We all seem to carry at least one mole on our body, those black spots or nevi that may become larger as we grow in childhood but then stay put in size for the rest of our lives, except… Yes, except if they start growing again, which may give rise to melanoma. Curiously, over 80% of these moles are mutant in the BRAF gene, usually carrying the BRAF^V600E allele, the same that serves as a driver in melanoma. Yet, nevi are normally growth‐arrested. And that is the problem that Rolando Ruiz‐Vega et al. address in one of the most inspiring and most elegantly presented papers that have recently come across our screens.

Ever since the discovery that oncogene expression presents a cellular stress that results in cellular senescence (oncogene‐induced senescence or OIS) (Serrano et al., 1997), it became generally accepted knowledge that nevi are growth‐arrested exactly for this reason. In vitro tests have indeed shown hallmarks of senescence, such as enhanced beta‐galactosidase expression, the senescence‐associated secretory phenotype, chromatin alterations, and metabolic changes [for a review, see (Pawlikowski et al., 2013)], but what constitutes a senescent nevus cell in vivo and distinguishes it from any other growth‐arrested cell was always less than clear. Recent advances in technologies for profiling gene expression in isolated individual cells now make it feasible to address this issue. That's what Ruiz‐Vega et al. did, using as tissue source the excellent mouse model produced by Bosenberg, McMahon and collaborators (Dankort et al., 2009) in which multiple nevi can be induced by tamoxifen/Tyr‐CreER‐mediated introduction of the Braf^V600E mutation at the endogenous Braf locus. If you now turn up your nose at another one of those mouse studies with little relevance to humans, you may want to reconsider.

Ruiz‐Vega et al.'s analysis of gene expression profiles of cells isolated from growth‐arrested nevi and their surrounding skin revealed a number of different cell populations: pigment cells of various subgroups, two of which were unique to induced transgenic skin and likely represented typical nevus cells, along with other usual suspects, such as fibroblasts and keratinocytes. Then, the authors tested for various classical senescence signatures in the growth‐arrested nevus cells by comparison with other cell types, including non‐nevus melanocytes. Surprisingly, while fibroblasts seemed to best fit one of these senescence signatures, nevus cells did not fit any, even though they had stopped growing. By comparison, BRAF^V600E‐expressing hair follicle melanocytes continued to express proliferation‐associated genes, clearly distinguishing them from BRAF^V600E‐expressing nevus cells.

This may have been less of a surprise to the authors, who are experts in senescence‐independent tissue growth control, than it may be to readers who have come to accept the notion that OIS is what operates in growth arrest of nevi. But to the authors, it was just the beginning of their study, because the question remained, how do nevus cells stop growing? And here, the authors showed great creativity.

First, they looked at the sizes of the nevi as they grow and stop growing. A peculiarity of nevi in both mice and humans is that they contain “nests” of nevus cells, and the sizes and cell numbers of these nests can be assessed by various morphometric means. After painting 4‐OH‐tamoxifen on the largely pigment‐free glabrous skin of the forepaws of mice early after birth, the nevus nests grew to a volume of about 40,000 μm³ before they stopped growing after 2–3 weeks, yielding an estimate of close to 700 cells/nevus on average at this time point. Ruiz‐Vega et al. then checked various theoretical models that would account for a clonal growth matching this average size. The first one assumes that there is simply a probabilistic switch: After oncogene induction, cells grow and then arrest over time with a fixed probability. This is a mathematical problem that is not too difficult to solve, but we can spare us the details here as they are well explained in the paper. Suffice it to say that with the best of assumptions, this model leads to nevus sizes that are not matched by those actually seen in vivo. One may argue, of course, that oncogene expression as a one‐hit phenomenon does not take into account oncogene activity and susceptibility of its targets, both of which may depend on numerous additional steps, and so one may dispute the validity of this simple one‐hit simulation‐based argument. Much to the reader's delight, the authors then consider precisely this problem and continue to evaluate simulations based on a multi‐step growth arrest. They conclude that generation of nevi of the observed sizes in mice would require on the order of 6–7 steps or even more for the larger sizes of human nevi. In other words, cells would have to have some intrinsic clock mechanism capable of counting time or number of divisions before arresting growth, or perhaps gauging the accumulation of random events, provided that each of these would occur with tightly controlled probabilities. There are indeed instances where cells count proliferation‐associated events in a cell‐autonomous fashion, for instance by registering telomere shortening or by continuously distributing cellular components unevenly to mother and daughter cells, thereby regulating the mother cell's number of divisions (Manzano‐Lopez & Monje‐Casas, 2020). But none of these types of counting mechanisms seem to operate in nevus cells. And so, the authors propose a different, non‐cell‐autonomous mechanism that explains the observed nevus arrest times and sizes much more precisely, namely a (likely) paracrine effect that would emanate from randomly growth‐arrested cells and would increase the arrest probability of neighboring cells as a function of the number of already arrested cells in their surroundings. This scenario would predict that nevi in close spatial proximity might inhibit each other's growth. Although the authors could not find sufficient numbers of close‐enough nevi to experimentally verify the distance over which such effects may occur, they could do this for the nevus cell nests within nevi and came up with an estimate of this distance to be <150 μm. This would seem to accommodate the sizes of arrested nevi in mice and also argue for the validity of the mouse model for the similar nest distributions seen in human nevi.

In a conceptual Figure (Figure 6 in the paper, partially reproduced here as Figure 1), the authors contrast the OIS‐based growth arrest (Figure 1a) with their new model of feedback control that does not depend on senescence (Figure 1b). They acknowledge that some “hybrid” model could also be conceived in which oncogene‐induced senescent cells would, by way of their senescence‐associated secretory program, act to inhibit nevus cell growth (Figure 6b in their paper). Because single‐cell gene expression data did not show evidence for such a secretory program to be operating in nevus cells, they favor in the end the non‐cell‐autonomous model (Figure 1b) in which growth arrest is not a direct consequence of oncogene action but rather of growth itself.

Elegant, this feedback model, isn't it? Not only does it accommodate the absence of a senescence signature in nevus cells but also the fact that growth‐arrested nevi can reinitiate growth under certain circumstances and give rise to melanomas. Moreover, it places no constraints on the molecular nature of the paracrine factors, as long as they accumulate with increasing numbers of arrested cells and their diffusion/activity range is compatible with the observed distance of action. Such factors may include growth‐inhibitory members of the TGF‐ß superfamily such as activin and GDF11. The model also has the appeal of wide applicability, may guide our general thinking about how oncogenes act and generate malignancies, and renew our focus on extracellular growth inhibitors for cancer therapy.

Nevertheless, there are at least two conceptual questions we think are still in need of an answer. First, what are the suggested mechanisms leading to growth arrest of the first nevus cell or cells in a clone? The authors favor the view that initial growth arrest need not be oncogene‐related at all and that growing nevus cells might just have an intrinsic probability to stop growing, as seen in other models of tissue growth control. Yet, a formal proof for this is currently not available. Second, and related to the above problem, while the favored model of continuing growth arrest is clearly cell non‐autonomous, as presented it still is nevus‐autonomous. Given the well‐known importance of the tissue microenvironment, however, nevus growth arrest may well be initiated by non‐nevus cells, such as fibroblasts, which conceivably may have become growth‐arrested by senescence mechanisms. Even so, this should not distract from the fact, so convincingly shown in this paper, that nevus cells themselves are definitely not senescent. Hence, we have to start letting go of the notion that nevi are growth‐arrested by cell‐autonomous oncogene‐induced senescence.

我们所有人似乎都在我们的身体上携带至少一颗痣，这些黑点或痣可能会随着我们童年的成长而变大，但在我们的余生中保持原样，除非……是的，除非它们再次开始增长，这可能会引起黑色素瘤。奇怪的是，这些痣中有80％以上是BRAF基因的突变体，通常携带BRAF ^V600E等位基因，而后者是黑色素瘤的驱动因子。但是，痣通常被禁止增长。这就是Rolando Ruiz-Vega等人的问题。在最近出现在我们屏幕上的最具启发性和最精美呈现的论文之一中发表演讲。

自从发现致癌基因表达会引起细胞应激导致细胞衰老（致癌基因诱导的衰老或OIS）（Serrano等，  1997）以来，人们就普遍认识到，痣正是由于这个原因而被捕。体外试验确实显示出衰老的特征，例如增强的β-半乳糖苷酶表达，衰老相关的分泌表型，染色质改变和代谢变化[有关综述，请参阅（Pawlikowski等，  2013）]，但体内衰老痣细胞的组成以及与其他任何生长停滞细胞的区别始终不清楚。现在，用于在分离的单个细胞中分析基因表达的技术的最新进展使解决这个问题变得可行。这就是Ruiz-Vega等人的观点。那样，使用如组织来源由Bosenberg，McMahon和合作者所产生的优异的小鼠模型（Dankort等人，  2009），其中多个痣可以通过他莫昔芬/酪氨酸- CreER介导引入的诱导型Braf ^V600E突变在所述内源型Braf轨迹。如果您现在对另一项与人类无关的小鼠研究抬起鼻子，则可能需要重新考虑。

Ruiz-Vega等人对从生长停滞的痣及其周围皮肤中分离的细胞的基因表达谱进行了分析，发现了许多不同的细胞群：各个亚组的色素细胞，其中两个是诱导性转基因皮肤所特有的，可能代表典型的痣细胞，以及其他常见的疑似细胞，例如成纤维细胞和角质形成细胞。然后，作者通过与包括非痣黑素细胞在内的其他细胞类型进行比较，测试了生长停滞的痣细胞中各种经典的衰老特征。出人意料的是，尽管成纤维细胞似乎最适合这些衰老特征之一，但痣细胞即使已经停止生长也没有适合任何衰老特征。相比之下，BRAF ^V600E表达毛囊的黑素细胞继续表达与增殖相关的基因，从而将它们与表达BRAF ^V600E的痣细胞区分开。

对于那些独立于衰老的组织生长控制专家的作者来说，这可能不足为奇，而对于那些已经接受了OIS是在痣的生长停滞中起作用的观念的读者来说，这可能并不那么令人惊讶。但是对于作者来说，这只是他们研究的开始，因为问题仍然存在，痣细胞如何停止生长？在这里，作者表现出了极大的创造力。

首先，他们观察了痣的大小，随着它们的生长和停止生长。在小鼠和人类中，痣的独特之处在于它们含有“巢”的痣细胞，并且这些巢的大小和细胞数量可以通过各种形态计量学方法进行评估。在出生后早期小鼠的前爪的主要不含颜料的光滑皮肤后涂抹4-OH他莫昔芬，痣巢增长到约40000微米的音量³在2到3周后它们停止生长之前，在这个时间点平均估计接近700个细胞/痣。Ruiz-Vega等。然后检查了各种理论模型，这些模型将说明克隆增长与该平均大小相匹配。第一个假设仅存在一个概率转换：致癌基因诱导后，细胞生长，然后以固定的概率随时间停滞。这是一个很难解决的数学问题，但是我们可以在这里为我们省去细节，因为它们在本文中得到了很好的解释。可以说，最好的假设是，该模型导致痣的大小与体内实际看到的大小不符。当然，有人可能会说，癌基因表达是一种一击现象，没有考虑到癌基因的活性。目标的敏感性和敏感性，这两个步骤都可能取决于许多其他步骤，因此有人可能会质疑这种简单的一击式基于仿真的论证的有效性。令读者感到非常高兴的是，作者随后精确地考虑了这个问题，并继续基于多步增长停滞评估模拟。他们得出结论，要在小鼠中观察到大小的痣，要产生更大尺寸的人类痣，将需要大约6-7个步骤，甚至更多。换句话说，单元格必须具有某种固有的时钟机制，能够在阻止增长之前，或者在限制随机事件的积累之前，对分裂的时间或数量进行计数，前提是这些事件中的每一个都将以严格控制的概率发生。实际上，在某些情况下，细胞会以细胞自主方式计算与增殖相关的事件， 2020年）。但是，这些类型的计数机制似乎都不在痣细胞中起作用。因此，作者提出了一种不同的，非细胞自主的机制，可以更精确地解释观察到的痣的停滞时间和大小，即一种（可能的）旁分泌效应，这种效应来自于随机生长停滞的细胞，并会增加停滞的可能性。周围细胞的数量随周围环境中已被捕获的细胞数量的变化而变化。这种情况将预测在空间上靠近的痣可能会抑制彼此的生长。尽管作者找不到足够数量的近距离痣来通过实验验证可能发生这种作用的距离，但他们可以对痣内的痣细胞巢进行此操作，并得出此距离的估计值小于150μm 。

在概念图（本文中的图6，在此处部分复制为图1）中，作者将基于OIS的生长停滞（图1a）与他们的不依赖衰老的反馈控制新模型（图1b）进行了对比。他们承认，还可以设想一些“混合”模型，其中癌基因诱导的衰老细胞将通过其衰老相关的分泌程序来抑制痣细胞的生长（其论文中的图6b ）。由于单细胞基因表达数据没有显示出这种分泌程序在痣细胞中起作用的证据，因此最终他们倾向于非细胞自主模型（图1b），在该模型中生长停滞不是癌基因的直接后果。行动，而不是成长本身。

优雅，这个反馈模型，不是吗？它不仅适应了痣细胞中不存在衰老的特征，还阻止了生长停滞的痣可以在某些情况下重新开始生长并引起黑色素瘤的事实。此外，它对旁分泌因子的分子性质没有任何限制，只要它们随着停滞细胞数量的增加而积累并且它们的扩散/活性范围与观察到的作用距离相适应即可。这些因素可能包括TGF-ß超家族的生长抑制成员，例如激活素和GDF11。该模型还具有广泛的应用吸引力，可能会指导我们对癌基因如何发挥作用并产生恶性肿瘤的一般思考，并重新使我们更加关注用于癌症治疗的细胞外生长抑制剂。

然而，我们认为至少有两个概念性问题仍需要答案。首先，建议的机制是什么导致克隆中第一个痣细胞生长停止？作者赞成这样一种观点，即最初的生长停滞根本不需要与癌基因相关，并且正在生长的痣细胞可能只是具有停止生长的内在可能性，如在其他组织生长控制模型中所见。但是，目前尚无正式证据。其次，与上述问题有关，尽管持续增长停滞的偏爱模型显然是细胞非自主的，但如前所述，它仍然是痣自主的。鉴于组织微环境的众所周知的重要性，痣的生长停滞很可能是由非痣细胞（例如成纤维细胞）引发的。可以想象，这可能已被衰老机制抑制了生长。即便如此，这也不应偏离本论文中令人信服地表明的事实，即痣细胞本身绝对不是衰老的事实。因此，我们必须开始放开神经被细胞自主癌基因诱导的衰老所抑制的观念。