Compromising asymmetric stem cell division in Drosophila central brain: Revisiting the connections with tumorigenesis

ABSTRACT

Asymmetric cell division (ACD) is an essential process during development for generating cell diversity. In addition, a more recent connection between ACD, cancer and stem cell biology has opened novel and highly intriguing venues in the field. This connection between compromised ACD and tumorigenesis was first demonstrated using Drosophila neural stem cells (neuroblasts, NBs) more than a decade ago and, over the past years, it has also been established in vertebrate stem cells. Here, focusing on Drosophila larval brain NBs, and in light of results recently obtained in our lab, we revisit this connection emphasizing two main aspects: 1) the differences in tumor suppressor activity of different ACD regulators and 2) the potential relevance of environment and temporal window frame for compromised ACD-dependent induction of tumor-like overgrowth.

Introduction

Asymmetric cell division (ACD) is a key process to generate cell diversity from bacteria to mammals. Intriguingly, over the past years, it has become established a direct link between defects in stem ACD and tumorigenic processes. This connection was first shown using Drosophila larval brain neural stem cells, called neuroblasts (NBs), as a model system [1]. In this work, pieces of larval brains mutant for particular ACD regulators were transplanted into the abdomen of wild-type (wt) hosts generating after weeks massive tumors that killed the host. Pioneering studies by Elisabeth Gateff and colleagues already used this transplantation technique, along with studies in situ in the larval brain, to demonstrate the tumor suppressor activity of several Drosophila genes, such as discs large1 (dlg1), lethal (2) giant larvae (l(2)gl) and brain tumor (brat) among others [2–5]. Interestingly, some of these genes including dlg1, l(2)gl and brat were shown a posteriori to regulate the process of ACD, further supporting the connection between compromised ACD and tumorigenesis [6–12]. In fact, brat NB mutant clones in the larval brain displayed strong ACD failures that led to massive clones showing tumor-like overgrowth [7–9,12]. Since then, mutations in other ACD regulators causing similar effects have been described [13–17]. However, through all these years different conditions or experimental setups have been used to analyze the effect of compromising ACD on (tumor-like) overgrowth. Here we revisit this aspect highlighting the potential relevance of the environment, context-dependent signals and the type of ACD regulator for the ACD-associated tumor suppressor activity.

The (main) players

The neural stem cells of Drosophila central nervous system (CNS), the NBs, are one of the best paradigms for studying the process of ACD [18, 19]. NBs divide to give rise to another NB and a committed ganglion mother cell (GMC), which will terminally divide to generate two neurons or glial cells. The fate of a GMC relies on cell-fate determinants that asymmetrically distribute at the basal pole of the NB and segregate into the GMC, activating in this daughter cell a differentiation program (Fig. 1A). The main cell-fate determinants described in Drosophila include the transcription factor Prospero (Pros), the translational repressor Brat and the Notch signaling inhibitor Numb [7–9,12,20-24]. The basal sorting of these determinants in the NB depends in turn on the activity of a complex protein network located at the NB apical pole (Fig. 1B). This apical complex includes Cdc42, the conserved partitioning defective proteins Par-6 and Par-3 (Bazooka, Baz in Drosophila) and the atypical protein kinase C (aPKC) [25–29]. L(2)gl forms initially part of this complex instead of Baz; a cascade of events, starting with Par-6 phosphorylation by Aurora A (AurA) and consequent aPKC activation and ending up with L(2)gl phosphorylation by aPKC, leads to the replacement of L(2)gl by Baz in the apical complex [30]. Baz directly interacts with the adaptor protein Inscuteable (Insc), which in turn binds Partner of Inscuteable (Pins; LGN in mammals); this allows the association of Pins with the membrane anchored G protein subunit Gα and the beginning of a proper spindle orientation orchestration [31-38]. For that, the actin-binding protein Canoe (Cno; afadin in mammals) phosphorylated by the kinase Warts binds Pins, displacing Insc, and contributes to recruit the Pins binding proteins Dlg1 and Mushroom body defect (Mud; NuMA in mammals) to the apical domain [39,40]. Dlg1 associates with the Pins^LINKER middle domain and with the Kinesin heavy chain 73 (Khc-73); Mud binds the Pins^TPRs domain, displacing Cno that like Insc binds the same Pins region, and the Dynein-Dynactin complex [35,41–45]. Both Khc-73 and the Dynein-Dynactin complex interact with astral microtubules reinforcing each other and ensuring the proper orientation of the mitotic spindle. Some cell cycle regulators, such as the kinases AurA and Polo also contribute to modulate the ACD process [15–17,46]. For example, AurA apart of phosphorylating Par-6 as mentioned above, event that ultimately leads to a correct Numb localization [30], it also phosphorylates the Pins^LINKER domain, modification that is necessary for Dlg1 binding [43]. Polo phosphorylates the adaptor protein Partner of Numb (Pon), phosphorylation that is essential for Pon to asymmetrically localize the cell-fate determinant Numb [17]. Another adaptor protein called Miranda (Mira) is required for facilitating the localization of the cell-fate determinants Pros and Brat at the basal pole of the NB [47–51].

Figure 1.

The neural stem cells or NBs of the Drosophila CNS divide asymmetrically. (A) Once an axis of cell polarity is established in the NB, the mitotic spindle orientates along it and cell-fate determinants asymmetrically localize at the basal pole of the NB. An intricate complex of proteins apically located at metaphase NBs regulates these latter events. When the NB divides only the most basal cell, the GMC, receives the cell-fate determinants, which initiate a program of differentiation in this daughter cell. The GMC divides to give rise through a terminal and asymmetric division to two neurons (or glial cells). The other daughter cell does not inherit cell-fate determinants and keeps on dividing asymmetrically like the mother NB. A: apical; B: basal; the two small yellow dots in the NB represent the centrosomes. (B) A simplified and “static” view of main apical complex components and cell-fate determinants in dividing NBs. Note, for example, that all Insc, Cno and Mud bind the same region of Pins (Pins^TPRs domain) and cannot be simultaneously interacting with Pins (see text). The apical complex is linked with the mitotic spindle through the motor proteins Khc-73 and the Dynein-Dynactin complex, both of which physically interact with astral microtubules (see also main text).

The NBs

The function of most of these factors in ACD has been traditionally studied using the embryonic NBs as a model system. These NBs undergo a limited, up to 20, number of asymmetric divisions and enter quiescence at the end of embryogenesis. At larval stages (late first instar larvae) NBs resume proliferation; unlike embryonic NBs, larval NBs increase in size after each division and can undergo hundreds of them. Two main types of larval NBs have been characterized, type I and type II (Fig. 2). Type I NBs occupy most of the central brain and are similar to embryonic NBs in that they divide to generate an NB and a GMC that will terminally divide to give rise to two neurons or glial cells [52]. Type II NBs were more recently characterized [9,53,54]. They are only eight per brain hemisphere and are specifically located in the dorso-medial part of the brain. Unlike type I NBs, type II NBs give rise to an NB and an intermediate progenitor (INP) instead of a GMC. The INP then, after a maturation process, divides generating another INP and a GMC. In these type II NB lineages, as in type I ones, cell-fate determinants are going to prevent self-renewal and to promote differentiation in the daughter cell destined to become an INP or a GMC. Over the past years, some of the factors and mechanisms that particularly operate in type II NB lineages to allow the specification of an INP have been described. For example, in type II NBs, the cell-fate determinants Brat and Numb inhibit the self-renewal transcriptional repressors Deadpan (Dpn), Klumpfuss (Klu) and Enhancer of split mγ (E(spl)mγ) in the INP in which they are asymmetrically segregated. The repression of dpn and klu mRNA translation by Brat then leads to the activation of an immature INP identity factor called Earmuff (Erm) [55–58]. In the NB, Notch is active and contributes to suppress this factor [59]. Hence, the lack of Brat or the Notch inhibitor Numb in the INP (i.e. in brat or numb mutant clones) leads to Dpn and Klu activation and erm repression; consequently, the prospective INP adopt an NB identity [56,59–61]. Otherwise, the correct asymmetric segregation of Brat and Numb allows the proper maturation of the INP and its subsequent divisions. Given this additional phase of proliferation, type II NB lineages are larger and more prone to show tumor-like overgrowth when ACD is compromised [9].

Figure 2.

Two types of NB lineages coexist in the Drosophila larval central brain. (A) Diagram of the Drosophila third instar larval CNS in a dorsal view. OL: optic lobe; CB: central brain; VNC: ventral nerve cord; A: anterior; P: posterior. Type I NBs (orange) occupy most of the CB; type II NBs (purple) are only eight per brain hemisphere and are restricted to the dorso-medial part of the CB. (B) Type II NB lineages are larger than type I NB lineages due to an intermediate phase of proliferation carried out by intermediate progenitors (INPs; in red); iINP: immature INP; mINP: mature INP.

They are not all the same

The initial demonstration and subsequent studies regarding the tumor suppressor activity of ACD regulators were performed using larval NBs as a model system. Some of these studies were carried out before type II NBs were described. Thus, the analysis of ACD regulator mutations in type I NBs, or even comparisons between control and mutant condition without taking into account the type of NB lineage could have been misleading. Even in these “uncontrolled” situations we can find clear cases in which the loss of particular ACD modulators does not lead to tumor-like overgrowth. For example, in aPKC mutant clones, analyzed in mushroom body NB lineages (type I NB-like), the lack of aPKC results in premature NB cell cycle exit or arrest and consequently in reduced NB lineages [62]. It would be expected a similar effect in type II NB lineages. Also, the clonal analysis of aurA and polo mutant NB lineages (it was not specified whether type I or type II NBs, as the latter had not yet been described) revealed the presence of extra NBs but not clone overgrowth [16,17]. Other apical components, such as Cdc42, Par-6, Baz (Par-3), Insc, Pins or Mud have not been subjected to NB mutant clonal analysis at all.

Regarding those analyses performed in clearly identified type II NBs, these are also already showing that not all ACD regulators have similar effects on overgrowth when they are compromised (Table 1). For example, we recently described the effect of cno, l(2)gl, dlg1 and scribble (scrib) loss in type II NB lineages by clonal analysis [63]. All single mutant NB clones contain extra NBs but none of them show overgrowth; on the contrary, mutant clones tend to be smaller than wt clones and scrib mutant clones are eliminated by apoptosis. A completely different phenotype is observed in cell-fate determinant NB mutant clones, particularly in brat [7-9] and pros [7-9,12] mutant clones, in which tumor-like overgrowth is very apparent. These clonal analyses were first carried out in NB lineages all over the brain (type II NB had not yet been described) [7,8,12], and the results were further confirmed later in a study that revealed type II NB lineages as the ones responsible of giving rise those big tumoral masses when brat or pros were mutated [9]. In fact, these studies were among the first in demonstrating the connection between ACD defects and tumorigenesis, as mentioned before. Just a consideration about the latter type II NB pros mutant clones. Pros, unlike Brat, is not present in type II NBs; it is cytoplasmic in the INPs, as well as in type I NBs, and nuclear in GMCs and neurons in both types of NB lineages. Thus, the extra “progenitor cells” described within those tumoral masses in type II NBs pros mutant clones are INPs and not NBs, or more accurately “tumor-like NBs”, as it happens in type II NB brat mutant clones (see also Table 1).

Table 1.

ACD regulators and their tumor suppressor activity.

ACD regulator	Homozygous mutant NB clone analysis (in a wt* background)	References	Homozygous mutant brain analysis	References	Allograft culture	References
aPKC	Reduced NB lineages: NB stops dividing (in MB NBs)	#[62]	Reduced NB lineages: NB stops dividing (die at second instar)	#[62]	No overgrowth	#[1]
aPKCΔN(GOF, constitutively active aPKC)	Mild increase in NBs (with insc-Gal4)	#[8]			ND
	Mild increase in NBs (with worniu-Gal4)	#[66]
aPKCCAAXWT (GOF, aPKC targeted to the plasma membrane)	Large increase in NBs (with worniu-Gal4)	#[66]
L(2)gl	No clone overgrowth (extra NBs; analysis in type II NB clones)	#[63]	Brain overgrowth (affects larval optic NBs and GMCs)	#[2,3,5]	Invasive, lethal, metastatic growth (in wt adult hosts)	#[2-5]
			Brain overgrowth (affects the dorsal part of the brain lobes only)	#[8]	No overgrowth, abnormal differentiation (in wt larval hosts)	#[5]
			Extra NBs (within individual NB lineages positively labeled, and in general, in the brain)	#[66]
				#[69]
Cno	No clone overgrowth (occasional extra NBs; analysis in type II NB clones)	#[63]	ND (lethal) No clone overgrowth in cnonull/cnohypomorph (viable)	Bañón and Carmena, (unpublished results)	ND
Pins	ND		Decreased number of NBs (0–1 NB within individual NB lineages positively labeled and, in general, in the CB)	#[66]	Invasive, lethal, metastatic growth	#[1]
Mud	ND		Increased number of MB and CB NBs	#[42]	ND
Dlg1	No clone overgrowth (extra NBs; analysis in type II NB clones)	#[63]	General brain overgrowth	#[2,3]	Invasive, lethal, metastatic growth	#[2,3]
Scrib	No clone overgrowth (extra NBs; apoptotic clones; analysis in type II NB clones)	#[63]	General brain overgrowth	#[68]	ND
Mira	Clone overgrowth? (extra Cyclin E+ cells in dorsal brain (miraRNAi with insc-Gal4)	#[8]	ND (lethal)		Invasive, lethal, metastatic growth	#[1]
Brat	Clone overgrowth (clone analysis all over the brain)	#[7,8]	General overgrowth (affects adult optic NBs and GMCs)	#[3]	Invasive, lethal, metastatic growth	#[2,3]
	Clone overgrowth (analysis in type II NBs: clones filled with NBs; no differentiated progeny)	#[9]	Increase in CB NBs	#[12]
			Overgrowth (affects posterior part of the brain lobes only)	#[8]
Pros	Clone overgrowth (clone analysis all over the brain: clones filled with “NB-like cells” and smaller cells)	#[7,8,12]	ND (lethal)		Invasive, lethal, metastatic growth	#[1]
	Clone overgrowth (analysis in type II NBs: clones filled with many INPs; no or few differentiated progeny)	#[9]
Numb	Mild clone overgrowth (extra NBs; clone analysis all over the brain) Clone overgrowth	#[16]	ND (lethal)		Invasive, lethal, metastatic growth	#[1]
	Clone overgrowth (analysis in type II NBs: clones filled with NBs; no differentiated progeny	#[15]
		#[9]
Polo	No clone overgrowth (extra NBs)	#[17]	General overgrowth	#[17]	Invasive, lethal, metastatic growth	#[70]
AurA	No clone overgrowth (extra NBs)	#[16]	General overgrowth	#[15,16]	Invasive, lethal, metastatic growth	#[70]

^*Heterozygous; MB: mushroom body; CB: central brain; ND: not determined

Numb is another main cell-fate determinant whose impairment in NB clones leads to clone overgrowth [9,15,16] (Table 1), though apparently with a milder effect than brat and pros mutants [16]; in this latter study all NB clones were considered and not specifically type II NBs (again, not yet described), this fact perhaps influencing the milder phenotype described for numb mutant clones.

Overall, looking at all ACD regulators whose loss has been analyzed in larval NB clones, we could conclude that only cell-fate determinant impairment leads to tumor-like overgrowth; other ACD modulator loss causes as much as some extra NBs. This fact is not so surprising in light of well-known “hallmarks” in cancer biology [64]. Specifically, it has been established that not one but several mutational hits are required to promote tumorigenesis by inducing cooperative changes that facilitate tumor progression. In the context of ACD, it is also well established for a long time ago the cooperation between different ACD regulators, so that only the loss of two or more of these modulators strongly impairs this highly redundant process [65] and can synergistically induce NB overproliferation in the larval brain [14,66,67]. In our recent study we show that the simultaneous loss of two ACD regulators cno and scrib in larval NB clones has an additional consequence, the upregulation of the Ras signaling pathway, normally repressed by Cno. These three cooperative changes (mutation of two ACD regulators plus the upregulation of Ras, which in turn helps to evade apoptosis in the mutant clones) entail a much drastic, tumor-like overgrowth phenotype in NB larval mutant clones [63], in line with the increasing number of mutational changes required for tumor formation.

Then, what about the strong phenotypes shown by single cell-fate determinant mutant NB clones? These determinants are key cell identity regulators that directly impinge on multiple targets by controlling gene transcription (Pros) or translation (Brat). Hence, these would be particular single mutations that imply numerous and simultaneous failures mimicking several mutational hits, as mentioned above [64]. It is relevant to have this into account to evaluate the consequences of particular ACD regulator mutation. The presence of extra NBs in the mutant NB clones analyzed could be already indicative of the tumor suppressor capacity of the gene under study and a sensitized genetic background, in which further mutational hits would lead to tumor-like overgrowth.

Look what surrounds you

Another important aspect when considering the tumor suppressor activity of an ACD regulator could be the environment that surrounds the mutant cells. In principle, clonal mutant analysis in NB lineages would be an excellent model system to explore the proliferative potential of a given ACD mutation. In this situation, like in a tumor formation, group of mutant cells abnormally divide in an otherwise wt (heterozygote) tissue background. Likewise, mutant tissue transplantation into normal tissue is and has been for a long time a classic set up for testing growth capacity not only in Drosophila but also in mammals (Fig. 3A, C). However, reviewing the literature and in light of some recent results from our lab, we realized that both systems do not render the same results in most cases (Table 1). For example, while both l(2)gl and dlg1 mutations were described to induce invasive, lethal and metastatic growth after allograft culture in adult hosts [2-5], they do not induce tumor-like overgrowth in situ, in mutant larval type II NB clones [63]; both mutant clones show extra NBs (up to seven in the case of l(2)gl mutant clones) but they do not overgrow, and, at least l(2)gl mutant clones, are even smaller than wt clones. It has been proposed that the L(2)gl protein might be very stable and this could explain the persistence of some protein in NBs after the clonal removal of l(2)gl and the consequent milder phenotype [8]. However, at least in the type II NB l(2)gl (and dlg1) mutant clones we induced, we were unable to detect any L(2)gl (Dlg1) protein by immunofluorescence [63]. Similar results (lack of tumor-like overgrowth) are observed in the case of aurA and polo NB mutant clones (Table 1). In fact, only cell-fate determinant mutants seem to show tumor-like overgrowth both in allograft cultures and in situ in larval mutant NB clones (Table 1). Why these differences in growth potential? Given that in both experimental systems, mutant cells are surrounded by normal tissue, one would expect that signals associated to the different temporal window frame (larva versus adult) are responsible of such differences. In fact, l(2)gl mutant cells transplanted in wt larval hosts stop growing [5]. This has been explained by the mutant tissue response to the hormones secreted during metamorphosis that normally induce cessation of growth at this stage. However, something else must be restraining growth of the larval NB mutant clones mentioned above (see also Table 1), as lack of clone overgrowth is already observed before metamorphosis starts.

Figure 3.

Different experimental conditions to analyze the tumor suppressor capability of ACD regulators. (A) Analysis of ACD mutations in NB clones (type II NB mutant clones are labeled in green in the diagram) in an otherwise wt larval background. (B) Analysis of ACD mutations in homozygous mutant brains/larvae (labeled in green in the diagram). (C) Analysis of ACD mutations in allograft cultures. Mutant brain tissue GFP-labeled is injected in the abdomen of wt adult hosts.

In addition, homozygous mutant analysis in the larval brain (Fig. 3B) shows in most of the cases studied very different results to that shown by analysis in mutant clones (Table 1). For example, all dlg1, l(2)gl, aurA and polo, as well as scrib homozygous mutant brains display tumor-like overgrowth [2,3,5,8,15–17,66,68]. However, none of these mutants show overgrowth in NB clones, as described above for dlg1, l(2)gl, aurA and polo; and, in the case of scrib mutant NB clones, they are even significantly smaller than wt clones and finally removed by apoptosis [63]. Thus, the phenotype of homozygous mutants is more similar to that observed after allograft culture transplants. From those ACD gene regulators analyzed in both homozygous mutant larvae and in allograft cultures, there is only one reported case in which both phenotypes are not coincident. This is the case of pins, which in homozygosis does not show overgrowth but decreased number of NBs, and in allograft transplants renders invasive and metastatic growth [1,66] (Table 1). This is intriguing, but it might be that the potential “antigrowth signal” that seems to be eliminated in other ACD homozygous mutant condition is not overcome in pins homozygotes. However, it must be pointed out that at least some of the ACD mutant homozygote analyses, such as those involving l(2)gl, dlg1 or scrib did not describe whether/how central brain NB lineages are affected. In fact, the tumor-like overgrowth observed in these studies (those involving l(2)gl and dlg1) was basically attributed to an expansion of ectoderm-derived tissues as optic lobe NBs and GMCs (derived from the optic lobe neuroepithelium), as well as to a massive growth and fusion of epithelial imaginal discs [2,3,5]. In addition, some data seems to indicate that larval brain NB lineages in l(2)gl homozygotes show extra NBs but not overgrowth [66,69], similar phenotype to that we have recently found in l(2)gl mutant NB clones (in a wt/heterozygous background) [63]. In this line, NB lineages in cno homozygotes (cno^hypomorph/ cno^hypomorph and in cno^hypomorph/ cno^null) do not overgrow (Bañón and Carmena, unpublished results). This may suggest that independently of the NB clone surrounding cells (i.e. wt or mutant), NB clones mutant for most ACD gene regulators do not show massive overgrowth even though they do show frequently extra NBs. This would be in line with the high redundancy of the ACD process mentioned before. A more detailed analysis of individual NB lineages, especially of type II NBs, in a homozygous mutant background for the different ACD regulators (that can survive until third larval instar) would be required to clarify that.

Then, the differences in promoting overgrowth of some ACD regulators in larval NB mutant clones (in a wt/heterozygous background) versus allograft cultures (in wt adult hosts) could merely depend on some larval stage-associated signal that tend to restrain the overgrowth of mutant NB lineages independently of the mutant or wt genetic background that surround the clones. Only when in the NB clone the process of ACD is strongly compromised beside pro-growth signals are favored (as we have recently seen in cno scrib NB double mutant clones, or as it happens in cell-fate determinant mutant NB clones), a massive overgrowth is promoted in situ overcoming the larval stage-associated inhibitory signal. In allograft cultures, the lack of that signal would favor a tumor-like overgrowth, even in single mutant conditions, such as in l(2)gl, or dlg1 implants. In these latter cases, we cannot discard though that other mutant tissues of the homozygote implants are mainly contributing to the tumor-like overgrowth described [2–5]. It would be interesting to analyze allograft cultures using implants specifically carrying NB II single mutant clones instead of pieces of homozygote brain tissue for mutants such as l(2)gl or dlg1, as well as cno and scrib. This would indicate whether in this experimental setup these ACD single mutants are already able to induce tumor-like overgrowth.

Conclusions

Studies carried out over the past decade have uncovered the tumor suppressor capability of ACD regulators. While in NB clones most ACD regulator mutants, with the exception of cell-fate determinants, do not induce tumor-like overgrowth, they frequently cause the formation of extra NBs. Both, the high redundancy of the ACD process plus some larval-associated signals probably collaborate to restrict overgrowth avoiding stronger phenotypes in situ, in larval NB clones. Only when the ACD process is strongly compromised, as it happens in cell-fate determinant larval NB mutant clones in situ, or the larval signals are eliminated, as it occurs on allograft cultures, massive tumoral masses are induced. Both aspects, ACD redundancy and environment influence, must be taking into account for correctly interpreting the tumorigenic potential of ACD regulators.

Disclosure of potential conflicts of interest

No potential conflicts of interest are disclosed.

Funding

Our research was supported by Spanish Government (Ministerio de Economía y Competitividad, MINECO) grants BFU2012-33020, BFU2015-64251-P and by FEDER (European Regional Development Fund). The Instituto de Neurociencias in Alicante is a Center of Excellence Severo Ochoa.