Introduction

How cancer arises from a single normal cell is still the subject of active debate, affecting intervention strategies. While many cells may harbor oncogenic mutations, only a few unpredictably end-up developing a full-blown tumor^1,2,3,4. Various theories have been proposed to explain that transition^5,6,7,8, but none has been tested in vivo at the single cell level. Cancer initiation is thus believed to be a rare event taking place at the level of individual cells^{9,10,11,12,13} arising as a result of the accumulation of genetic mutations in so called Mut-driver genes (oncogenes such as KRAS^14,15, tumor suppressor genes¹⁶ such as TP53 or life-span¹⁷ genes such as TERT). Notably, mutant KRAS is the most frequent driver of several cancers^18,19 : about 27% of all human cancers, 45% of colorectal and 90% of pancreatic cancers²⁰.

Recently, it has been shown that genes involved in embryonic development, pluri/multipotency and cell reprogramming such as VENTX/NANOG and POU5/OCT4 are abnormally reactivated in late cancer stages, where acting as Epigenetic Drivers (Epi-Drivers) they empower cancer cells with Cancer Stem Cell (CSCs) features, resistance to anti-cancer therapies and potential for cancer recurrence/relapse^{21,22,23,24,25,26,27}. Although it is evident that such Epi-Drivers confer a selective advantage to CSCs in a full-blown cancer, whether they play a role during the early phases of malignant transformation is still unknown.

Current approaches to the study of cancer use constitutive or conditional expression of Mut-driver (or Epi-driver genes²⁸) in specific tissues, i.e. in many cells, even though only a small subset of these cells eventually leads to the growth of tumors²⁹, often observed when the tumor already consists of many thousands of heterogeneous abnormal cells. Hence, carcinogenetic processes observed among sibling organisms^30,31,32 occur with variable latency period from the onset of induction, at different locations and develop asynchronously. As a result, the initial stages of tumorigenesis are difficult to study, the state of the cell(s) of origin difficult to assess and control while its cellular environment is perturbed by the induced expression of the Mut- or Epi-driver genes. Due to the rarity of the event in vivo³², a statistically relevant single-cell tracking and characterization of the early stages of tumorigenesis has therefore never been done. This emphasises the need to predict or control the cell undergoing malignant transformation in vivo in order to pave the way for a study of the cellular and molecular events involved in the initial stages of tumorigenesis.

To address these issues, we have developed an optogenetic approach to control the expression of an oncogene in a single cell³³, with the goals of: (1) measuring the probability of the malignant transformation of a single cell in a live organism in various backgrounds and (2) tracking and characterising the development of a tumor from the original cell. In the following we show that the synergy between only two factors: the oncogene kRasG12V and a reprogramming factor (Ventx, Nanog or Oct4) increases the probability of carcinogenesis from a single cell by many orders of magnitude when compared to the expression of either of these genes (or none).

Optogenetics approaches allow for the photocontrol and monitoring of the activity of biomolecules in vivo. The approach we developed uses a photoactivable analog of tamoxifen (caged cyclofen, cCYC) to control the activity of proteins fused to the ERT-receptor^34,35 (a modified estrogen binding domain³⁶) (Fig. 1A). These protein constructs are sequestered by cytoplasmic chaperones. Once cCYC is uncaged by light (with one-photon illumination at ∼375 nm or two-photon at ∼750 nm), cyclofen (CYC) is released³⁴. It binds to the ERT-receptor and releases the fused protein (e.g. a Cre-ERT recombinase) from its complex with cytoplasmic chaperones³⁵ (Fig. 1A).

FIGURE 1:

Results

We used this approach to photocontrol the activity of a Cre/loxP recombination system in a transgenic zebrafish line (Tg(actin:loxP-EOS-stop-loxP-KRASG12V-T2A-H2B-mTFP; ubi:Cre-ERT; myl7:EGFP) carrying a floxable (loxP flanked) EOS gene (coding for a green fluorescent protein) upstream from the oncogene^33,34 (KRASG12V). In such a transgenic zebrafish line hereafter referred as KRASG12V line (Fig.1B, Fig.S1B) the expression of EOS can be switched to KRASG12V and H2B-mTFP (nuclear blue fluorescence) upon light-mediated uncaging of cCYC and CRE-ERT activation (Fig. 1C); mTFP fluorescence can be observed about 30min post-illumination (Fig. S1B) and is stably maintained in zebrafish (Fig. S1C). Consistent with an acquired refractory cell state and a loss of oncogenic competence³⁰, whole body expression of the oncogene at 1dpf did not result in tumorigenesis.

To test for the possible impact of a reprogramming factor on the malignant transition, embryos from this transgenic line were injected at the one-cell stage with the mRNA of a construct consisting of a glucocorticoid receptor (GR) fused with a reprogramming factor (Ventx, Nanog or Oct4)^24,37 (Fig. S3). The resulting protein (e.g. Ventx-GR) is sequestered by cytoplasmic chaperones and transiently activated upon incubation of zebrafish in Dexamethasone (Dex) (Fig. S2A). Activation of the oncogene³³ at 1dpf, only if followed by transient activation of Ventx (or Nanog or Oct4) did yield reproductible hyperplasic outgrowths (Fig. S3B-E). Immuno-histo-chemistry detection of phospho-ERK activity (Fig. S4A), Hematoxylin & Eosin staining (Fig. S4B) and RT-qPCR of selected genes (Fig. S5 and S6A,B) all display features associated with tumorigenesis. None of the controls (activation of kRasG12V only, Ventx-GR only, incubation in cyclofen or dexamethasone, etc.) developed hyperplasia, but rather grew into normal zebrafish (Fig. S3).

Since both kRasG12V and Ventx have been reported to play a role in brain cancer³⁸, we decided to activate the oncogene kRasG12V at 1dpf in a single normal cell of the brain of a transgenic zebrafish (injected with the mRNA of Ventx-GR at 1 cell stage). Activation of the oncogene in a single cell was achieved by illumination at 405nm to uncage cCyc in a small region (diameter ∼80 μm) of the brain (in the vicinity of the otic vesicle) (Fig. 1). On average the oncogene is then activated in a single cell, identified within ∼1h by the blue fluorescence of its nuclear marker (H2B-mTFP), see Fig. 1. Notice that single-cell activation of the oncogene alone does not give rise to cancer (Fig. S8A). However, if following the local expression of the oncogene (and its fluorescent marker), Ventx-GR is transiently activated the cell divides and proliferates (Fig. 2). We observed that at 1-day post induction (dpi) 50% of the induced cells had divided and expanded clonally by a factor ∼3 (Fig. 2B), while the other 50% had neither divided nor died. Surprisingly, at 3-5 dpi, we observed in all zebrafish larvae that the induced cell gave rise to progeny that display short-range dispersion (Fig. S7A, B) and then give rise to a tumor mass in the brain with local infiltration of malignant cells (Fig. 2C). Hematoxylin-Eosin (H&E) staining of the brain tissue (Fig. 2D, E) and metastases (Fig.3) further confirm the malignant state of the induced cell and its progeny.

FIGURE 2:

In parallel with tumor mass formation in the brain (Fig. 2), we observed that some cells of the progeny re-localized to new loci far from the brain, such as the heart (Fig. 3A), the digestive tract (Fig. 3B, Fig. S6D) and the trunk (Fig. S6C). We therefore deduce that the transient activation of Ventx alters the state of a cell expressing a mutated oncogene (i.e. KRASG12V) and induces its malignant transformation in vivo, with tumorigenic potential and the capacity to generate invasive progeny.

FIGURE 3:

Of the larvae in which a single cell was expressing the oncogene and in which Ventx was transiently activated (KR+VX cell), all (n=15) developed tumors within 5 dpi. The frequency of tumor development is therefore F₁=1. Conversely the probability for such a cell to not develop a tumor is F₀=0. Due to the finite size of the sample, we estimate the probability of tumor development from a single cell to be: P₁ > 80% (χ² = 3,75; df=1).

To definitely confirm the carcinogenic nature of the KR+VX induced cells we isolated and injected a single cell from a hyperplasic tissue (identified by the blue fluorescence of its nucleus) into a naïve host zebrafish larva. This led to integration, migration and colonization of the host tissues by the progeny of the transplanted cell (Fig. 4A,B), with strong pERK activity detected in tumor masses of the host (Fig. 4B). Out of 52 transplanted host zebrafish, 31 developed tumors, a probability of tumor development (60%) consistent with previously reported efficiency of tumor cells transplantation in zebrafish³⁹. This result definitely confirms the cancerous nature of cells expressing a mutated oncogene and exposed to the transient activation of Ventx, with subsequent alteration of the homeostasis. Notice that injection into a naïve larva of an oncogene expressing cell (identified by its nuclear blue fluorescence) which did not experience the transient activation of a reprogramming factor does not yield a tumor (Fig. S7B).

FIGURE 4:

Discussion

The surprising frequency of somatic mutations occurring in physiologically normal tissues^40,41 raises the question of what combinations of events are sufficient for the malignant transformation of a single cell and the rise of the cell of origin of cancer. The robust deterministic process of single-cell cancer induction that we uncovered suggests that the aberrant reactivation of Epi-Driver genes involved in reprogramming/pluripotency (such as VENTX/NANOG, POU5/OCT4) might be relevant to the irreversible malignant transformation of a cell. Reprogramming Epi-Drivers are important regulators of cell viability, survival and proliferation in several cellular contexts, from embryonic stem cells to cancer cells²²-²⁷. Due to their capacity to modulate epigenetic memory and cell plasticity, these reprogramming factors may drive the early stages of malignant transformation in vivo once (re)activated aberrantly. They likely share some mechanism(s) with processes such as induced nuclear-reprogramming²²,²³,²⁴, pluripotency maintenance and/or endogenous cell reprogramming during development³⁷. Thus, consistent with our results, the reactivation³⁷ of the Neural Crest (NC) Progenitor program (possibly via the stochastic expression of VENTX/NANOG and/or POU5/OCT4) has been shown in BRAF/p53 double mutant cells to yield NC-related tumors in vivo³⁰-³².

Since cells carrying cancer-causing mutations do not deterministically developed cancers in vivo³²,⁴⁰, we suggest that the probability of a mutated cell to enter into malignant transformation, or to maintain its physiological functions despite mutations, correlates with the probability of the reactivation of reprogramming factors. Our results support a “Vogelgram”⁴² two-step model (FIG. 5) for irreversible single-cell malignant transformation in vivo, mirroring the two hits hypothesis proposed by Berenblum and Shubik⁴³.

FIGURE 5:

Thus, the appearance of Mut-Drivers in a normal (preprocancer⁴¹) cell⁴⁰ within healthy tissue would act as a permissive but insufficient initiator for malignant cell transformation. Despite the frequency of mutational insults, which are often caused by external cues (chemical carcinogens, UV, etc…) or senescence/ageing, it is known that surveillance mechanisms like cell competition between wild-type and mutated cells safeguard tissue homeostasis and results in active elimination of mutant cells from the tissue^44,45,46. Our data imply that the aberrant reactivation of Epi-Driver genes involved in reprogramming/pluripotency may promote the irreversible and deterministic malignant transformation in the cell of origin of cancer leading to carcinogenesis.

Our hypothesis is further strengthened by the observation that the variation in cancer risk among tissues can be related to the number of stem cell divisions and tissue renewal⁶. The aberrant reactivation (or maintenance) of Epi-Driver genes in these regenerative events, in cooperation with incident mutations, might be the key deterministic factor in the switching of a normal/healthy cell into the first cell of origin of cancer.

Furthermore, the results reported here are compatible with the recently proposed “ground state theory of cancer initiation7“. According to that theory a malignant transformation may occur in a cell harboring an oncogenic mutation upon a change of its functional state (its “ground state”). Here the transient activation of a reprogramming factor (i.e. Ventx, Nanog or Oct4) possibly synergizes with the oncogene to alter the epigenetic and functional state of the cell allowing its transformation into a tumorigenic cell.

Importantly, our observation that the progeny of the cell(s) of origin of cancer display a short-range dispersal within the tissue during the earliest phases of tumorigenesis and prior to the effective appearance of tumor mass corroborates a recent theoretical model predicting such an early dispersal during carcinogenesis to explain heterogeneity, therapeutic resistance and tumor relapse⁸.

Being the first experiments to predictably control the malignant transformation of a single cell in an unaltered microenvironment our approach opens a new vista on the study of “the cell of origin of cancer”⁴⁷, an acknowledged enigma of cancer research. As such our results raise many questions: what mechanisms are at work in the transition from the preprocancer state to a tumorigenic state? Is the aberrant reactivation of reprogramming factors a key step in the initiation of carcinogenesis, independent of age, tissue or oncogenic mutation? Do drugs against reprogramming factors reduce the probability of carcinogenesis? Conversely, do some carcinogens (e.g. environmental pollutants, pesticides, endocrine disruptors etc…) act by reactivating those factors?

By allowing for specific and reproductible single cell malignant transformation in vivo, our optogenetic approach opens the way for a quantitative study of the initial stages of cancer at the single cell level (e.g. tracking and characterization), that will allow one to address many of these questions.

Materials and methods

Fish lines and maintenance

Zebrafish were raised and maintained in an approved Fish Facility (C75-05-32 at IBENS) on a 14–10 h light-dark diurnal cycle with standard culture methods⁴⁸. Embryos collected from natural crosses were staged according to Kimmel⁴⁹. The Tg(actin:loxP-EOS-stop-loxP-KRASG12V-T2A-H2B-mTFP) was generated by injecting the plasmid pT24-actin:loxP-EOS-stop-loxP-KRASG12V-T2A-H2B-mTFP, which contains the homologous cDNA sequence of KRASG12V from human, with tol2 mRNA transposase. Founder transgenic fish were identified by global expression of Eos. The Tg(ubi:Cre-ERT; myl7:EGFP) was previously described⁵⁰. The double transgenic line Tg(actin:loxP-EOS-stop-loxP-KRASG12V-T2A-H2B-mTFP; ubi:Cre-ERT; myl7:EGFP) was created by crossing Tg(actin:loxP-EOS-stop-loxP-KRASG12V-T2A-H2B-mTFP) and Tg(ubi:Cre-ERT; myl7:EGFP). Founder double transgenic fish were selected by global expression of Eos and expression of EGFP in the developing heart. Zebrafish were imaged for phenotypic analysis throughout early development, from embryonic to larval stage, and then fixed (at 6dpf) with PAXgene Tissue Container Product (Qiagen) for RT-qPCR or with 4% PFA (Thermofisher) for histological (Hematoxylin & Eosine, H&E) analyses.

Dexamethasone induction

transgenic zebrafish were injected at the one cell stage with the mRNA of a gene-construct Ventx-GR (Xenopus ventx2, 450pg), Nanog-GR (mouse Nanog, 100pg) or Pou5/Oct4 (Xenopus pou5f3.1/oct91, 100pg) together with a red fluorescent expression marker mRFP (50 pg). The protein products of these constructs are sequestered by cytoplasmic chaperones and released upon incubation of the embryos in a medium containing 10 µM Dexamethasone (DEX). In single cell activation experiments, zebrafish were selected from their strong intensity of fluorescent mRFP signal for local activation of kRasG12V via uncaging of 6µM cCYC (a gift of I.Aujard and L.Jullien).

Reverse transcriptase quantitative PCR (RT-qPCR)

Zebrafish larvae were fixed (as mentioned above) at 6 dpf. Total RNAs were extracted using the RNeasy micro kit (Qiagen) according to the manufacturer’s protocols. Sample quantity and purity, reverse transcription, pre-amplification and High throughput qPCR were performed as in Zhang et al.⁵¹. Quantitative PCR was performed using the high throughput platform BioMark™ HD System and the 48.48 GE Dynamic Arrays (Fluidigm). RTqPCR measurements were done in triplicate on pooled zebrafish larvae and single zebrafish larva (for KR+VX condition), as indicated in the text. Normalization and quantification were obtained with the Δ(ΔCt) method using rpl13a as a reference gene.

Immunohistochemistry (IHC)

The zebrafish obtained in the various conditions and the stages mentioned in the text were fixed in 4% PFA overnight at 4 °C, followed by dehydration with 100% methanol at −20 °C for more than 1 day. After gradual rehydration of methanol and wash with PBS/Tween 0.1%, the embryos were incubated in a blocking solution: 1%Triton, 1% DMSO, 1% BSA and 10% sheep serum (Sigma) in PBS on a shaker for 1 h at room temperature, followed by incubation on a shaker overnight at +4°C with 1:500 anti-phospho-Erk antibody (phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) (D13.14.4E) XP^®). Rabbit mAb detects endogenous levels of p44 and p42 MAP Kinase (Erk1 and Erk2) when dually phosphorylated at Thr202 and Tyr204 of Erk1 (Thr185 and Tyr187 of Erk2), and singly phosphorylated at Thr202 - this antibody does not cross-react with the corresponding phosphorylated residues of either JNK/SAPK or p38 MAP kinases) (ref 4370S Cell Signalling). After washing with PBS/Tween, embryos were incubated overnight at +4°C with 1/1000 secondary fluorescent conjugated antibody, Donkey anti-Rabbit Alexa Fluor 488 (Thermo Fisher A-21206). To visualize the distribution of Ventx-GR in the fish, embryos were incubated with 1:500 anti-HA antibody, tagging the HA sequence fused to Ventx-GR, as the conjugated antibody (Anti-HA, mouse IgG1, clone 16B12 Alex Fluor™ 488 conjugated, Thermofisher Ref. A21287) on a shaker overnight at 4 °C. All images were taken on a Nikon Ti microscope equipped with a Hamamatsu Orca camera, except for confocal microscopy (Fig.2C, S4B) carried out on a Leica SP5 and Zeiss LSM800 microscope.

Caged Cyclofen (cCYC) treatment and UV uncaging

To induce kRASG12V expression, 24hpf embryos were directly incubated in 6μM caged cyclofen (cCYC) for 1 hour (in the dark), briefly rinsed in E3 medium followed by 5 min. photo-activation with a ∼365 nm UV lamp. They were then transferred back to E3 medium + 10 µM DEX (to release Ventx-GR from its complex with cytoplasmic chaperones). The embryos at 1 dpi were washed 3x in E3 medium to remove DEX. For this experiment, we used a benchtop UV lamp (Fisher VL-6-L) which emits a peak wavelength at 365 nm with a FWHM (full width at half maximum) of 40 nm and delivers on the illuminated sample a typical photon flux of ∼4.3 Einstein/(s·m²).

Localized uncaging was performed by illumination for 7 min on a Nikon Ti microscope equipped with a light source peaking at 405 nm, Fig.1. The size of the uncaging region was controlled by an iris that defines a circular illumination of diameter ∼ 80 μm. After 0,5-1 h following uncaging the illuminated region of each embryo was imaged to identify and count mTFP positive cell(s). Embryos were transferred to a 12-well plate with E3 + 10 µM DEX, incubated in total darkness overnight and washed 3x with E3 at 1 dpi (to remove DEX).

Cell transplantation

Zebrafish larvae were grown until 6dpi, dissociated in dissociation medium⁵² (025% trypsin-EDTA plus 1/10 Collagenase 100mg/ml) 10min at 30°C, mechanically homogenized, resuspended in DMEM-10% Fetal Bovine Serum (FBS), filtered through 70µm nylon mesh (Corning Cell Strainer Ref. 431751) and resuspended in DMEM-10% FBS at 1cell/1nl concentration by using LUNA- II cell counter (Logos BioSystem).

Microscopy

Fluorescent images were taken on a Nikon Ti microscope equipped with a Hamamatsu ORCA V2+ camera and a 10X plan fluo objective. Filter setting of CFP: excitation at 438 ± 24 nm, emission at 483 ± 32 nm; mEosFP and Alexa 488: excitation at 497 ± 16 nm, emission at 535 ± 22 nm. Image analysis was done using ImageJ software.

Acknowledgements

We acknowledge financial support from the ITMO Cancer of Aviesan within the framework of the 2021-2030 Cancer Control Strategy, on funds administered by Inserm. The gene expression analysis was carried out on the high throughput qPCR-HD-Genomic Paris Centre core facility and was supported by grants from the Région Ile de France. We are grateful to I.Aujard and L.Jullien (ENS, Dept. of Chemistry) for a gift of caged cyclofen. We acknowledge useful comments on a draft of this paper by Z.Feng, M.Distel and D.Louvard. P.S. acknowledges useful discussions with C. Rogard, M. Smilla, L.A.M. Scerbo and E.P.A. Scerbo. We thank the Graphic Atelier of Timon Ducos for scientific illustrations (https://timonducos.com/).

Supplementary figures

FIGURE S1:

FIGURE S2:

FIGURE S3:

FIGURE S4:

FIGURE S5:

FIGURE S6:

FIGURE S7:

FIGURE S8: