ABSTRACT

Many studies over the years have shown that non-genetic mechanisms for producing cell-to-cell variation can lead to highly variable behaviors across genetically identical populations of cells. Most work to date has focused on gene expression noise as the primary source of phenotypic heterogeneity, yet other sources may also contribute. In this Commentary, we explore organelle-level heterogeneity as a potential secondary source of cellular ‘noise’ that contributes to phenotypic heterogeneity. We explore mechanisms for generating organelle heterogeneity and present evidence of functional links between organelle morphology and cellular behavior. Given the many instances in which molecular-level heterogeneity has been linked to phenotypic heterogeneity, we posit that organelle heterogeneity may similarly contribute to overall phenotypic heterogeneity and underline the importance of studying organelle heterogeneity to develop a more comprehensive understanding of phenotypic heterogeneity. Finally, we conclude with a discussion of the medical challenges associated with phenotypic heterogeneity and outline how improved methods for characterizing and controlling this heterogeneity may lead to improved therapeutic strategies and outcomes for patients.

Introduction

Living cells are complex machines, comparable perhaps in complexity to computers. The analogy of machines is often used to describe molecules and cells, with important elements of the central dogma being performed by ‘protein machines,’ yet several features set biological systems apart from human-made machines. One of the dominant features of living systems is the high level of noise or variability. Biological molecular interactions are dominated by non-covalent bonds, the formation of which depend on inherently stochastic processes. Moreover, many important regulatory molecules are present only in low numbers, such that small fluctuations in number over time, or between cells, can lead to large changes in the activity of molecular pathways. Because of the low numbers, weak interactions and constant random motion of cellular components, cells are subject to continual variation across many parameters. Despite this variability, cells are able to perform important physiological functions, respond to external stimuli and develop into complex multicellular organisms. The question of the mechanisms and consequences of intercellular heterogeneity is not purely academic given that the treatment of many diseases is complicated by heterogeneity that renders some cells more resistant to treatment than others. In this Commentary, we review evidence concerning the underlying sources of organelle heterogeneity, how this heterogeneity affects cellular behavior and, finally, the challenges posed by such heterogeneity in the treatment of various diseases.

At a molecular level, the ‘state’ of a cell will depend on a large number of microscopic variables. One must consider not just expression levels of genes and proteins, but also the vast multitude of protein phosphorylation sites and other post-translational modifications, not to mention alternative splice isoforms, small RNAs, second messengers and metabolites. Although tools exist to study various state variables, there are currently no methods to comprehensively measure all molecular state variables at the same time within the same cell. Moreover, most of the methods for measuring molecular state variables are destructive to cells, rendering it impossible to study temporal variation or correlate molecular states with downstream behavior. Is there a way to comprehend how variation in these thousands or millions of molecular-state variables map onto observable variation in cell behavior?

To address this question, this Commentary will focus on heterogeneity at the level of subcellular structure, that is to say, at the level of organelle size, shape and distribution (Rafelski and Marshall, 2008; Marshall, 2011; Chan and Marshall, 2010), and attempt to characterize the sources and consequences of fluctuations at this scale. It is at such a scale that molecular processes become cellular processes, and it is therefore possible, at least in theory, that organelle scale variation reflects molecular-level variations. Many organelles serve as reaction vessels for biochemical reactions, and just as in the design of a chemical plant, a chemical engineer must carefully tune the size and geometry of reaction vessels to achieve a desired yield and purity, so too can we imagine that the size and geometry of organelles may be tied to their functional output. For example, the volume of an organelle will influence its capacity for storing reaction intermediates, whereas its surface area will influence the flux of molecules between the cytoplasm and the organelle lumen. For cellular structures with more mechanical functions (e.g. flagella, focal adhesions and stress fibers), the size and shape of structures will directly influence their mechanical properties. It is therefore reasonable to hypothesize that variation in organelle size and shape may lead to variation in a number of cellular behaviors and functions, such as motility, proliferation, apoptosis and tumorigenicity, amongst others. These differences in behavior will be important on a cellular level but might also be important at the population level as a means to phenotypically diversify the population, and either better meet the physiological needs of the organism (e.g. cells of the adaptive immune system that express highly variable antigen receptors) or improve the overall fitness of a population (e.g. Escherichia coli persistence, where a few cells in each population adopt a drug-tolerant state).

A second motivation for studying heterogeneity at the subcellular level is that it may lend fundamental insight into how the cellular structure is formed. Often in physics, once one has a model that can account for the average behavior of a system, a second way to test the validity of the model is to ask whether or not it can account for the fluctuations as well as the average behavior of the system. We therefore believe that quantitative analysis of noise in a biological system will be a powerful way to probe the fundamental principles that govern its behavior. Analyses of fluctuations and noise have, in the past, been used to probe a wide range of molecular phenomena, such as ion channel conductance (Neher and Stevens, 1977), flagellar rotation (Samuel and Berg, 1995; Shaevitz et al., 2005) and neurotransmitter exocytosis (Neher and Sakaba, 2001). We believe that similar approaches may provide a way to probe the mechanisms that regulate cellular architecture and behavior.

In this Commentary, we begin with an overview of cellular heterogeneity and illustrate the many contexts in which it is observed. We continue with a detailed discussion of possible sources of heterogeneity, which include temporal cell cycle progression, noisy gene expression and fluctuations in organelle dynamics. We then address questions of physiological relevance – in other words, does heterogeneity at the level of cellular structure and composition lead to differences in cellular behavior and function? Finally, we motivate the study of cellular noise and heterogeneity with the observation that many clinically important diseases are characterized by extensive heterogeneity and thus an improved understanding of this heterogeneity will be integral in developing better therapeutic strategies.

Types of intercellular phenotypic variability

At the behavioral level, phenotypic heterogeneities can be broadly subdivided into two classes: (i) ‘directed’ heterogeneities that play a role in normal developmental processes, and (ii) ‘non-directed’ heterogeneities, which do not exist as part of coordinated developmental pathways, but instead occur spontaneously owing to the inherent stochasticity of molecular processes. We note that the distinction between directed and non-directed is purely teleological, rather than mechanistic, and is presented primarily to inform the reader of different modes by which heterogeneities arise rather than to imply any mechanism.

Directed heterogeneities are the result of cellular decision-making processes that are coordinated within a population of cells to produce variable behavior across cells in a spatially precise and replicable fashion (Fig. 1A). During the development of multicellular organisms, a single cell undergoes many cycles of division to form a collection of cells that, although genetically clonal, exhibits specialized functions within each cell or group of cells. A common source of directed heterogeneity is asymmetrical cell division, in which a cell divides to produce two daughter cells of differing fates. A classic example of this type of deterministic heterogeneity is found in Drosophila melanogaster development, where ganglion mother cells consistently undergo asymmetric division to give rise to two daughter cells – a larger cell that will become a motor neuron and a smaller sibling cell of unspecified developmental fate (Bhat et al., 2014).

Fig. 1.

‘Directed’ vs ‘non-directed’ heterogeneity. (A) Schematic illustration of an example of ‘directed’ heterogeneity in which a morphogen gradient produces a coordinated, replicable and spatially precise pattern of cellular differentiation (top). This type of process is responsible for the coordinated process of body plan development in Drosophila melanogaster (bottom). (B) Schematic illustration of an example of ‘non-directed’ heterogeneity shows each cell making a stochastic cell fate decision that is statistically independent from those of its neighbors (top). This type of process is responsible for the stochastic choice of which color-sensitive variant of rhodopsin will be expressed in each photoreceptor cell of the Drosophila melanogaster compound eye (bottom). The image of the Drosophila compound eye is courtesy of Justin P. Kumar (Indiana University, Bloomington, IN), and the image in the inset has been reproduced from Wernet et al. (2006) with permission (@Nature Publishing Group).

Fig. 1.

‘Directed’ vs ‘non-directed’ heterogeneity. (A) Schematic illustration of an example of ‘directed’ heterogeneity in which a morphogen gradient produces a coordinated, replicable and spatially precise pattern of cellular differentiation (top). This type of process is responsible for the coordinated process of body plan development in Drosophila melanogaster (bottom). (B) Schematic illustration of an example of ‘non-directed’ heterogeneity shows each cell making a stochastic cell fate decision that is statistically independent from those of its neighbors (top). This type of process is responsible for the stochastic choice of which color-sensitive variant of rhodopsin will be expressed in each photoreceptor cell of the Drosophila melanogaster compound eye (bottom). The image of the Drosophila compound eye is courtesy of Justin P. Kumar (Indiana University, Bloomington, IN), and the image in the inset has been reproduced from Wernet et al. (2006) with permission (@Nature Publishing Group).

In contrast, non-directed heterogeneity arises through mechanisms whereby cells in a population make individual statistically independent decisions. One example of this is the generation of color-specific photoreceptors in the compound eye of D. melanogaster (Fig. 1B). During eye maturation, stochastic transcription factor expression directs each photoreceptor cell to express either a blue- or green-sensitive variant of rhodopsin (Wernet et al., 2006; Mikeladze-Dvali et al., 2005). Once these variants are expressed, the cells become behaviorally heterogeneous, exhibiting preferential sensitivity to different wavelengths of light. It is important to note that photoreceptor choice is made stochastically at the individual cell level such that each cell's choice occurs independently of its neighbors'. In this case, there is no global program dictating the precise spatial arrangement of photoreceptors, only an overall average dictated by the statistical properties of each choice.

In addition to the distinction between ‘directed’ and ‘non-directed’ heterogeneities, another important distinction can be drawn between irreversible differentiation and reversible fluctuations. Some cell-to-cell variation reflects permanent committed heterogeneities (e.g. lineage commitments, photoreceptor color choice, apoptosis), whereas other heterogeneities fluctuate and are reversible, as is the case in Escherichia coli persistence (Balaban et al., 2004) and Bacillus subtilis competence (Maamar et al., 2007).

These observations underline the complexities inherent in intercellular heterogeneities and highlight the important role they play in organismal development and behavior. Although important in normal healthy biology, these heterogeneities are also prevalent in many diseases and present significant therapeutic challenges, a topic we will explore in the second half of this Commentary.

Noise at different scales – molecular versus organelle-level variation

Cellular structure is a multi-scale phenomenon, with molecular-scale processes interacting with those at the organelle and cellular levels. We draw a distinction between molecular events, by which we mean events of the central dogma (e.g. transcription, translation and protein turnover) and organelle dynamics – i.e. the synthesis, inheritance and degradation of organelles. Given the fact that organelles can encapsulate entire biochemical pathways, fluctuations in organelle abundance or size are expected to have significant effects on biochemical output (Marshall, 2012). It is important to note that fluctuations at the organelle and cellular scales do not necessarily follow directly from molecular-level fluctuations, underlining the importance of studying fluctuations at many different scales.

Perhaps the most extensively studied source of variation is molecular variation. All cellular processes are under genetic control, often mediated by complex signaling pathways. Because gene expression, signaling and epigenetic regulation all involve potentially small numbers of molecules and their weak intermolecular interactions, statistical fluctuations can have large effects on gene expression and signaling (Raser and O'Shea, 2004; Elowitz et al., 2002; Volfson et al., 2006; Eldar and Elowitz, 2010; Ladbury and Arold, 2012). Fluctuations arise not only from variation in the production rates of molecules but also from variation in the partitioning of small numbers of molecules during cell division (Huh and Paulsson, 2011). Because noise at the level of gene expression has been extensively reviewed elsewhere, in this Commentary, we focus instead on variation at the organelle level, with the understanding that heterogeneity at this level may contribute to intercellular heterogeneity in ways different from those engendered by molecular heterogeneity.

Sources of organelle-level heterogeneity

Temporal heterogeneity resulting from the cell cycle

Random sampling from an asynchronous population of cells undergoing deterministic temporal variation in the form of the cell cycle will produce heterogeneity that is impossible to distinguish from noise. However, this type of variation is likely to be an important factor in intercellular heterogeneity because many aspects of cellular structure and function undergo regulated changes as a function of the cell cycle.

In mitosis, microtubules undergo a dramatic reorganization during which interphase microtubules are disassembled to form shorter mitotic spindles that facilitate chromosome condensation and congression (Cassimeris, 1999). Similarly, the endoplasmic reticulum (ER) has been shown to rearrange both its structure and subcellular localization during mitosis (Bergman et al., 2015), presumably as an aid to segregating the organelle to both daughters.

During the G1–S transition, mitochondria become highly fused. Maintenance of this hyperfused state through inhibition of the mitochondrial fission protein Dnp1 permanently blocks entry into S phase (Mitra et al., 2009), highlighting the link between mitochondrial morphology and cell cycle progression.

During G1, cells do not undergo drastic morphological changes but instead undergo continuous growth, such that cells further along in G1 are larger. This effect ramifies down to the organelle level, as many organelles have been shown to scale with cell size, with the size of an organelle (as judged by its surface area or volume) often scaling as a power law with the volume of the cell (Wuehr et al., 2008; Chan and Marshall, 2010; Uchida et al., 2011; Rafelski et al., 2012; Chan and Marshall, 2014). The surface area and volume can, and often do, show different scaling exponents (Chan and Marshall, 2014), and as a result, as a cell grows, not only does organelle size increase but organelle shape can additionally change as a function of the surface-to-volume ratio of the organelle. Quantifying the relationship between organelle and cell size scaling for a given cell type is thus critical for dissecting the possible contribution of cell growth to heterogeneity. If either the absolute size of an organelle, or the size of an organelle relative to the cell size, is functionally relevant, this type of allometric size scaling will result in functional heterogeneity among a population of cells at different stages of growth. There are thus two sources of size variation for any organelle – variation linked to cell size through a scaling relationship, and individual variation that may arise either spontaneously at the organelle level or as a result of fluctuations at the molecular level. These can in principle be distinguished by first estimating the scaling relationship that best describes the average behavior of many cells and then quantifying the degree to which individual organelle measurements deviate from the organelle size predicted by this scaling relationship.

Several studies have been undertaken to quantify the contribution of the cell cycle state to the overall population-level heterogeneity, with differing conclusions depending on the method of statistical analysis. Single-cell RNA sequencing (RNA-Seq) comparisons of known cell cycle-related transcripts with transcripts of unknown importance to the cell cycle suggest that only 7% of intercellular transcriptional variability arises from differences in cell cycle state (Buettner et al., 2015; McDavid et al., 2016). However, these studies focus on transcriptional variability at a fixed time point, which can give a deceptive view of the level of stochasticity at work in individual cells, suggesting that live-cell measurements will be essential in developing a more comprehensive understanding of intercellular heterogeneity.

Temporal heterogeneity resulting from the circadian cycle

The cell cycle is an important source of temporal variation but is by no means the only one. In many cells, the circadian cycle can drive variation in cell behavior and structure. This is most apparent in photosynthetic cells for which the time of day is of direct physiological relevance. In plant and algal cells, the circadian clock regulates remodeling of chloroplast architecture (Nassoury et al., 2005). In flagellated green algae, such as Chlamydomonas reinhardtii, flagella typically begin forming just before the onset of daylight and then gradually grow throughout the day. During the dark phase of the cycle, when cells undergo division, the flagella resorb (Kater, 1929). When flagellar lengths are measured, the variance of the length distribution is substantially greater in cultures that are ‘free running’ – i.e. grown under continuous light, compared to cultures in which the cells are synchronized and measured at specific time points in the cycle (W.F.M., unpublished data). This phenomenon has been shown to extend to mammalian cells, where the circadian clock drives periodic changes in actin dynamics (Gerber et al., 2013) and Golgi organization (Muueller and Gerber, 1985).

Both sources of temporal variation described here (cell cycle and circadian cycle) are deterministic predictable processes, although it is worth noting that if cells are not synchronized, the contribution of these processes to intercellular heterogeneity becomes difficult to unravel. For example, if cells with a circadian rhythm are grown in the absence of an entraining stimulus, at any point in time, different cells in the population will be at different phases of their circadian cycles, thus generating what looks like an apparently random source of heterogeneity at the population level, even though the underlying source is a deterministic process at the single cell level.

Heterogeneity resulting from fluctuations in organelle biogenesis rates

As discussed earlier, the molecular processes of life are subject to substantial fluctuations owing to the stochasticity of thermodynamic forces that drive molecular interactions. Consequently, all cellular trafficking, biosynthetic and self-assembly processes are subject to noise and variation, and given that the structure of cells and organelles is the result of many such processes, it is to be expected that overall organelle and cellular structure may also exhibit heterogeneity. This hypothesis has been proven using live-cell imaging experiments, where stochastic fluctuations in growth rates have been observed for neurite outgrowth (Odde and Buettner, 1998) as well as endocytotic activity (Dey et al., 2014). Similar fluctuations are seen when cellular processes are reconstituted in vitro – for example, microtubules display momentary fluctuations in stability that influence microtubule growth and shrinkage rates (Duellberg et al., 2016). Many organelles also undergo fission and fusion, processes which themselves are potentially subject to stochastic variation. In such cases, although the total mass of an organelle can only change as a result of biogenesis or degradation, fission and fusion events can drive changes in the size of individual organelles without altering their total mass within the cell. Interestingly, a recent model predicts that – assuming organelle-intrinsic rate constants that are independent of the number of organelles – fission and fusion events, coupled with organelle synthesis and degradation, can produce substantial heterogeneity in organelle abundance (Fig. 3) (Mukherji and O'Shea, 2014). Such models have been found to be in good agreement with measurements of intercellular variability in the number of yeast vacuoles, peroxisomes and Golgi (Mukherji and O'Shea, 2014). Recent quantitative analysis of size in living cells indicates that the rate of organelle biogenesis is independent of vacuole size or cell size (Chan et al., 2016), consistent with the assumptions of the Mukherji and O'Shea model.

Organelle biogenesis entails two processes, both of which are potentially subject to noise. The first is biosynthesis of the component molecules. As discussed previously, noise in gene expression leads to variation in the quantity of precursor molecules produced (Fig. 3). The second facet of biogenesis is intracellular trafficking. In many cases, components of an organelle are synthesized in one cellular location before being trafficked to the final site of assembly, such that variation in trafficking pathways can directly lead to variations in organelle size (Fig. 3) (Chan and Marshall, 2014). These trafficking processes themselves have been shown to be subject to large variation. For example, assembly of cilia and flagella require kinesin-based intraflagellar transport (IFT) of tubulin and other building blocks, a process mediated by IFT protein complexes. These complexes organize into linear arrays called IFT trains, where both the size of these trains and the timing of their entry into the cilia are variable. Similar fluctuations in processes contributing to organelle biogenesis have been observed in nuclear trafficking (Kubitscheck et al., 2005), cytoskeletal transport (Wacker et al., 1997; Bouzat, 2016) and membrane vesicle fusion during exocytosis (Bennett and Kearns, 2000).

Heterogeneity resulting from asymmetry in organelle partitioning

Partitioning of organelles between daughter cells during cell division can be an important source of variability, but this depends on (a) how evenly the partitioning system distributes organelles between the two daughter cells and (b) how rapidly, if at all, organelle abundance in the daughter cells is restored to the average. From a statistical perspective, if organelles are randomly partitioned into two equally sized daughter cells, the number of organelles inherited will follow a binomial distribution. Some partitioning events are clearly not random, as evidenced by the near 1:1 segregation of chromosomes and centrosomes during mitosis. However, other organelles, such as mitochondria and endosomes, are known to exhibit asymmetrical partitioning (Boldogh et al., 2001; Catlett and Weisman, 2000; Knoblach and Rachubinski, 2015). Interestingly, asymmetrical partitioning can produce distributions both more and less unequal than that predicted by binominal statistics (Fig. 2). For example, the budding yeast Saccharomyces cerevisiae exhibits more-than-random asymmetrical partitioning, where mitochondria are evenly partitioned between the mother and bud cells during cell division, despite the bud cell being significantly smaller in volume (Boldogh et al., 2001). In contrast, studies of GFP-tagged Golgi complexes in HeLa cells show less-than-random asymmetrical partitioning, a phenomenon attributed to the association of Golgi complexes into mitotic clusters that lower the overall number of units of distribution (Shima et al., 1997).

Fig. 2.

Variable models for organelle partitioning during cellular division. (A) Schematic illustration of four different modes of organelle partitioning. Strict 1:1 partitioning (green) partitions organelles evenly between two daughter cells. Noise-driven partitioning produces partitioning of organelles with a binomial distribution (blue). Cell-intrinsic mechanisms that actively contribute to asymmetrical partitioning may produce asymmetries that are either more (red) or less (yellow) random than those predicted by binomial statistics. (B) A simulated probability density function showing the possible range of distributions available in each partitioning model. In 1:1 partitioning, only one possible partitioning distribution is available, represented by a single vertical line (green). Models following less-than-random partitioning statistics (yellow) are more tightly constrained than those following binomial partitioning statistics (blue), whereas models following more-than-random partitioning statistics (red) are less tightly constrained. Depending on the consistency of cell-intrinsic partitioning mechanisms, these distributions may change [e.g. in a 100% consistent mechanism for producing more-than-random partitioning, the area surrounding x=0 (symmetrical partitioning) would not be accessible].

Fig. 2.

Variable models for organelle partitioning during cellular division. (A) Schematic illustration of four different modes of organelle partitioning. Strict 1:1 partitioning (green) partitions organelles evenly between two daughter cells. Noise-driven partitioning produces partitioning of organelles with a binomial distribution (blue). Cell-intrinsic mechanisms that actively contribute to asymmetrical partitioning may produce asymmetries that are either more (red) or less (yellow) random than those predicted by binomial statistics. (B) A simulated probability density function showing the possible range of distributions available in each partitioning model. In 1:1 partitioning, only one possible partitioning distribution is available, represented by a single vertical line (green). Models following less-than-random partitioning statistics (yellow) are more tightly constrained than those following binomial partitioning statistics (blue), whereas models following more-than-random partitioning statistics (red) are less tightly constrained. Depending on the consistency of cell-intrinsic partitioning mechanisms, these distributions may change [e.g. in a 100% consistent mechanism for producing more-than-random partitioning, the area surrounding x=0 (symmetrical partitioning) would not be accessible].

These deviations from binomial partitioning highlight the presence of cell-intrinsic mechanisms that result in asymmetrical partitioning, although these mechanisms are likely to differ depending on the degree of asymmetry (e.g. some mechanisms may augment asymmetry, whereas others may diminish it). Importantly, the existence of non-binomially distributed partitioning does not eliminate noise as a potential contributor to asymmetry, it only tells us that noise in and of itself cannot fully explain the observed data. It remains possible that observed distributions reflect a convolution of cell-intrinsic mechanisms coupled with stochastic noise.

A second factor that influences organelle heterogeneity is the timescale on which organelle abundance in daughter cells recovers to the population mean. This reversion to the mean has been observed for centrioles; cells have a mechanism for adjusting copy number by dampening centriole synthesis when too many centrioles are present and conversely for triggering synthesis when no centriole is present (Marshall, 2007). Similar observations have been made in algal cells, which modulate chloroplast synthesis to dampen any variation in chloroplast number after their random partitioning in mitosis (Hennis and Birky, 1984). We note that the propensity for and rate of recovery is likely to vary from organelle to organelle, as well as between cell types.

Heterogeneity resulting from fluctuations in organelle degradation rates

Degradation of organelles is a final contributor to the steady-state quantity of any organelle (Fig. 3). The statistical nature of the decision-making process that drives organelle degradation has not yet been extensively studied. Under the simplest model, if organelle degradation is triggered by a single random decision, organelle lifetimes are expected to follow an exponential distribution. If, however, organelle degradation is triggered by changes in their functional state or by accumulated damage, their lifetimes may be more narrowly distributed. In mitochondria, accumulated oxidative damage is associated with a slowdown in rates of degradation, producing enlarged mitochondria with lower rates of ATP production and increased rates of cell death (Moreira et al., 2010). These observations suggest the existence and importance of cell-regulatory pathways that control degradation rates.

Fig. 3.

Schematic illustration of several mechanisms that contribute to organelle morphology. In this case, mitochondrial (pink) morphology is influenced by transcriptional noise (blue) of molecular precursors, as well as variability in intracellular trafficking of these precursors to the site of assembly. Other mechanisms include organelle biogenesis, degradation (by the lysosome, green), fusion and fragmentation. Temporally driven processes, such as the cell and circadian cycles, also influence organelle morphology but are not pictured here.

Fig. 3.

Schematic illustration of several mechanisms that contribute to organelle morphology. In this case, mitochondrial (pink) morphology is influenced by transcriptional noise (blue) of molecular precursors, as well as variability in intracellular trafficking of these precursors to the site of assembly. Other mechanisms include organelle biogenesis, degradation (by the lysosome, green), fusion and fragmentation. Temporally driven processes, such as the cell and circadian cycles, also influence organelle morphology but are not pictured here.

From molecular and organelle heterogeneity to functional heterogeneity

We have discussed a number of sources of heterogeneity, both at the molecular and organelle levels, but for these heterogeneities to have biological consequence, they must relate, in some way, to cellular phenotype and behavior. These differences could include differences in cell-intrinsic behaviors and outputs (e.g. cell growth, cell-intrinsic apoptosis and signaling), as well as differential sensitivity to external inputs (e.g. drug response, cell-extrinsic apoptosis and differentiation cues). Currently, intercellular phenotypic variability has mostly been studied at the level of molecular variability in gene expression, with the contributions of organelle variability remaining largely unexplored. In the second half of this Commentary, we present evidence linking organelle size and shape to cellular behavior to underline the potential role that organelle heterogeneity plays in intercellular phenotypic variability. We conclude with an overview of the medical challenges posed by intercellular heterogeneity and discuss how organelle-level studies may contribute to a fuller understanding of this heterogeneity.

The hypothesis that organelle-level heterogeneity may lead to intercellular behavioral heterogeneity is premised on extensive links between organelle morphology and cellular behavior that have been observed through the years. As an example, apoptosis, or programmed cell death, is characterized by extensive mitochondrial fragmentation (Karbowski et al., 2004). Inhibition of the proteins responsible for fragmentation results in hyperfused mitochondria and a consequent decrease in apoptosis (Thomenius et al., 2011), suggesting that fragmentation may be integral to apoptosis rather than strictly a secondary consequence. At the clinical level, changes in mitochondrial morphology have been linked to several neurodegenerative disorders such as Alzheimer's (Fig. 4A) (Wang et al., 2008), optic atrophy (Frank, 2006) and Charcot–Marie–Tooth neuropathy (Frank, 2006). Similar findings have been observed for the ER, with ER morphology expanding in response to cellular stress (Wikstrom et al., 2013), as well as in the pancreatic β-cells of type 2 diabetic patients compared to healthy control patients (Fig. 4B). At the nuclear level, changes in morphology have long been diagnostic of cancer, with the extent of nuclear aberration often trending with the severity of prognosis (Fig. 4C) (Cibas and Ducatman, 2009).

Fig. 4.

Organelle morphology changes observed in diseased states. (A) Fluorescence images taken of fibroblast cells taken from a healthy control patient (left) and a patient diagnosed with Alzheimer's disease (right). Note the severe condensation and perinuclear localization of mitochondria (yellow-orange) in the Alzheimer's patient as compared to in the healthy control. (B) Electron micrographs of a pancreatic β-cell taken from a healthy control patient (left) and a patient diagnosed with type II diabetes (right). ER, endoplasmic reticulum; IG, insulin granules; M, mitochondria; N, nucleus. Note the enlarged ribbons of ER in the diabetic patient as compared to in the healthy control. (C) Cervical cells derived from a healthy control patient (left) and a patient diagnosed with late-stage cervical cancer (right). Note the severely enlarged nuclei (blue) and diminished cell-to-nucleus ratio in the cancer patient as compared to the healthy control. Images in A reproduced courtesy of Wang et al. (2008), ©Elsevier; images in B reproduced courtesy of Marchetti et al. (2007), ©Springer; images in C reproduced courtesy of Cibas and Ducatman (2009), ©Elsevier.

Fig. 4.

Organelle morphology changes observed in diseased states. (A) Fluorescence images taken of fibroblast cells taken from a healthy control patient (left) and a patient diagnosed with Alzheimer's disease (right). Note the severe condensation and perinuclear localization of mitochondria (yellow-orange) in the Alzheimer's patient as compared to in the healthy control. (B) Electron micrographs of a pancreatic β-cell taken from a healthy control patient (left) and a patient diagnosed with type II diabetes (right). ER, endoplasmic reticulum; IG, insulin granules; M, mitochondria; N, nucleus. Note the enlarged ribbons of ER in the diabetic patient as compared to in the healthy control. (C) Cervical cells derived from a healthy control patient (left) and a patient diagnosed with late-stage cervical cancer (right). Note the severely enlarged nuclei (blue) and diminished cell-to-nucleus ratio in the cancer patient as compared to the healthy control. Images in A reproduced courtesy of Wang et al. (2008), ©Elsevier; images in B reproduced courtesy of Marchetti et al. (2007), ©Springer; images in C reproduced courtesy of Cibas and Ducatman (2009), ©Elsevier.

Although these correlative observations do not denote a causal link between organelle morphology and cellular behavior, the ensemble of these observations leads to the compelling question of whether organelle morphology itself may play an integral role in determining cellular state and function, especially if we consider that organelle-level heterogeneity need not be a strict corollary of molecular heterogeneity, as previously discussed. An improved understanding of the link between organelle morphology and cellular behavior, whether it be correlative, causal, or a mix of the two, will prove critical in furthering our understanding of the principles that govern cellular architecture and behavior.

Intercellular heterogeneity as a key challenge in medicine

In this final section, we examine intercellular heterogeneity as it contributes to the development, progression, and treatment of two of the most pervasive and clinically challenging diseases in the world today – cancer and HIV/AIDS. We illustrate the challenges presented by intercellular heterogeneities and outline future avenues by which these heterogeneities may be controlled in the hopes of advancing therapeutic strategies and improving patient outcomes.

Cancer is a highly heterogeneous disease characterized by a vast array of subtypes, each with its own unique set of genomic signatures and molecular markers. Compounding this complexity is the observation of significant genomic heterogeneity within single tumors (Navin and Hicks, 2010; Torres et al., 2007; Zhang et al., 2014), such that chemotherapeutics targeting single pathways or oncogenic drivers often fail at fully abolishing a tumor (Fisher et al., 2013; Diaz Jr et al., 2012; Gore and Larkin, 2011; Szerlip et al., 2012). This difficulty in treatment can translate to poorer patient outcomes, as studies have shown that higher intratumoral genetic heterogeneity can correlate with a higher likelihood for patient relapse (Zhang et al., 2014).

To complicate matters further, intratumoral heterogeneity extends to the epigenetic and transcriptional levels as well. Recent studies have demonstrated that, within populations of clonally expanded cancer cells, individual cells can exhibit variable chemotherapeutic responses as a function of either their transcriptional (Lee et al., 2014) or epigenetic (Sharma et al., 2010) states. Furthermore, these chemotherapeutic sensitivities have been found to be transient and reversible, with outgrowth of drug-resistant cells producing a mixed population of drug-resistant and drug-sensitive cells. These observations highlight the importance of non-genetic drivers of behavioral heterogeneity as important determinants of chemotherapeutic response. Non-genomic markers have additionally proven clinically useful in diagnosing patients and in assessing their prognosis, as is the case in glioblastoma, where the promoter methylation state of a DNA repair enzyme has been identified as the most clinically predictive biomarker (Parker et al., 2016).

The ensemble of these observations underlines the considerable challenges posed by intercellular heterogeneity in the treatment of cancer. Although extensive, our understanding of this heterogeneity is still limited, and given previous arguments for the existence of organelle-level heterogeneity and the potential consequences for cellular behavior, we believe that future studies of organelle-level heterogeneity may aid in developing a more comprehensive understanding of intratumoral heterogeneity and the consequences posed therein for tumor behavior.

The task of controlling and mitigating intercellular heterogeneity is extremely challenging, but recent work in the field of HIV/AIDS presents promising evidence that this may be possible. HIV/AIDS treatment was revolutionized in 1997 with the development of combination antiretroviral therapies (cART) (Hammer et al., 1997; Gulick et al., 1997). This advancement dramatically increased the life expectancy of HIV/AIDS patients by changing HIV/AIDS from an acute disease to a chronic one in which HIV loads can be managed at low levels over many decades. However, cART can never be discontinued, as a stable reservoir of latent HIV-1 remains integrated in resting CD4+ T cells (Chun et al., 1995), with spontaneous reactivation in the absence of cART quickly producing viral loads similar to those observed before treatment (Dahabieh et al., 2015).

One promising approach to developing a cure for HIV/AIDS would be to reactivate the latent pool of HIV-1 provirus within the protective presence of cART; eventually, the latent pool would be cleared and cART could be discontinued. However, this approach has proven challenging, as the mechanisms of HIV-1 reactivation are multifaceted, and studies have shown that upon 99%+ reactivation of T cells harboring latent HIV-1 provirus, less than 1% of cells show successful viral induction (Ho et al., 2013). Although many cells had genetic mutations or deletions that rendered the virus replication-incompetent, a significant fraction (11.7%) of cells remained replication-competent from a genetic standpoint. Additionally, the authors found that, upon a second round of T-cell activation, a fraction of previously non-induced cells began actively producing virus; this transient nature of HIV latency has been attributed to stochastic gene expression of Tat, the HIV trans-activator of transcription (Weinberger et al., 2005).

In theory, one way to achieve more complete induction of latent HIV-1 provirus would be to increase the noisiness of HIV induction such that, over a given period of time, a higher proportion of cells falls into an induction-competent state. This principle was illustrated in a recent study that screened for compounds altering the expression noise of the HIV LTR promoter; the authors found that, in combination with T-cell-activating compounds, noise-enhancing compounds could successfully reactivate HIV in up to 50% of resting T cells (Dar et al., 2014).

These findings highlight the potential promise of therapies targeting intercellular heterogeneity, and although much work remains to be done to develop and refine modalities for modulating this heterogeneity, these recent findings present a promising proof-of-principle that addressing sources of intercellular heterogeneity can productively alter cellular behavior and disease pathology.

Conclusions and future prospects

Studies in recent years have highlighted the incredible complexity of biological organisms, with many layers of noise and variability built on top of an otherwise deterministic genome. These layers contribute to the phenomenon of intercellular phenotypic variability, whereby genetically identical cells can vary widely in their behavior. Many studies to date have identified molecular-level variability as an important source of phenotypic variability. However, given significant mechanisms for organelle-level heterogeneity, as outlined in this paper, coupled with evidence of links between organelle morphology and cellular behavior, one naturally wonders what role, if any, organelle-level heterogeneity may play in phenotypic variability.

To address this question, we should begin with improving our understanding of baseline organelle heterogeneity in healthy populations of cells, and how this heterogeneity relates to phenotypic variability. We may then build upon this foundation by examining how organelle heterogeneity is altered in disease contexts, e.g. do some diseases exhibit more or less heterogeneity when compared to non-diseased states? Furthermore, what role does heterogeneity play in the development and progression of complex diseases, and what therapeutic strategies can be developed to modulate this heterogeneity and meaningfully impact disease progression and patient outcomes?

The questions concerning both the sources and consequences of intercellular heterogeneity remain extensive, and it is our hope that future studies of organelle-level heterogeneity will complement work on genetic and molecular heterogeneity to produce a more comprehensive understanding of all aspects of intercellular heterogeneity. We hope that this Commentary has opened the reader's eyes to the potential importance of organelle-level heterogeneity and convinced them that further scientific study of this heterogeneity may offer new insights into a wide range of biological processes ranging from developmental biology to the initiation and progression of complex diseases.

Acknowledgements

We would like to acknowledge past and present members of the Marshall lab for guiding our thoughts about cellular heterogeneity. We also thank Adam Olshen for helpful discussions about statistics.

Footnotes

Funding

The authors acknowledge the support of a National Science Foundation (NSF) Graduate Research Fellowship (A.Y.C.), a University of California, San Francisco (UCSF) Discovery Fellowship (A.Y.C.), National Institutes of Health grant [R01 GM097017 (to W.F.M.)] and an NSF grant (1515456 to W.F.M.). Both authors are members of the NSF Center for Cellular Construction, supported by NSF grant 1548297. Deposited in PMC for release after 12 months.

References

Balaban
,
N. Q.
,
Merrin
,
J.
,
Chait
,
R.
,
Kowalik
,
L.
and
Leibler
,
S
. (
2004
).
Bacterial persistence as a phenotypic switch
.
Science
305
,
1622
-
1625
.
Bennett
,
M. R.
and
Kearns
,
J. L
. (
2000
).
Statistics of transmitter release at nerve terminals
.
Prog. Neurobiol.
60
,
545
-
606
.
Bergman
,
Z. J.
,
Mclaurin
,
J. D.
,
Eritano
,
A. S.
,
Johnson
,
B. M.
,
Sims
,
A. Q.
and
Riggs
,
B
. (
2015
).
Spatial reorganization of the endoplasmic reticulum during mitosis relies on mitotic kinase cyclin A in the early Drosophila embryo
.
PLoS ONE
10
,
e0117859
.
Bhat
,
K. M.
(
2014
).
Notch signaling acts before cell division to promote asymmetric cleavage and cell fate of neural precursor cells
.
Sci. Signal.
7
,
ra101
.
Boldogh
,
I. R.
,
Yang
,
H. C.
and
Pon
,
L. A.
(
2001
).
Mitochondrial inheritance in budding yeast
.
Traffic
2
,
368
-
374
.
Bouzat
,
S
. (
2016
).
Models for microtubule cargo transport coupling the Langevin equation to stochastic stepping motor dynamics: caring about fluctuations
.
Phys. Rev. E.
93
,
012401
.
Buettner
,
F.
,
Natarajan
,
K. N.
,
Casale
,
F. P.
,
Proserpio
,
V.
,
Scialdone
,
A.
,
Theis
,
F. J.
,
Teichmann
,
S. A.
,
Marioni
,
J. C.
and
Stegle
,
O.
(
2015
).
Computational analysis of cell-to-cell heterogeneity in single-cell RNA-sequencing data reveals hidden subpopulations of cells
.
Nat. Biotechnol.
33
,
155
-
160
.
Cassimeris
,
L
. (
1999
).
Accessory protein regulation of microtubule dynamics throughout the cell cycle
.
Curr. Opin. Cell Biol.
11
,
134
-
141
.
Catlett
,
N. L.
and
Weisman
,
L. S.
(
2000
).
Divide and multiply: organelle partitioning in yeast
.
Curr. Opin. Cell Biol.
12
,
509
-
516
.
Chan
,
Y.-H. M.
and
Marshall
,
W. F.
(
2010
).
Scaling properties of cell and organelle size
.
Organogenesis
6
,
88
-
96
.
Chan
,
Y.-H. M.
and
Marshall
,
W. F.
(
2014
).
Organelle size scaling of the budding yeast vacuole is tuned by membrane trafficking rates
.
Biophys. J.
106
,
1986
-
1996
.
Chan
,
Y.-H. M.
,
Reyes
,
L.
,
Sohail
,
S. M.
,
Tran
,
N. K.
and
Marshall
,
W. F.
(
2016
).
Organelle size scaling of the budding yeast vacuole by relative growth and inheritance
.
Curr. Biol.
26
,
1221
-
1228
.
Chun
,
T.-W.
,
Finzi
,
D.
,
Margolick
,
J.
,
Chadwick
,
K.
,
Schwartz
,
D.
and
Siliciano
,
R. F.
(
1995
).
In vivo fate of HIV-1-infected T cells: quantitatve analysis of the transition to stable latency
.
Nat. Med.
1
,
1284
-
1290
.
Cibas
,
E. S.
and
Ducatman
,
B. S.
(
2009
).
Cytology: Diagnostic Principles and Clinical Correlates
.
Philadelphia
:
Saunders Elsevier
.
Print
.
Dahabieh
,
M.
,
Battivelli
,
E.
and
Verdin
,
E.
(
2015
).
Understanding HIV latency: the road to an HIV cure
.
Annu. Rev. Med.
66
,
407
-
421
.
Dar
,
R. D.
,
Hosmane
,
N. N.
,
Arkin
,
M. R.
,
Siliciano
,
R. F.
and
Weinberger
,
L. S
. (
2014
).
Screening for noise in gene expression identifies drug synergies
.
Science
344
,
1392
-
1396
.
Dey
,
G.
,
Gupta
,
G. D.
,
Ramalingam
,
B.
,
Sathe
,
M.
,
Mayor
,
S.
and
Thattai
,
M.
(
2014
).
Exploiting cell-to-cell variability to detect cellular perturbations
.
PLoS ONE
9
,
e90540
.
Diaz
,
L. A.
, Jr
,
Williams
,
R. T.
,
Wu
,
J.
,
Kinde
,
I.
,
Hecht
,
J. R.
,
Berlin
,
J.
,
Allen
,
B.
,
Bozic
,
I.
,
Reiter
,
J. G.
,
Nowak
,
M. A.
, et al. 
(
2012
).
The molecular evolution of acquired resistance to targeted EGFR blockade in colorectal cancers
.
Nature
486
,
537
-
540
.
Duellberg
,
C.
,
Cade
,
N. I.
,
Holmes
,
D.
and
Surrey
,
T
. (
2016
).
The size of the EB cap determines instantaneous microtubule stability
.
Elife
5
,
e13470
.
Eldar
,
A.
and
Elowitz
,
M. B.
(
2010
).
Functional roles for noise in genetic circuits
.
Nature
467
,
167
-
173
.
Elowitz
,
M. B.
,
Levine
,
A. J.
,
Siggia
,
E. D.
and
Swain
,
P. S.
(
2002
).
Stochastic gene expression in a single cell
.
Science
297
,
1183
-
1186
.
Fisher
,
R.
,
Pusztai
,
L.
and
Swanton
,
C
. (
2013
).
Cancer heterogeneity: implications for targeted therapeutics
.
Br. J. Cancer
108
,
479
-
485
.
Frank
,
S
. (
2006
).
Dysregulation of mitochondrial fusion and fission: an emerging concept in neurodegeneration
.
Acta Neuropathol.
111
,
93
-
100
.
Gerber
,
A.
,
Esnault
,
C.
,
Aubert
,
G.
,
Treisman
,
R.
,
Pralong
,
F.
and
Schibler
,
U
. (
2013
).
Blood-borne circadian signal stimulates daily oscillations in actin dynamics and SRF activity
.
Cell
152
,
492
-
503
.
Gore
,
M. E.
and
Larkin
,
J. MG
. (
2011
).
Challenges and opportunities for converting renal cell carcinoma into a chronic disease with targeted therapies
.
Br. J. Cancer
104
,
399
-
406
.
Gulick
,
R. M.
,
Mellors
,
J. W.
,
Havlir
,
D.
,
Eron
,
J. J.
,
Gonzalez
,
C.
,
McMahon
,
D.
,
Richman
,
D. D.
,
Valentine
,
F. T.
,
Jonas
,
L.
,
Meibohm
,
A.
, et al. 
(
1997
).
Treatment with indinavir, zidovudine, and lamivudine in adults with human immunodeficiency virus infection and prior antiretroviral therapy
.
N. Engl. J. Med.
337
,
734
-
739
.
Hammer
,
S. M.
,
Squires
,
K. E.
,
Hughes
,
M. D.
,
Grimes
,
J. M.
,
Demeter
,
L. M.
,
Currier
,
J. S.
,
Eron
, Jr. 
J. J.
,
Feinberg
,
J. E.
,
Balfour
,
H. H.
,
Deyton
,
L. R.
, et al. 
(
1997
).
A controlled trial of two nucleoside analogues plus indinavir in persons with human immunodeficiency virus infection and CD4 cell counts of 200 per cubic millimeter or less
.
N. Engl. J. Med.
337
,
725
-
733
.
Hennis
,
A. S.
and
Birky
,
C. W
. (
1984
).
Stochastic partitioning of chloroplasts at cell division in the alga Olisthodiscus, and compensating control of chloroplast replication
.
J. Cell Sci.
70
,
1
-
15
.
Ho
,
Y.-C.
,
Shan
,
L.
,
Hosmane
,
N. N.
,
Wang
,
J.
,
Laskey
,
S. B.
,
Rosenbloom
,
D. I.
,
Lai
,
J.
,
Blankson
,
J. N.
,
Siliciano
,
J. D.
and
Siliciano
,
R. F
. (
2013
).
Replication-competent noninduced proviruses in the latent reservoir increase barrier to HIV-1 cure
.
Cell
155
,
540
-
551
.
Huh
,
D.
and
Paulsson
,
J.
(
2011
).
Random partitioning of molecules at cell division
.
Proc. Natl. Acad. Sci. USA
108
,
15004
-
15009
.
Karbowski
,
M.
,
Arnoult
,
D.
,
Chen
,
H.
,
Chan
,
D. C.
,
Smith
,
C. L.
and
Youle
,
R. J.
(
2004
).
Quantitation of mitochondrial dynamics by photolabeling of individual organelles shows that mitochondrial fusion is blocked during the Bax activation phase of apoptosis
.
J. Cell Biol.
164
,
493
-
499
.
Kater
,
J. M.
(
1929
).
Morphology and Division of Chlamydomonas with Reference to the Phylogeny of the Flagellate Neuromotor System
.
University of California Press
,
Berkeley, California
.
Knoblach
,
B.
and
Rachubinski
,
R. A.
(
2015
).
Motors, anchors, and connectors: orchestrators of organelle inheritance
.
Annu. Rev. Cell Dev. Biol.
31
,
55
-
81
.
Kubitscheck
,
U.
,
Grünwald
,
D.
,
Hoekstra
,
A.
,
Rohleder
,
D.
,
Kues
,
T.
,
Siebrasse
,
J. P.
and
Peters
,
R.
(
2005
).
Nuclear transport of single molecules: dwell times at the nuclear pore complex
.
J. Cell Biol.
168
,
233
-
243
.
Ladbury
,
J. E.
and
Arold
,
S. T
. (
2012
).
Noise in cellular signaling pathways: causes and effects
.
Trends Biochem. Sci.
37
,
173
-
178
.
Lee
,
M.-C. W.
,
Lopez-Diaz
,
F. J.
,
Khan
,
S. Y.
,
Tariq
,
M. A.
,
Dayn
,
Y.
,
Vaske
,
C. J.
,
Radenbaugh
,
A. J.
,
Kim
,
H. J.
,
Emerson
,
B. M.
and
Pourmand
,
N
. (
2014
).
Single-cell analysis of transcriptional heterogeneity during drug tolerance transition in cancer cell by RNA sequencing
.
Proc. Natl. Acad. Sci. USA
111
,
E4726
-
E4735
.
Maamar
,
H.
,
Raj
,
A.
and
Dubnau
,
D.
(
2007
).
Noise in gene expression determines cell fate in bacillus subtilis
.
Science
317
,
526
-
529
.
Marchetti
,
P.
,
Bugliani
,
M.
,
Lupi
,
R.
,
Marselli
,
L.
,
Masini
,
M.
,
Boggi
,
U.
,
Filipponi
,
F.
,
Weir
,
G. C.
,
Eizirik
,
D. L.
and
Cnop
,
M.
(
2007
).
The endoplasmic reticulum in pancreatic beta cells of type 2 diabetes patients
.
Diabetologia
50
,
2486
-
2494
.
Marshall
,
W. F
. (
2007
).
Stability and robustness of an organelle number control system: modeling and measuring homeostatic regulation of centriole abundance
.
Biophys. J.
93
,
1818
-
1833
.
Marshall
,
W. F.
(
2011
).
Origins of cellular geometry
.
BMC Biol.
9
,
57
.
Marshall
,
W. F
. (
2012
).
Organelle size control systems: from cell geometry to organelle-directed medicine
.
BioEssays
34
,
721
-
724
.
McDavid
,
A.
,
Finak
,
G.
and
Gottardo
,
R
. (
2016
).
The contribution of cell cycle to heterogeneity in single-cell RNA-seq data
.
Nat. Biotechnol.
34
,
591
-
593
.
Mikeladze-Dvali
,
T.
,
Wernet
,
M. F.
,
Pistillo
,
D.
,
Mazzoni
,
E. O.
,
Teleman
,
A. A.
,
Chen
,
Y.-W.
,
Cohen
,
S.
and
Desplan
,
C.
(
2005
).
The growth regulators warts/lats and melted interact in a bistable loop to specify opposite fates in Drosophila R8 photoreceptors
.
Cell
122
,
775
-
787
.
Mitra
,
K.
,
Wunder
,
C.
,
Roysam
,
B.
,
Lin
,
G.
and
Lippincott-Schwartz
,
J.
(
2009
).
A hyperfused mitochondrial state achieved at G1-S regulates cyclin E buildup and entry into S phase
.
Proc. Natl. Acad. Sci. USA
106
,
11960
-
11965
.
Moreira
,
P. I.
,
Carvalho
,
C.
,
Zhu
,
X.
,
Smith
,
M. A.
and
Perry
,
G.
(
2010
).
Mitochondrial dysfunction is a trigger of Alzheimer's disease pathophysiology
.
Biochim. Biophys. Acta
1802
,
2
-
10
.
Mukherji
,
S.
and
O'Shea
,
E. K.
(
2014
).
Mechanisms of organelle biogenesis govern stochastic fluctuations in organelle abundance
.
Elife
3
,
e02678
.
Muuller
,
O. M.
and
Gerber
,
H. B.
(
1985
).
Circadian changes of the rat pancreas acinar cell. A quantitative morphological investigation
.
Scand. J. Gastroenterol.
20
(
Suppl. 112
),
12
-
19
.
Nassoury
,
N.
,
Wang
,
Y.
and
Morse
,
D.
(
2005
).
Brefeledin a inhibits circadian remodeling of chloroplast structure in the dinoflagellates gonylaulax
.
Traffic
6
,
548
-
561
.
Navin
,
N. E.
and
Hicks
,
J.
(
2010
).
Tracing the tumor lineage
.
Mol. Oncol.
4
,
267
-
283
.
Neher
,
E.
and
Stevens
,
C. F.
(
1977
).
Conductance fluctuations and ionic pores in membranes
.
Ann. Rev. Biophys. Bioeng.
6
,
345
-
381
.
Neher
,
E.
and
Sakaba
,
T.
(
2001
).
Combining deconvolution and noise analysis for the estimation of transmitter release rates at the calyx of held
.
J. Neurosci.
21
,
444
-
461
.
Odde
,
D. J.
and
Buettner
,
H. M
. (
1998
).
Autocorrelation function and power spectrum of two-state random processes used in neurite guidance
.
Biophys. J.
75
,
1189
-
1196
.
Parker
,
N. R.
,
Hudson
,
A. L.
,
Khong
,
P.
,
Parkinson
,
J. F.
,
Dwight
,
T.
,
Ikin
,
R. J.
,
Zhu
,
Y.
,
Cheng
,
Z. J.
,
Vafaee
,
F.
,
Chen
,
J.
, et al. 
. (
2016
).
Intratumoral heterogeneity identified at the epigenetic, genetic and transcriptional level in glioblastoma
.
Sci. Rep.
6
,
22477
.
Rafelski
,
S. M.
and
Marshall
,
W. F.
(
2008
).
Building the cell: design principles of cellular architecture
.
Nat. Rev. Mol. Cell Biol.
9
,
593
-
602
.
Rafelski
,
S. M.
,
Viana
,
M. P.
,
Zhang
,
Y.
,
Chan
,
Y.-H.
,
Thorn
,
K. S.
,
Yam
,
P.
,
Fung
,
J. C.
,
Li
,
H.
,
Costa Lda
,
F.
and
Marshall
,
W. F
. (
2012
).
Mitochondrial network size scaling in budding yeast
.
Science
338
,
822
-
824
.
Raser
,
J. M.
and
O'Shea
,
E. K.
(
2004
).
Control of stochasticity in eukaryotic gene expression
.
Science
304
,
1811
-
1814
.
Samuel
,
A. D.
and
Berg
,
H. C.
(
1995
).
Fluctuation analysis of rotational speeds of the bacterial flagellar motor
.
Proc. Natl. Acad. Sci. USA
92
,
3502
-
3506
.
Shaevitz
,
J. W.
,
Block
,
S. M.
and
Schnitzer
,
M. J.
(
2005
).
Statistical kinetics of macromolecular dynamics
.
Biophys. J.
89
,
2277
-
2285
.
Sharma
,
S. V.
,
Lee
,
D. Y.
,
Li
,
B.
,
Quinlan
,
M. P.
,
Takahashi
,
F.
,
Maheswaran
,
S.
,
McDermott
,
U.
,
Azizian
,
N.
,
Zou
,
L.
,
Fischbach
,
M. A.
, et al. 
(
2010
).
A chromatin-mediated reversible drug-tolerant state in cancer cell subpopulations
.
Cell
141
,
69
-
80
.
Shima
,
D. T.
,
Haldar
,
K.
,
Pepperkok
,
R.
,
Watson
,
R.
and
Warren
,
G.
(
1997
).
Partitioning of the Golgi apparatus during mitosis in living HeLa cells
.
J. Cell Biol.
137
,
1211
-
1228
.
Szerlip
,
N. J.
,
Pedraza
,
A.
,
Chakravarty
,
D.
,
Azim
,
M.
,
McGuire
,
J.
,
Fang
,
Y.
,
Ozawa
,
T.
,
Hollande
,
E. C.
,
Huse
,
J. T.
,
Jhanwar
,
S.
, et al. 
(
2012
).
Intratumoral heterogeneity of receptor tyrosine kinases EGFR and PDGFRA amplification in glioblastoma defines subpopulations with distinc growth factor response
.
Proc. Natl. Acad. Sci. USA
109
,
3041
-
3046
.
Thomenius
,
M.
,
Freel
,
C. D.
,
Horn
,
S.
,
Krieser
,
R.
,
Abdelwahid
,
E.
,
Cannon
,
R.
,
Balasundaram
,
S.
,
White
,
K.
and
Kornbluth
,
S.
(
2011
).
Mitochondrial fusion is regulated by Reaper to modulate Drosophila programmed cell death
.
Cell Death Differ.
18
,
1640
-
1650
.
Torres
,
L.
,
Ribeiro
,
F. R.
,
Pandis
,
N.
,
Andersen
,
J. A.
,
Heim
,
S.
and
Teixeira
,
M. R.
(
2007
).
Intratumor genomic heterogeneity in breast cancer with clonal divergence between primary carcinomas and lymph node metastases
.
Breast Cancer Res. Treat.
102
,
143
.
Uchida
,
M.
,
Sun
,
Y.
,
McDermott
,
G.
,
Knoechel
,
C.
,
Le Gros
,
M. A.
,
Parkinson
,
D.
,
Drubin
,
D. G.
and
Larabell
,
C. A.
(
2011
).
Quantitative analysis of yeast internal architecture using soft X-ray tomography
.
Yeast
28
,
227
-
236
.
Volfson
,
D.
,
Marciniak
,
J.
,
Blake
,
W. J.
,
Ostroff
,
N.
,
Tsimring
,
L. S.
and
Hasty
,
J.
(
2006
).
Origins of extrinsic variability in eukaryotic gene expression
.
Nature
439
,
861
-
864
.
Wacker
,
I.
,
Kaether
,
C.
,
Krömer
,
A.
,
Migala
,
A.
,
Almers
,
W.
and
Gerdes
,
H. H.
(
1997
).
Microtubule-dependent transport of secretory vesicles visualized in real time with a GFP-tagged secretory protein
.
J. Cell Sci.
110
,
1453
-
1463
.
Wang
,
X.
,
Su
,
B.
,
Fujioka
,
H.
and
Zhu
,
X
. (
2008
).
Dynamin-like protein 1 reduction underlies mitochondrial morphology and distribution abnormalities in fibroblasts from sporadic Alzheimer's disease patients
.
Am. J. Pathol.
173
,
470
-
482
.
Weinberger
,
L. S.
,
Burnett
,
J. C.
,
Toettcher
,
J. E.
,
Arkin
,
A. P.
and
Schaffer
,
D. V.
(
2005
).
Stochastic gene expression in a lentiviral positive-feedback loop: HIV-1 Tat fluctuations drive phenotypic diversity
.
Cell
122
,
169
-
182
.
Wernet
,
M. F.
,
Mazzoni
,
E. O.
,
Çelik
,
A.
,
Duncan
,
D. M.
,
Duncan
,
I.
and
Desplan
,
C.
(
2006
).
Stochastic spineless expression creates the retinal mosaic for colour vision
.
Nature
440
,
174
-
180
.
Wikstrom
,
J. D.
,
Israeli
,
T.
,
Bachar-Wikstrom
,
E.
,
Swisa
,
A.
,
Ariav
,
Y.
,
Waiss
,
M.
,
Kaganovich
,
D.
,
Dor
,
Y.
,
Cerasi
,
E.
and
Leibowitz
,
G.
(
2013
).
AMPK regulated ER morphology and function in stressed pancreatic β-cells via phosphortylation of DRP1
.
Mol. Endocrinol.
27
,
1706
-
1723
.
Wühr
,
M.
,
Chen
,
Y.
,
Dumont
,
S.
,
Groen
,
A. C.
,
Needleman
,
D. J.
,
Salic
,
A.
and
Mitchison
,
T. J.
(
2008
).
Evidence for an upper limit to mitotic spindle length
.
Curr. Biol.
18
,
1256
-
1261
.
Zhang
,
J.
,
Fujimoto
,
J.
,
Zhang
,
J.
,
Wedge
,
D. C.
,
Song
,
X.
,
Zhang
,
J.
,
Seth
,
S.
,
Chow
,
C.-W.
,
Cao
,
Y.
,
Gumbs
,
C.
, et al. 
(
2014
).
Intratumor heterogeneity in localized lung adenocarcinomas delineated by multiregion sequencing
.
Science
346
,
256
-
259
.

Competing interests

The authors declare no competing or financial interests.