemness via symmetric cell division under IGF1 stimulation promotes angiogenesis by promoting the production of CXCL1 and PlGF in CD133⁺ CSCs (Fig. 4E).

In addition, previous studies have shown that CSCs could accelerate vasculogenic mimicry (VM) formation via epithelium-to-endothelium transition for promoting cancer progression (37). Therefore, we performed the double staining of laminin 5 γ2-chain/pCK, laminin 5 γ2-chain/PAS staining, and CD31/PAS staining in recurrent tumors to confirm the role of VM during lung cancer recurrence under IGF1 stimulation. Results clearly showed the VM networks in recurrent tumors, suggesting that resurrecting LCSCs might also accelerate VM formation during lung cancer recurrence (SI Appendix, Fig. S5 C and D).

High-Serum IGF1 Is Related to Recurrence of NSCLC Patients.

Next, we tested the serum levels of IGF1 between healthy persons (n = 38) and NSCLC patients (n = 74) by ELISA. The serum samples of NSCLC patients were collected within 1 wk after orthotopic lung cancer resection. First, results showed the higher serum level of IGF1 in NSCLC patients than that in healthy persons, and NSCLC patients with recurrence had elevated serum IGF1 than those with remission (P < 0.001, Fig. 5A and SI Appendix, Fig. S6A). More importantly, patients with high serum level of IGF1 (>126.58 ng/mL) were dramatically related to the high rate of recurrence (P < 0.001, Fig. 5B). Kaplan–Meier survival analysis showed that high serum IGF1 level was significantly correlated with adverse progression-free survival for NSCLC patients (P = 0.0056, Fig. 5C), which suggested that IGF1 was a hazardous factor of NSCLC recurrence. Furthermore, IGF1R highly expressed in most human lung cancer cells, but IGF1 as a risk factor of recurrence only expressed in stroma cells (SI Appendix, Fig. S6B), indicating the potential of preventing tumor recurrence by blocking IGF1 pathway.

Fig. 5.

Targeting serum IGF1 prevents NSCLC recurrence. (A) ELISA was used to test the serum levels of IGF1 between healthy persons (n = 38) and NSCLC patients (n = 74). The serum samples of NSCLC patients were collected within 1 wk after orthotopic lung cancer resection. ***P < 0.001 (two-tailed unpaired t test). (B) Correlations between IGF1 serum levels and the recurrence rates of NSCLC patients. ***P < 0.001 (χ² test). (C) Progression-free survival curves of NSCLC patients based on serum IGF1 levels. P = 0.0056 (log-rank test). (D) Outline of the assay that mAb against IGF1 blocking IGF1-incduced recurrence in vivo. (E) Tumor recurrence induced by IGF1 (2 μg per mouse) could be blocked by neutralizing antibodies against IGF1 (50 μg per mouse). n = 6 mice for each group. (F) Outline of assay that using CXCL1 and PlGF mAbs to stop the IGF1-incduced recurrence 1 d after IGF1 treatment. (G) CXCL1 and PlGF mAbs could stop the IGF1-incduced recurrence 2 d after IGF1 stimulation.

Neutralizing Serum IGF1 Prevents NSCLC Recurrence.

Herein, we explored the feasibility of preventing the IGF1-induced tumor recurrence in mouse model by blocking serum IGF1 with neutralizing antibodies (Fig. 5D). Results showed that simultaneous injection of IGF1 mAb (50 µg per mouse) and IGF1 (2 µg per mouse) in 1 d could effectively inhibit the IGF1-induced tumor relapse (Fig. 5E). However, if injecting the IGF1 mAb at 2 d after IGF1 treatment, it could not prevent the activation of suspensive tumor (SI Appendix, Fig. S6 C and D). In addition, the serum levels of CXCL1 and PlGF were also increased in NSCLC patients with recurrence than those with remission (SI Appendix, Fig. S6E). Therefore, in this scenario, using mAbs-targeting CXCL1 and PlGF at 1 d after IGF1 stimulation could still stop the IGF1-induced tumor recurrence (Fig. 5 F and G). Therefore, timely and effectively lowering serum IGF1 in combination with antiangiogenesis might be an effective approach to prevent NSCLC relapse (Fig. 6).

Fig. 6.

A schema for illustrating the mechanism of IGF1 initiating the NSCLC recurrence. Suspensive tumor constructs a steady hollow microsphere and is characterized by resting growth, avascularity inside, and a steady-state level of proliferating and apoptotic cells. Moreover, a high proportion of LCSCs in suspensive tumor underwent asymmetric cell division (ACD). Superfluous serum IGF1 can induce resting tumor recurrence by promoting LCSCs from ACD to symmetric cell division (SCD), which leads to the expansion of the LCSC pool. In addition, increased LCSC population induced by IGF1 also results in angiogenesis by promoting the expression of CXCL1 and PlGF in CD133⁺ LCSCs. Hence, prophylactically blocking serum IGF1 is a promising approach for preventing NSCLC recurrence.

Discussion

Indolent cancer cells can relapse into the clinically detectable tumor after years of dormancy, but what eventually will switch them from latency to rebirth remains perplexing mysteries (38). Our lucky discovery that xenografting human lung cancer tissues can stimulate suspensive tumor recurrence in nude mice provides a chance to understand tumor latency and relapse. Here, we found that a high level of IGF1 could trigger lung cancer recurrence in mouse model, which was supported by the close association between high serum IGF1 and increasing risk of relapse in NSCLC patients. Therefore, our studies indicate the crucial role of IGF1 in NSCLC recurrence. IGF1 is a very interesting cytokine involving in cellular proliferation and apoptosis, angiogenesis, and metastasis, although it is old cytokines (39). A previous study shows that IGF1 varies with every healthy individual, and interestingly, patients with Laron Syndrome due to congenital IGF1 deficiency are protected from cancer development (40). Other studies showed that serum levels of IGF1 in lung cancer patients were significantly increased compared with the healthy control group, and NSCLC patients with metastasis had elevated serum IGF1 levels compared with those with nonmetastatic stage in NSCLC (41, 42). This evidence suggests that IGF1 plays a vital role in tumorigenesis and progression.

Accumulating evidence suggested that resting tumor cells possessed the characteristics of CSC, such as slow cycle (43), drug resistance (44), stemness (45), and tumor initiation (46). Consistently, we also found that suspensive tumor was composed of a high percentage of CD133⁺ LCSCs, which highly expressed multidrug resistant proteins (ABCB1 and ABCG2) and stemness-related proteins (Oct4, Nanog, and Sox2). Furthermore, label-retaining assay in paired daughter cells showed that most LCSCs in indolent tumors underwent asymmetric cell division. Asymmetric cell division is a common mechanism used by stem cell self-renewal to generate one daughter cell keeping stemness and a more differentiated daughter cell to perform functions, and symmetric cell division is necessary for the expansion of stem cell pool (47). Herein, we focused on the effects of IGF1 on the LCSCs self-renewal. After IGF1 treatment, the proportion of symmetric cell division of LCSCs was dramatically increased in suspensive tumors. This finding suggested that IGF1 could switch LCSC self-renewal from asymmetry to symmetry to induce a rapid expansion of an LCSC pool during recurrence. Further studies found that IGF1 could promote LCSC symmetric cell division via the PI3K/Akt/β-Catenin pathway. Therefore, our findings suggest that the asymmetric cell division of LCSC is responsible for the maintenance of the dynamic balance between proliferation and apoptosis.

Numerous studies have indicated that the presence of CSCs has been associated with angiogenesis (48). Inhibitors of angiogenesis induce hypoxia, which increases tumor growth rate and metastasis due to an increase in the CSC population (49, 50). Moreover, Bao et al. (51) have found that CSCs produce higher levels of VEGF in normal or hypoxic conditions than the non-CSC population in glioma. In our study, we found that CD133⁺ LCSCs highly expressed CXCL1 and PlGF in suspensive tumors. More importantly, IGF1 stimulation further induces the production of CXCL1 and PlGF in CD133⁺ LCSCs, resulting in angiogenesis and tumor recurrence. Hence, antiangiogenic therapy and therapy targeted at CSCs may provide a reasonable and promising option of treatment.

High circulating concentrations of IGF1 indicated the high risk of breast cancer (52), colorectal cancer (53), and lung cancer (54). Using a mice model, our study showed the high level of IGF1 is a high risk for tumor recurrence. We also found it, in NSCLC patients, was closely associated with the high risk of recurrence. Inhibition of IGF1 signaling, in most cases through targeting IGF1R, has been extensively studied in large-scale randomized clinical trials in lung cancer patients. However, these trials have showed no significant benefit for patients; even the drug appeared to worsen the patients’ prognosis (55). We think that wide expression of IGF1R in the multiple organs results in serious side effects for patients; however, targeting the receptor itself leads to a lot of adaptive responses, such as up-regulation of the insulin receptor and others. Therefore, targeting the receptor’s ligand IGF1 directly may avoid the adaptive responses. The results demonstrated that targeting IGF1 with neutralizing antibodies in mouse could effectively inhibit tumor recurrence, which provided a promising therapeutic strategy to prevent NSCLC relapse by lowering the serum level of IGF1. In addition, nutrition and dietary habits are key regulators in the level of circulating IGF1, and IGF1 levels can be reduced by fasting (56). Previous preclinical studies have been shown that dietary fat reduction could inhibit cancer progression in the mouse model (57, 58). Therefore, a low fat or plant-based diet may also inhibit lung cancer recurrence by reducing serum IGF1 levels. Moreover, prolonged fasting abated the immunosuppression through protection of lymphocytes from chemotoxicity in cancer patients (59). Furthermore, combining dietary fat reduction with IGF-1R blocking antibody therapy may contribute to the additive inhibition of lung cancer relapse.

Based on all our findings, we propose a schematic diagram of IGF1-mediated NSCLC recurrence (Fig. 6). Suspensive tumor constructs a characteristic steady microsphere characterized by growth-arrest and a steady-state level of proliferating and apoptotic cells. Moreover, a high proportion of LCSCs in suspensive tumor underwent asymmetric cell division. High serum IGF1 can induce resting tumor recurrence by promoting LCSCs from asymmetric cell division to symmetric cell division, which leads to the expansion of LCSC pool. Next, a high level of LCSCs under IGF1 stimulation promotes angiogenesis by increasing the expression of CXCL1 and PlGF in CD133⁺ LCSCs. Hence, preventatively blocking serum IGF1 is an attractive approach for preventing NSCLC recurrence.

Materials and Methods

Clinical Samples and Cell Lines.

All clinical specimens (NSCLC cancer tissues, normal lung tissues, and human serum samples) were collected from the Sun Yat-sen University Cancer Center (Guangzhou, China) and approved by the Committees for Ethical Review of Research at Sun Yat-sen University. The human NSCLC cell lines LSC1 (also known as SCC210011) and LAC1 (also known as ACC212102) were established in our laboratory (60). All of the cell lines were cultured in high-glucose DMEM (Life Technologies) supplemented with 10% FBS (Life Technologies) and incubated at 37 °C in a humidified chamber containing 5% CO₂.

Animal Studies.

All animal experiments were approved by the Animal Ethics Committee at Sun Yat-sen University Cancer Center (Guangzhou, China). To establish the indolent tumor model, different numbers of tumor cells were s.c. inoculated into the right dorsal flank of 4-wk-old female nude mice. After injection, the length (L) and width (W) of tumor were measured every week by calipers, and tumor volumes were calculated as volume (mm³) = L × W² × 0.5. If the tumor volume exceeds the threshold between indolent tumor and aggressive tumor (>50 mm³) within 4 wk, the mice would be excluded from follow-up experiments. Other mice with indolent tumor were randomly assigned according to random number table. For bioluminescent tracking, tumor cells expressing a lentiviral luciferase were tracked by i.p. injection of d-luciferin (150 mg/kg; Promega) and imaged with an IVIS Spectrum (PerkinElmer) at indicated times. For tissue heteroplasty, fresh human NSCLC and normal lung tissues were cut into pieces equally (∼80 mm³) with scalpels and s.c. transplanted into the left side of nude mice. After transplantation, all wounds were sterilized with 75% alcohol. If necessary, suturing of incisions was carried out in mice with leakage. For cytokines screening, 11 human recombinant proteins were disposed according to the specifications to maintain high bioactivity and were administrated by tail i.v. injection (SI Appendix, Table S2).

EdU and BrdU Administrations.

For BrdU incorporation assay in vitro, the DNA of parental cells was prelabeled with BrdU (10 μM; Roche) for 2 wk. Then, the cells were washed intensively, seeded on new culture dishes in medium without BrdU for 30 d. CD133-positive cells were enriched by flow cytometry, scattered into single cells for paired-cell formation, and seeded on coated coverslips in normal medium (DMEM+10% FBS) with IGF1 (500 ng/mL) or PBS control. After 24 h culture, the paired cells were treated according to the standard protocol of BrdU staining kit (11-299-964-001; Roche). The BrdU-positive cells were detected by primary monoclonal anti-BrdU (Roche) and Alexa Fluor 594 conjugated secondary antibody (Invitrogen). For EdU and BrdU dual labeling analysis in vivo, EdU (5 mg/kg per d; RiboBio) were administered first by i.p. injection for 2 wk when the nude mice had carried an indolent tumor for 30 d. After 30 d of EdU administration, the proliferating cells were labeled by injecting BrdU (30 mg/kg; Roche) into the lateral tail vein of mice and analyzed the labeled cells in the next day by fluorescent staining. EdU-labeling cells were analyzed according to the standard protocol (C10322-2; RiboBio). In addition, all tumor cells were stained with primary anti-pCK (Boster) and Alexa Fluor 488 conjugated secondary antibody (Invitrogen). Finally, slides were counterstained with DAPI (Life Technology) and visualized with a fluorescence microscope (OLYMPUS FV1000). Any label-negative paired cells and ambiguous segregation of BrdU or EdU were excluded from analysis.

Angiogenesis Antibody Array Assay.

Angiogenic factors in suspensive tumor and recurrent tumor were detected by angiogenesis antibody array (AAH-ANG-1-4; Raybiotech) that can analyze 20 proteins associated with angiogenesis. In brief, suspensive tumor and recurrent tumor were homogenized with protein lysis buffer (RIPA supplemented with 1 mM phenylmethylsulfonyl fluoride), respectively. Two milliliters of blocking buffer was added into each well and incubated for 30 min at room temperature. After aspirating blocking buffer from each well with a pipette, 100 μg of total protein was added into each well and incubated for overnight at 4 °C. Biotinylated Antibody Mixture was added into each well and incubated for 2 h at room temperature. After rinse with wash buffer, HRP-Streptavidin was incubated for 2 h at room temperature. Five hundred microliters of the detection buffer mixture was gently pipetted onto each membrane and incubated for 2 min. The sandwiched membranes were transferred to the chemiluminescence imaging system and expose. ImageJ software was used to measure the intensity of spots and local background around each spot. Normalized values were calculated by subtracting the background and normalizing to the positive control signals.

Statistics.

Statistical analyses were performed using the SPSS 18.0 (SPSS, Inc.). The independent Student’s t test was used to assess the statistical significance between any two preselected groups. χ² test was applied for comparison of dichotomous variables. The significance of Kaplan–Meier survival analysis was determined using log-rank tests. All statistical tests were two-sided and were considered statistically significant if P ≤ 0.05.

All other material and methods can be found in SI Appendix, Supplemental Material and Methods.