“When a plant goes to seed, its seeds are carried in all directions; but they can only live and grow if they fall on congenial soil.” – S. Paget, 1889.

In 1889, Stephen Paget made an insightful observation following autopsy of 735 patients who died with breast cancer. He observed that there is a predilection for breast carcinomas to metastasize to bone. From this observation, he inferred that the tumor cell < < seed > > needed an appropriate <<soil>> in order to complete the last stages of the metastatic process [1]. This so-called seed and soil hypothesis challenged prior assertions that organotropism of metastasis was solely based upon anatomical blood distribution patterns, a debate that persisted for more than a century. In subsequent years, researchers have confirmed that selective colonization is dependent upon delivery of cells as well as cellular responses when they get to different sites (reviewed in [2]).

Often less appreciated is that Paget’s insight represented one of the first recorded recognitions of tumor-microenvironment interactions. If we take that initial insight to first principles, Paget was asserting that tumor cells were responding to signals from other cells, extracellular matrices, soluble molecules, and environmental conditions. The questions being asked in this special issue of Cancer & Metastasis Reviews relate to how those signals are translated into modified cellular behavior. [Note: tumor cells also send signals to the microenvironment; however, for purposes of this editorial, only incoming signals are being considered. The same principles are operational in reverse as well.]

Briefly, cellular sensing of signals must be translated into differential expression of genes which, broadly speaking, reflect changes in epigenetics. In addition to manipulation of signaling cascades from the cell surface to the nucleus, there are changes in chromatin which regulate gene expression that, in turn, change cellular behavior. Much of the focus of this special issue relate to covalent modifications of histones that either open or close chromatin which exposes or restricts access of transcription factors to the DNA helix. Ultimately, the hope of this line of investigation is that, by identifying common histone marks, critical regulators of metastasis will be identified and later exploited in therapeutic settings. As readers will note, realization of this promise is more challenging than desired.

In this editorial, I want to briefly summarize state-of-the-art as well as point out some knowledge gaps that deserve additional research focus to achieve the ultimate goal of clinical control of cancer metastasis. Without disparaging the superb work published here and elsewhere, I will point out limitations, both technical and theoretical, which I believe should be considered more extensively moving forward.

The vast majority of metastases are clonal in origin. As a result, the well-recognized heterogeneity that exists in every cancer [3, 4] makes bulk analysis too insensitive since fewer than 1 in 10,000 cells successfully colonize secondary sites [5]. The relevant epigenetic changes are likely masked by the majority cells’ chromatin states. Single cell analysis will be required. The challenge is that it is currently impossible to determine which cells would have successfully metastasized since the analysis itself requires killing cells. Challenge #1: How can the signalto-noise ratio be improved? Challenge #2: Is there potentially a way in which single cell analysis can preserve cell viability?

Another limitation is that there are literally hundreds of signals impinging upon metastatic cells at a given time. Eventually, it will be necessary to sift through those interactions to determine which are the (most) critical ones. It should also be noted that pO2 changes dramatically over a few millimeters. Why is this relevant? Previous studies have demonstrated that hypoxia can stimulate metastasis [6]. What level of hypoxia is required? This parameter, as an example, illustrates how spatial context is critical. The advent of techniques for single cell analysis of transcriptomes may provide some of these answers and may solve Challenge #3: maintaining spatial information. Nonetheless, the lost spatial information in current TCGA (or similar) databases should be acknowledged when interpreting expression data, especially when dealing with metastasis as an endpoint. Moreover, the relative lack of data comparing paired samples from primary tumors and metastases should compel deliberate, focused collection of such samples for future studies. While much has been learned from cell lines, the metastasis data from human materials is woefully inadequate.

Moreover, temporal changes need to be addressed. The transit time for primary tumor cells to reach a distant organ, such as the lungs – even when not being directly deposited into (seeding) that organ – is seconds to minutes. A cell that must detach from the primary tumor mass must then reestablish attachments at another site within that short time. The transit time is too short to initiate a cell signal, alter chromatin, initiate transcription, translate the mRNA, and localize the newly synthesized protein to the cell surface. Adding to the complexity, during transit neoplastic cells are interacting with plasma proteins, leukocytes, red blood cells, endothelial cells, platelets and sometimes other cancer cells. Each of those interactions likely alters cellular epigenetics in some way. Challenge #4: Can temporal information be captured, especially for short-lived processes? Relatedly, what technical modifications will be required that could allow single cell analyses while maintaining viability? As a result, both spatial and temporal parameters must be taken into account.

The eventual colonization of ectopic tissues requires a seeded cell to positively respond to the local milieu. That cell, like all cancer cells, generates new heterogeneity within the metastatic mass. Those attributes are parallel properties of stem cells, which has led to the concept that successful metastasis involves specialized cancer stem cell populations. Normally (relatively) quiescent cells when present in a specialized niche can be induced to proliferate, differentiate, or regenerate daughter stem cells, depending upon incoming signals [7, 8]. While some of those signals can be intuitively described, Challenge #5 will require identification of those signals and the cells’ epigenetic responses to them [9, 10].

Relatedly, one of the hallmarks of metastasis [2] is cellular plasticity. The confluence of stemness, EMT, genetic instability, differentiation and adaptation are part of the rubric of metastatic potential. Yet, the underlying mechanisms of cellular plasticity are relatively unexplored, in part, because they integrate most of the challenges listed above [11]. Capturing a cell state at a particular space and time along with knowing which signal(s) are impinging upon that cell will ultimately address Challenge #6: How do epigenetic changes define plasticity or susceptibility to adaptation?

In recent years, other epigenetic mechanisms have been emerging which have not been directly linked to altered chromatin structure. Specifically, tRNA fragments (tRF) confer heritable changes in progeny but their mechanisms of action are varied [12]. In one example, Qi Chen and colleagues bred normal chow-fed or high fat diet-fed male mice with females and observed that metabolism in the pups was differentially affected by the sires’ passing along of tRF [13]. Furthermore, those changes in the embryos were not associated with changes in DNA methylation.

tRF are cleaved from precursor and/or mature tRNA [12] and have been shown to regulate translation, ribosome biogenesis, epigenetic inheritance, retrotransposons, viral replication, hematopoiesis, tumorigenesis, and metastasis. Their impacts can be long-lived or transient. As a result, they could represent an explanation for the just-in-time delivery of critical molecules during the metastatic process. Similarly, ribosomal RNA fragments (rRF) have been identified and are differentially expressed [14], although they have not been as extensively studied as tRF or other small non-coding RNA, like miRNA or lncRNA.

My group has been exploring polymorphic tRF as regulators of metastatic efficiency. Unfortunately, we have encountered many of the issues raised above. Nonetheless, the notion that there are as-yet-unknown epigenetic marks and mechanisms seems quite probable. Challenge #7: How extensive are the roles of these newer epigenetic mechanisms? Perhaps new mechanisms for regulating gene expression may resolve some of the complications raised above.

Increasing evidence suggests that epigenetic marks can influence gene expression at long distances within the nuclear matrix and chromatin [15]. As a result, multiple genes could be impacted by a single mark. And since metastasis is a complex disease, involving dozens of genes, defining the quantitative trait loci (QTL) and the epigenetic QTL (eQTL) will eventually need to be assessed. Current state-of-the-art mostly effectively measures one locus at a time, emerging assays and informatics will eventually integrate the petabytes of data to formulate a more complete picture of the epigenetic changes that occur during the process of metastasis which contribute to dynamic heterogeneity of the phenotype. Exciting times are ahead.