Cell-in-cell phenomena across the tree of life

Cells in obligately multicellular organisms by definition have aligned fitness interests, minimum conflict, and cannot reproduce independently. However, some cells eat other cells within the same body, sometimes called cell cannibalism. Such cell-in-cell events have not been thoroughly discussed in the framework of major transitions to multicellularity. We performed a systematic review of 508 articles to search for cell-in-cell events across the tree of life, the age of cell-in-cell-related genes, and whether cell-in-cell events are associated with normal multicellular development or cancer. Out of the 38 cell-in-cell-related genes found in the literature, 14 genes were over 2.2 billion years old, i.e., older than the common ancestor of some facultatively multicellular taxa. Therefore, we propose that cell-in-cell events originated before the origins of obligate multicellularity. Cell-in-cell events are found almost everywhere: across some unicellular and many multicellular organisms, mostly in malignant rather than benign tissue, and in non-neoplastic cells. Thus, our results show that cell-in-cell events exist in obligate multicellular organisms, but are not a defining feature of them. The idea of eradicating cell-in-cell events from obligate multicellular organisms as a way of treating cancer, without considering that cell-in-cell events are also part of normal development, should be abandoned.


Introduction
(entos?s OR "homotypic cell cannibalism" OR "cell cannibalism" OR emperitos?s OR enclysis) AND (vertebrat* OR urochord* OR cephalochord* OR echinoderm* OR protostom* OR cnidaria* OR ascomycot* OR basidiomycot* OR amoebozoa* OR embryophyt* OR chlorophyt* OR rhodophyt* OR stramenopila* OR bacter*) During January 2022 to January 2023, we also searched for articles that mentioned "phagocytosis" and cases of cannibalism specifically in the taxa shown in the phylogenetic tree of Aktipis et al 25 . This led to 352 articles. When an article mentioned "entos?s", "homotypic cell cannibalism", "cell cannibalism", "emperitos?s", and referenced another article, we searched for the original publication and included the original publication in Table   1 if the content was relevant to our search. This method has been previously used when conducting systematic reviews 26 . Searching back for such citations added 156 articles to our list. For a more objective assessment of the literature, two of the authors read several of the articles independently. L.H.C. assessed 249 articles, reading from the oldest to the most recent, and S.E.K. assessed 360 articles, reading from the most recent to the oldest. Out of these, a total of 101 articles were assessed by both L.H.C. and S.E.K. We removed patents, web resources, books, newsletter articles, government documents, book chapters, newspaper articles, conference proceedings, reviews, and non-English articles, which led to 337 articles for further assessment. We also excluded 222 articles that turned out to be about irrelevant topics, review articles that directed us to more relevant articles (specifically to articles with original data), articles with no information about cell-in-cell phenomena in specific taxa, or with no information regarding the fate of the engulfed or host cell in terms of one or both remaining alive after the cell-in-cell event (Supplementary Table).
Our final assessment included 115 articles (Supplementary Table; Supplementary   Figure). We collected the following information from these 115 articles: (1) whether the cellin-cell phenomenon was between heterospecifics or conspecifics; (2) whether the host cell engulfed the whole prey cell; (3) whether both cells remained alive after the cell-in-cell event or at least one cell died; (4) whether both cells were non-neoplastic cells or at least one cell was a neoplastic cell; and (5) the specific taxon of the host cell. We included this information in our across-species comparisons (Table 1; Figure 1).
Our third aim was to examine associations between cell-on-cell phenomena and the evolution of multicellularity across 20 taxa based on previous scales of multicellularity 25,27 .
We performed the following ordinal categorical scaling: we categorized taxa in Table 1A according to their multicellularity levels as "unicellular" [0], "simple or aggregative multicellularity" [1], or "complex multicellularity" [2]. If several such multicellularity levels were found in a taxon, we assigned that taxon to the highest reported level of multicellularity.
We also categorized cell-in-cell phenomena from Table 1A based on the level of 'selfishness' of the interacting cells. The cell-in-cell categories were "no cell-in-cell phenomena reported/found" [0], "heterospecific cell-in-cell phenomena where both cells remain alive" [1], "heterospecific cell-in-cell phenomena where at least one cell dies" [2], "conspecific cellin-cell phenomena where both cells remain alive" [3], "conspecific cell-in-cell phenomena where at least one cell dies" [4], "conspecific cell-in-cell phenomena where both cells remain alive and at least one of the cells is a neoplastic cell" [5], "conspecific cell-in-cell phenomena where at least one cell dies and at least one of the cells is a neoplastic cell" [6]. If several such cell-in-cell categories were found in a taxon, we assigned that taxon to the highest reported 'selfishness' index.

Gene functional information
Within the 115 articles shown in Table 1, we searched for any mentioned markers of entosis, cannibalism, phagocytosis, emperitosis, and emperipolesis. We collected the names of the cell-in-cell-related genes and information about their cell-in-cell-related function from these articles.

Gene age data
We found the human homologs of the cell-in-cell-related genes that we had identified ( Table 2) and obtained their evolutionary age using a human gene age database published in previous work 28 . These gene ages are determined as the maximum phylogenetic divergence time between humans and the species represented in each gene ontology, as given in the TimeTree database 29,30 . We could not find human homologs of two cell-in-cell-related genes (AlyA and FspA) 31 .
AlyA is a gene found in Klebsiella pneumoniae, a gram-negative Enterobacterium.
This gene encodes for Alginate lyase, and it is also found in brown (Phaeophyceae) and red algae (Rhodophyta) 32 , which leads to an age of at least 4250 MYA based on the estimated time of divergence 30 .
FspA is a gene found in Campylobacter jejuni, also a gram-negative bacterium. It encodes for the Type 3 secretion system protein 33 and it is found in a great number of eubacteria 34 . This indicates that this gene most likely emerged around 4250 MYA with the evolution of bacteria 30 .

Results
Cell-in-cell phenomena have been found in 16 taxonomic groups across the tree of life. Cell-in-cell phenomena across the seven different phyla examined in our study can be separated into six different categories from the perspective of social evolution ( Fig. 1; Table   1): 1. Heterospecific killing between non-neoplastic cells where at least one of the resulting cells dies, is the most common phenomenon out of the six categories of cell-in-cell phenomena appearing in all 7 examined unicellular, facultative or obligately multicellular phyla ( Fig. 1; Table 1).
2. Conspecific killing between non-neoplastic cells where at least one of the resulting cells dies is the second most common phenomenon appearing in three out of the seven examined unicellular, facultative, or obligately multicellular phyla ( Fig. 1; Table 1).

Heterospecific cell-in-cell phenomena between non-neoplastic cells where
both of the cells remain alive have been found in three out of the seven examined unicellular or facultatively multicellular taxa ( Fig. 1; Table 1).

Conspecific cell-in-cell phenomena between non-neoplastic cells where both
of the cells remain alive have been found in one out of the seven examined obligately multicellular phyla (echinoderms) ( Fig. 1; Table 1).
In the domains of archaea and bacteria, only heterospecific killing between nonneoplastic cells has been found in bacteria where at least one of the resulting cells dies.
Across the four examined divisions of Ascomycota, Basidiomycota, Chlorophyta, and Rhodophyta, only the Chlorophyta have a form of cell-in-cell event which is heterospecific killing between non-neoplastic cells where at least one of the resulting cells dies. Across the two examined clades of stramenopiles and embryophyta, the former have conspecific killing between non-neoplastic cells where at least one of the resulting cells dies, whereas the latter have heterospecific killing between non-neoplastic cells where at least one of the resulting cells dies. Across the three examined subphyla of tunicates, cephalochordata, and vertebrates, the tunicates have conspecific killing between non-neoplastic cells where at least one of the resulting cells dies, the cephalochordata have heterospecific killing between non-neoplastic cells where at least one of the resulting cells dies, and all cell-in-cell categories exist in vertebrates except for heterospecific cell-in-cell events between non-neoplastic cells where both of the resulting cells remain alive ( Fig. 1; Table 1).
In many taxa out of the 20 (Fig. 1), many of the above phenomena have not been found or have not been searched for (Table 1A). Within vertebrates, cell-in-cell phenomena have been described in eight species (Table 1B).
Conspecific cell-in-cell phenomena also occur between non-neoplastic cells and appear in multicellular organisms that have no known cancer or cancer-like development.
According to the literature (Table 1), there are several examples of cell-in-cell phenomena between non-neoplastic cells. Also, porifera display conspecific cell-in-cell phenomena but have no known cancer-like growth (Table 1). An important caveat here is that the fact that no cancer-like phenomena have been reported in these taxa does not mean that they do not get cancer. They may not have been adequately studied yet to know if they can get cancer 25 .
They are related to the AMPK pathway, folate-sensing pathway, cell-cell adhesion, entosis, phagocytosis, intracellular bacterial killing, cell cannibalism, lysozyme activity, and lysosomal maturation ( Table 2). 24 cell-in-cell-related human genes originated after the common ancestor of amoeba and humans. These genes (  Table 1A.
Cell-in-cell phenomena across the tree of life. We obtained data on the levels of multicellularity and cancer from Aktipis et al. 25

Cannibal neoplastic and non-neoplastic cells
Cell-in-cell phenomena happen between neoplastic and non-neoplastic cells.
Depending on the details of the cell interaction, the outcome of cannibalism can be either the growth or shrinkage of the neoplasm. Thus, cannibalism cannot be considered a characteristic of only uncontrollably dividing 'selfish' neoplastic cells.
Due to their relatively higher nutritional demands, neoplastic cells are often cannibals 159 . Cell-in-cell phenomena can lead to aneuploidy and/or tetraploidy, and more aggressive and invasive tumors 59,160 . Therefore, it is no surprise that cannibalism is more often found in malignant than benign biopsies and urine cytology samples 161,162 , in metastatic tumors than in primary melanoma 17 , in aggressive than in non-aggressive giant cell granuloma 163 68 . Phagocytosis is also a common mechanism for the immune system to discard pathogenic foreign cells, as well as cells from the same organism that have been programmed for disposal ( Table 1).

Origins of cell-in-cell-related genes
We searched for associations between 38 cell-in-cell-related genes and the origins of obligate multicellularity, as well as correlations between the levels of 'selfishness' of cell-incell phenomena and the level of multicellularity, but did not find any significant associations.   (Table 3)

Open question: connection between microscopic and macroscopic cannibalistic phenomena
There may also be connections between microscopic and macroscopic cannibalistic phenomena. Factors that drive cannibalism in the microscopic world (micrometer scale) may also drive cannibalism in the macroscopic world (centimeter or meter scale). Examples of such analogies can be seen in Table 3. Under starvation, both cells and obligate multicellular organisms (e.g., honeybees, mantises, and mice) perform cannibalism. Mammals with large litter sizes also cannibalize their young in periods of food scarcity 187 . Under attack by the enemy (immune system or specific tribe), cancer cells and humans (respectively) cannibalize their enemy. Upon landing in a new environment, entotic uterine cells and cannibalistic beetles are more likely to survive (Table 3). However, no one has yet quantitatively estimated all the different factors (natural selection, random genetic drift, mutation, migration) that may drive cannibalistic processes both at the microscopic and macroscopic scales.

Conclusion
Overall, this study is the first to systematically analyze cell-in