When Dicty Met Myco, a (Not So) Romantic Story about One Amoeba and Its Intracellular Pathogen

In recent years, Dictyostelium discoideum has become an important model organism to study the cell biology of professional phagocytes. This amoeba not only shares many molecular features with mammalian macrophages, but most of its fundamental signal transduction pathways are conserved in humans. The broad range of existing genetic and biochemical tools, together with its suitability for cell culture and live microscopy, make D. discoideum an ideal and versatile laboratory organism. In this review, we focus on the use of D. discoideum as a phagocyte model for the study of mycobacterial infections, in particular Mycobacterium marinum. We look in detail at the intracellular cycle of M. marinum, from its uptake by D. discoideum to its active or passive egress into the extracellular medium. In addition, we describe the molecular mechanisms that both the mycobacterial invader and the amoeboid host have developed to fight against each other, and compare and contrast with those developed by mammalian phagocytes. Finally, we introduce the methods and specific tools that have been used so far to monitor the D. discoideum—M. marinum interaction.


INTRODUCTION
The Amoeba Dictyostelium discoideum Dictyostelium discoideum is a soil amoeba that feeds on bacteria by phagocytosis. The Amoebozoa branch split from the common lineage leading to fungi and animals, shortly after these crown organisms split from the Plant lineage (Eichinger et al., 2005). D. discoideum has a very ATrich (77.75%) haploid genome, organized into 6 chromosomes, which has been sequenced and annotated (dictyBase, 2004;Eichinger et al., 2005). The isolation in the 70's of axenic strains that could feed on liquid medium by pinocytosis (Watts and Ashworth, 1970) enabled the easy laboratory culture of D. discoideum. In addition, this amoeba is genetically tractable, being relatively easy to transform for gene expression from extrachromosomal plasmids, generation of knockins, and knockouts by homologous recombination (Veltman et al., 2009a,b;Wiegand et al., 2011;Mukai et al., 2016), and mutagenesis by Restriction Enzyme-Mediated Integration (REMI) (Kuspa, 2006). Moreover, D. discoideum can be used in multiple cell biology, biochemistry, and imaging assays, as well as for high-throughput screens and RNAseq analysis (Eichinger and Rivero, 2013;Kicka et al., 2014;Rosengarten et al., 2015). For these reasons, D. discoideum has been extensively used as a model system to study very diverse biological processes such as motility, chemotaxis, vesicular trafficking, gene expression, facultative multicellularity, and hostpathogen interactions (Barry and Bretscher, 2010;Maniak, 2011;Stevense et al., 2011;Loomis, 2014;Tosetti et al., 2014;Nichols et al., 2015).

Central Role of Autophagy in Development and Host-Microbe Interaction
In conditions of nutrient repletion, D. discoideum lives as a vegetative unicellular organism. However, when resources become scarce, the amoeba triggers a developmental program for the aggregation of hundreds of thousands of cells into a true multicellular organism (Raper, 1935;Dormann et al., 2000;Williams, 2006). This process is known as the "developmental cycle" of D. discoideum and has been extensively studied as one of the evolutionary origins of multicellularity. Aggregation and morphogenesis during development require a high level of cellular activity. To undertake this metabolic demand in the absence of nutrients, D. discoideum relays on macroautophagy (hereafter referred as autophagy), a major catabolic pathway in eukaryotes (Otto et al., 2004). Non-selective autophagy consists of the engulfment of bulk cytosolic material in a doublemembrane compartment called the autophagosome. Upon fusion with lysosomes, autolysosomes degrade, and recycle their content, enabling cell survival under starvation. One of the major regulators of autophagy is the target of rapamycin complex 1 (TORC1). Under nutrient repletion, TORC1 downregulates autophagy by repressing the expression of autophagy genes and/or by phosphorylation and inhibition of proteins involved in autophagosome formation (reviewed in Nakatogawa, 2015). In D. discoideum, the TORC1 complex consists of the TOR serine/threonine kinase and two TOR activators, Lst8, and Raptor (Wullschleger et al., 2006;Rosel et al., 2012). Like in mammals, starvation decreases TORC1 activity, leading to induction of autophagy (King et al., 2011;Rosel et al., 2012).
Since many of the well-studied proteins known to participate in mammalian autophagy are conserved in this amoeba, autophagy can be easily monitored in D. discoideum. For instance, like in mammals, the activity of D. discoideum TORC1 can be analyzed by monitoring the phosphorylation state of Raptor and the TORC1 effector protein 4E-BP1 by immunoblotting (Rosel et al., 2012;Cardenal-Muñoz et al., 2017). In addition, in D. discoideum also the changes in autophagosome formation can be distinguished from changes in autophagic degradation by the differential response of Atg18 [WD-repeat protein interacting with phosphoinositides (WIPI) in mammals] and Atg8a (LC3/GABARAP family proteins in mammals) . Accordingly, during induction of autophagosome formation Atg8 and Atg18 translocate from the cytosol to membranes of nascent and elongating phagophores (the yet unclosed autophagic double membrane compartment). Immediately after closure of the autophagosome, Atg8 stays in both inner and outer membranes, while Atg18 dissociates. Upon fusion of the autophagosome with lysosomes, the hydrolytic enzymes delivered from the lysosome not only degrade the autophagic cargo but also the inner membrane of the autolysosome and its associated Atg8, while the Atg8 on the outside is recycled to the cytosol (reviewed in Dominguez-Martin et al., 2017).
In recent years, D. discoideum has also become an interesting model to study the molecular mechanisms regulating xenophagy (reviewed in Mesquita et al., 2016). Xenophagy is a selective autophagy pathway specifically recognizing and digesting intracellular pathogens. It relies on selective receptors that recruit the cargo to be degraded [i.e., bacteria or remnants of their phagosome decorated with "eat-me" signals such as ubiquitin (Ub) or galectins in mammals] to autophagic membranes (Thurston et al., 2009(Thurston et al., , 2016Boyle and Randow, 2013;Noad et al., 2017). Several xenophagy receptors have been described in mammalian cells, but only one, p62, has been identified or studied so far in D. discoideum (Calvo-Garrido and Escalante, 2010;Gerstenmaier et al., 2015;Lampe et al., 2015;Cardenal-Muñoz et al., 2017). In addition, although D. discoideum lacks galectins, which recognize bacterial or host membrane glycans, another family of cytosolic lectins in this amoeba, the discoidins (Dsc), share many molecular and biological characteristics with galectins. Dsc are highly expressed upon starvation (Rosen et al., 1973), bind to self-glycans during development Eitle et al., 1993) and can be secreted to the extracellular medium . Interestingly, Dsc bind to sonicated extracts of Escherichia coli and Klebsiella aerogenes (Cooper et al., 1983), as well as to glutaraldehyde-fixed K. aerogenes bacteria (Madley and Hames, 1981). Recent studies have revealed that Dsc are enriched on compartments containing virulent Legionella pneumophila strains (i.e., Corby and JR32) but not the less virulent L. hackelieae, suggesting a role in innate immunity (Shevchuk et al., 2009;Urwyler et al., 2009). Moreover, one ortholog of the human TRIM37, a member of the Tripartite Motif protein superfamily (TRIMs) of autophagy receptors and regulators that often interact with galectins (Mandell et al., 2014;Kimura et al., 2015;Chauhan et al., 2016), has been identified in D. discoideum (Dunn et al., 2018). It will be interesting to uncover whether or not Dsc, TRIM-like protein and p62 cooperate in the immunity of D. discoideum against pathogenic bacteria.

The Bacterial Pathogen Mycobacterium marinum
Mycobacterium is the only genus of the Mycobacteriaceae family. It comprises more than 150 different species, among them two of the most important human pathogens, M. leprae and M. tuberculosis (Mtb), causing leprosy and tuberculosis, respectively. The latter represents the most severe bacterial diseases, responsible for 1.8 million deaths worldwide in 2015 (World Health Organization, 2015, http://www.who.int/ tb/publications/global_report). One of the important features of Mtb is its capacity to persist inside its host in a nonreplicating state with low metabolic activity. This phenomenon is called dormancy (or latency) and affects 2-3 billion people (i.e., one third of the world population). Latent tuberculosis is non-transmissible and presents no clinical manifestation. However, the disease reactivates in 5-10% of the cases usually as a consequence of immunosuppression, becoming contagious. Therefore, providing treatments to prevent the reactivation of latent tuberculosis is an important concern in the control of this disease (World Health Organization, 2015).
Mtb belongs to the group of genetically related mycobacteria, the Mtb complex, that cause tuberculosis in human or other animals. However, another group of human pathogens emerging as a major public health problem is the Non-Tuberculous Mycobacteria (NTM) or Mycobacteria Other Than Tubercle bacilli (MOTT), which includes among others the species M. ulcerans, M. abscessus, M. avium-intracellulare complex (MAC) and M. marinum (Ryu et al., 2016). In contrast to Mtb, with no environmental reservoir demonstrated so far, M. marinum is an intracellular pathogenic mycobacterium with ubiquitous distribution. It can be found in aquatic environments, and it affects a wide range of freshwater and marine vertebrates including fish, amphibians, and turtles (Decostere et al., 2004). During infection of these animals, M. marinum forms granulomatous lesions highly similar to the ones produced by Mtb in humans. In addition, M. marinum also develops a latent disease in the zebrafish model (Parikka et al., 2012). Very importantly, and reflecting their genetic proximity, M. marinum is also an opportunistic pathogen of humans, but due to its optimal growth at 30 • C, M. marinum infections are mostly localized and restricted to the skin and extremities. These lesions are usually painless and can be treated with anti-Mtb antibiotics when healing is not spontaneous (Aubry et al., 2017). At the cellular level, the infection course of M. marinum is also very similar to that of Mtb (reviewed in Ramakrishnan, 2004;Tobin and Ramakrishnan, 2008). Both bacteria avoid degradation within the host cells by arresting the maturation of their phagosomes and, contrary to the non-tuberculous M. avium, they escape to the host cytosol prior to dissemination (Stamm et al., 2003;Hagedorn et al., 2009;Simeone et al., 2015;Jamwal et al., 2016).
The genome of the M. marinum strain M was sequenced and annotated in 2008 by Stinear et al. (2008). Whole genome comparisons revealed that M. marinum is the ancestor of M. ulcerans and that both species are the closest relatives of the Mtb complex (Stinear et al., 2000(Stinear et al., , 2008. Among these species, M. marinum has the largest genome with 6.6 Mb, while the genome size of M. ulcerans and Mtb is 5.8 and 4.4 Mb, respectively. The genome of M. marinum shares 85% nucleotide identity with Mtb and 97% with M. ulcerans. Although phylogenetically and genetically close, Mtb and M. marinum present some differences such as the cell wall composition (Daffe et al., 1991;Tonjum et al., 1998), the abundance of prolineglutamic acid polymorphic guanine-cytosine-rich sequence (PE-PGRS) and proline-proline-glutamic acid (PPE) proteins (Cole et al., 1998;Bottai et al., 2012;Ates et al., 2015), and the synthesis of carotenoid pigments, which only occurs in M. marinum (Stinear et al., 2008).
Nevertheless, and more importantly, many virulence factors are conserved between M. marinum and Mtb. One example are the genes involved in dormancy (Dos-regulon) (Gerasimova et al., 2011). In addition, M. marinum and Mtb secrete virulence effector proteins through their complex cell walls thanks to very sophisticated secretion systems. In this context, the two mycobacteria possess five type VII secretion systems (T7SS) called ESX1-5 (reviewed in Groschel et al., 2016), in reference to the first secreted factor identified ESAT6 (Sorensen et al., 1995). The T7SS are not exclusive to mycobacteria and have been found in other bacterial genera such as Streptomyces and Bacillus (Gey Van Pittius et al., 2001;Pallen, 2002;Unnikrishnan et al., 2017). Despite their diversity, the T7SS share two common characteristics: the presence of FtsK/SpoIIIE transmembrane proteins and the secretion of small peptides of around 100 amino acids with the conserved motif Trp-X-Gly (WXG) (Pallen, 2002). The WXG motif is involved in the generation of a helixturn-helix structure and is present in two of the most studied T7SS secreted peptides, the 6 kDa early secretory antigenic target ESAT6 and the 10 kDa culture filtrate protein CFP10 (Renshaw et al., 2005). Each ESX system is encoded in its respective ESX cluster or locus, which also contains the genes for secreted peptides and other accessory proteins (Bitter et al., 2009). The ESX loci do not complement each other (reviewed in Abdallah et al., 2007), which suggests that they are involved in different functions. Only three of them have been shown to play a role in pathogenesis: ESX-1, 3, and 5. Whereas ESX-3 is required for iron and zinc acquisition (Serafini et al., 2009(Serafini et al., , 2013Siegrist et al., 2009Siegrist et al., , 2014, ESX-5 is involved in nutrient intake (Ates et al., 2015), cell wall integrity and secretion of PE and PPE proteins. The ESX-1 locus is the most studied, since it was shown to be very important for mycobacterial virulence (Pym et al., 2002). The attenuated M. microti and M. bovis BCG bacille Calmette-Guérin (BCG), used for vaccines, lack a partially overlapping region of the ESX-1 locus, which was called Region of Difference 1 (RD1). This region comprises two genes that encode the above mentioned secreted peptides ESAT6 and CFP10. Early studies using transposon libraries and purified proteins identified ESX-1, and in particular ESAT6 and CFP10, as membranolytic factors in both Mtb and M. marinum (Hsu et al., 2003;Gao et al., 2004). Importantly, the deletion of M. marinum RD1 can be complemented with homologous genes from Mtb, thus confirming the functional similarity of the RD1 regions in both species (Gao et al., 2004). Even though a 3.7 kb portion of M. marinum RD1 is partially duplicated within its genome (Stinear et al., 2008), deletion of RD1 prevents M. marinum escape from its phagosome as occurs with Mtb, suggesting that the duplicated fragment of M. marinum is not functional (Simeone et al., 2015;Cardenal-Muñoz et al., 2017).
These pathologic, phylogenetic, and genomic aspects make of M. marinum an interesting model to study mycobacterial infections. Importantly, M. marinum also has advantages for laboratory work. Despite the fact that this species is considered a slow growing mycobacterium, M. marinum grows faster than Mtb, with a doubling time of 6-8 h vs. the 20-24 h of Mtb. Furthermore, M marinum is a level 2 biosafety organism, and numerous molecular tools are available and transposable from Mtb to M. marinum, and vice versa. human diseases induced by diverse bacteria genera such as Francisella, Legionella, Salmonella, or Mycobacterium (Bozzaro and Eichinger, 2011;Weber and Hilbi, 2014;Brenz et al., 2017;Cardenal-Muñoz et al., 2017;Steiner et al., 2017). In both D. discoideum and human phagocytes, the main cellautonomous defenses rely on the highly conserved microbicidal machinery that arms the phagosome (Boulais et al., 2010). This includes a number of genes and pathways involved in bacteria recognition, phagocytic uptake, actin dynamics, phagosome identity, acidification and maturation, lysosomal degradation, production of Reactive Oxygen Species (ROS), metal poisoning and nutrition immunity (recently reviewed in Dunn et al., 2018).
The ability of M. marinum to replicate within D. discoideum was first described in 2003 (Solomon et al., 2003). Solomon and collaborators monitored monolayers of D. discoideum infected with GFP-producing M. marinum at 25.5 • C, a temperature optimal for the joint growth of host and bacterium. For these long-term infection assays, and to avoid rapid host cell death, D. discoideum needs to be infected at low multiplicity of infection (MOI ≤1-10). After an initial incubation without antibiotics, to allow the phagocytosis of M. marinum, streptomycin was added to the infection sample to impede the growth of extracellular bacteria. In this first approach, the increased growth of M. marinum was assessed by direct observation, but other quantitative methods such as luminescence recording, fluorescence-activated cell sorting (FACS) and colony-forming units (CFUs) counting have confirmed the intracellular proliferation of this bacterium within D. discoideum, both in static and shaking cultures (Solomon et al., 2003;Hagedorn and Soldati, 2007;Arafah et al., 2013). Other antibiotics such as amikacin or a mix of penicillin and streptomycin can be used to kill the extracellular M. marinum when performing infection assays in D. discoideum (Arafah et al., 2013). The presence of extracellular bacteria in the infection sample can be tested by incubating live cells with an anti-M. marinum serum (anti-mar) and performing immunofluorescence assays (Supplementary Table 1); Only the extracellular bacteria, and not those inside the amoeba, are accessible to the serum (Hagedorn et al., 2009).

M. MARINUM INFECTION COURSE IN D. DISCOIDEUM
The infection cycle of M. marinum in D. discoideum starts with the phagocytosis of the bacterium by the amoeba, and finishes with the bacterial egress from the cell (Figure 1 and Table 1). Three different growth stages of M. marinum can be distinguished during the infection cycle: (i) an initial phase during the first 12 h post-infection (hpi), when M. marinum actively manipulates the D. discoideum phagocytic pathway to convert the phagosome where it resides into a replicationpermissive niche [mycobacteria-containing vacuole (MCV)], (ii) an enhanced proliferation phase between 12 and 37 hpi, and (iii) a phase of arrested proliferation and/or decreased bacterial load due to bacterial death or release after 37 hpi (Hagedorn and Soldati, 2007). This infection cycle might be impacted by the number of bacteria initially taken up. When a single amoeba internalizes a clump of bacteria, the M. marinum clump either cannot initiate its intracellular growth, remaining in a dormant or undigested state as occurs in zebrafish, or causes cytotoxicity to D. discoideum (Solomon et al., 2003;Parikka et al., 2012).
Early M. marinum Infection Phase (∼0-12 hpi) During uptake, M. marinum is surrounded by a circular ruffle and ingested into a newly produced phagosome (Figure 2). The engulfment of the bacterium by the tight phagocytic cup has been observed by microscopy of live and fixed cells, using markers of the actin cytoskeleton of D. discoideum (Hagedorn et al., 2009). In addition, phagocytosis of GFP-producing or Alexa 488 hydrazide-labeled M. marinum has been quantitatively monitored by flow cytometry (Sattler et al., 2013). It is important to notice that the labeling of surface amines with Alexa Fluor 488tetrafluorophenyl ester (Alexa Fluor 488-TFP), normally used for other bacteria such as Klebsiella pneumoniae or Salmonella enterica, is not suitable for M. marinum, presumably due to its extremely hydrophobic and poor-in-proteins cell wall (Sattler et al., 2013).
Ten minutes after phagocytosis, the early endosomal marker Rab5a is withdrawn from the M. marinum phagosome, compared to 3 min for the non-pathogenic M. smegmatis (Barisch et al., 2015a). Subsequently, the late endosome and lysosome marker Rab7a is detected at the early MCV (Cardenal-Muñoz et al., 2017), as occurs in macrophages (Lerena and Colombo, 2011). However, M. marinum rapidly bypasses the phagocytic pathway and blocks the maturation of its compartment. Thus, late maturation markers such as vacuolin and cathepsin D are absent from the MCV from 0.3 to 6 hpi, and the vacuolar H + -ATPase (vATPase) is almost undetectable on the MCV at the end of this infection phase (12 hpi, Figure 2) (Hagedorn and Soldati, 2007). In this context, and contrary to M. smegmatis, only a small percentage of the phagocytosed M. marinum are killed by D. discoideum within the first hours (Hagedorn and Soldati, 2007), even in conditions of mild starvation that stimulate autophagy (i.e., phosphate buffer supplemented with 5% nutrient-rich medium) (Delince et al., 2016).
It has been shown that actin polymerization is required for the efficient maturation arrest of the M. marinum MCV during early infection of D. discoideum (Kolonko et al., 2014). Thus, Arp2/3 and WASH (Wiskott-Aldrich syndrome protein and Scar Homolog), two complexes involved in actin nucleation and polymerization on endosomal membranes (Duleh and Welch, 2010), can be observed in increasing association with the MCV from 0 to 6 hpi (Figure 2). This is concomitant with the presence of a patchy F-actin coat covering the MCV and with the absence of VatA, a peripheral subunit of the vATPase, on the same MCV at 6 hpi (Kolonko et al., 2014). On the contrary, amoebae either treated with the inhibitor of actin polymerization latrunculin A (LatA), producing a fusion construct between VacA and the actin-depolymerizing factor cofilin (VMC), or lacking the WASH complex (wshA-) harbor actin-negative (1) and rapidly manipulates its phagocytic pathway to reside within a replicative niche (2). The ESX-1 secretion system of M. marinum perforates the MCV (3), which induces the recruitment of phagophores for membrane repair (4). M. marinum proliferates within its MCV (5), which finally breaks (6) and release mycobacteria to D. discoideum cytosol (7). Bacteria continue growing in the host cytosol (8) prior to egress by ejection (9.1), lytic death (9.2), or exocytosis (9.3). Phagophores are also recruited to the site of ejection for plasma membrane repair. Recapture into lysosome-like compartments may precede late exocytosis. However, this has not been shown yet in D. discoideum. Early exocytosis (9.3) can be induced upon starvation. Intercellular dissemination occurs after M. marinum release from the amoeba (10).
MCVs which accumulate VatA and VatM (a transmembrane subunit of the vATPase), while losing the endocytic marker p80. Interestingly, p80 is not lost from the phagosome containing M. smegmatis and the avirulent mutant M. marinum L1D (Kolonko et al., 2014) (Supplementary Table 2). In agreement with the LatA-dependent accumulation of the vATPase on MCVs, LatA treatment also leads to acidification of the bacterial compartment, resulting in decreased bacteria viability (Kolonko et al., 2014). Interestingly, actin polymerization also prevents the acidification of the M. marinum and Mtb phagosomes in murine phagocytes (Kolonko et al., 2014). In addition to Arp2/3 and WASH, the D. discoideum homolog of mammalian flotillin, vacuolin B (VacB), is also essential for the establishment of the MCV, assisting M. marinum in the retrieval of the vATPase from its niche. On the contrary, the small GTPase RacH of D. discoideum contributes to the maturation of the MCV into phagolysosomes, reducing the intracellular bacterial load (Hagedorn and Soldati, 2007) (Supplementary Table 3).
It has to be noticed that M. marinum does not interfere with the contractile vacuole of D. discoideum (Kolonko et al., 2014). This osmoregulatory organelle consists of interconnected vesicles, cisternae and tubules that fuse with the plasma membrane to expel the excess of water ingested during macropinocytosis and maintain cell volume (Patterson, 1980). The contractile vacuole accumulates Rab11 GTPase, vATPase, and calmodulin (Du et al., 2008), but the latter does not localize to the MCV (Kolonko et al., 2014;Cardenal-Muñoz et al., 2017). In addition, tetramethylrhodamine isothiocyanate (TRITC)-dextran, which labels the endolysosomal system but not the contractile vacuole (Clarke et al., 2002), can be found inside the MCV (Gerstenmaier et al., 2015).
The D. discoideum ammonium transporter AmtA is a useful marker to determine whether or not mycobacteria are inside a compartment, since its fluorescent versions ubiquitously localize to all the endosomes and the MCV (Barisch et al., 2015b;Barisch and Soldati, 2017b) (Supplementary Table 4). Methods based on fluorescence resonance energy transfer (FRET), such as FRETbased microscopy or flow cytometry, have been performed in macrophages infected with Mtb and M. marinum to detect the mycobacteria-mediated phagosome disruption (Simeone et al., , 2015Acosta et al., 2014).
Late M. marinum Infection Phase (∼37-72 hpi) After 37 hpi, when M. marinum has escaped its MCV, the intracellular bacterial load stabilizes and even decreases (Hagedorn and Soldati, 2007). This might be the result of the convergence between recapture and killing of cytosolic bacteria in lytic compartments, and the release of the bacteria to the extracellular medium or neighboring cells (Figure 1). Events of recapture have not been reported so far in D. discoideum, but in macrophages cytosolic M. marinum is ubiquitinated and recaptured into non-autophagic double membrane compartments (LAMP-1-positive, LC3/Atg8-negative) (Collins et al., 2009). It was proposed that ESCRT, a protein machinery that sorts ubiquitinated cargo to degradation into intraluminal vesicles (Schmidt and Teis, 2012), would deliver these bacteria to multivesicular bodies (MVB) and finally to lysosomes, but the bactericidal activity of these putative recapture compartments was not established (Collins et al., 2009).
Regarding release, and as a consequence of the wide range of strategies used by bacteria to survive and replicate within their hosts, multiple egress strategies have emerged in the course of evolution (Friedrich et al., 2012) (Figure 1 and Table 1). In some cases, egress implies the disruption of one or several host membranes, which may lead to the host cell death. Mycobacteria modulate the death of their host cells, inducing apoptotic or necrotic processes with different biological implications. These processes vary depending on the mycobacterial species, host cell types, and infection stage (reviewed in Aguilo et al., 2013;Srinivasan et al., 2014). In this context, many studies have associated ESX-1 activity and mycobacterial vacuolar escape to an increased host cell cytotoxicity and death (Hsu et al., 2003;Derrick and Morris, 2007;Kaku et al., 2007;Kinhikar et al., 2010;Simeone et al., 2012;Augenstreich et al., 2017). Thus, incubation with purified ESAT-6, as well as infection with wildtype (wt) Mtb and M. marinum but not with RD1 mutant bacteria, lyse epithelial and red blood cells and macrophages (Hsu et al., 2003;Gao et al., 2004). This is of special relevance since mycobacterial egress from its host might influence spread of the infection in a tissue. Similarly, by using a microfluidic device (the InfectChip) to microscopically monitor single-cell long-term infection in D. discoideum, it has been shown that a quarter of the phagocytosed M. marinum induces death and lysis of the amoebae, contributing to bacteria release (Delince et al., 2016).
Alternative non-lytic mechanisms of bacterial egress also occur during mycobacterial infection in D. discoideum. For instance, release of M. marinum within the first hours of infection, when most of the bacilli are still intravacuolar, thereby suggesting an exocytic process, can be induced by mild starvation of D. discoideum (Delince et al., 2016). This phenomenon is likely akin to the known massive exocytosis of lysosomal enzymes that D. discoideum undergoes upon starvation (Smith et al., 2010). In addition, bacterial ejection occurs at later time points (or earlier if the MOI is increased to 50-100, Gerstenmaier et al., 2015) once mycobacteria have escaped the MCV and accessed the host cytosol (Hagedorn et al., 2009). During ejection, M. marinum and Mtb, but not the ESX-1-deficient M. avium, use the actin cytoskeleton of D. discoideum to form an actin-based short barrel structure formed at the plasma membrane, called the ejectosome, through which they exit the cell (Hagedorn et al., 2009). In this context, some D. discoideum actin-binding proteins such as coronin and myosin IB are enriched at the ejectosome (Hagedorn et al., 2009) (Supplementary Table 4). In addition, ejection cannot occur in amoebae lacking the actin polymerization complex WASH (Kolonko et al., 2014). Strikingly, other markers of the actin cytoskeleton such as the actin nucleator complex Arp2/3 and myosin II are not enriched at the ejectosome, suggesting that this structure may assemble de novo during ejection (Hagedorn et al., 2009). It has been proposed that the D. discoideum plasma membrane reseals at the posterior of the bacterium after ejection, preventing the amoebal lysis or leakage (Gerstenmaier et al., 2015) (section D. discoideum Plasma Membrane Damage and Sealing at Later Time Points). Exclusion of DAPI (a membraneimpermeant DNA binding dye) from the nuclei of live amoebae during ejection demonstrates that M. marnium does not kill or induce leakage of its host while egressing by this mean (Hagedorn et al., 2009).
Exocytosis and ejection can be discriminated by fluorescence microscopy with various D. discoideum markers. For instance, only during exocytosis the intracellular pole of the egressing bacterium is positive for endosomal and MCV markers such as p80 and vacuolins. In addition, the extracellular part of the exocytosed bacterium is free of any plasma membrane marker (Hagedorn et al., 2009;Gerstenmaier et al., 2015). On the contrary, upon ejection, extracellular Mtb and M. marinum are wrapped in plasma membrane remnants positive for p80 or PM4C4 (Supplementary Table 4). Clumps of bacteria can be exocytosed, but ejectosomes, which are often observed to form clusters, only release individual bacteria (Hagedorn et al., 2009).
Once bacteria exit the cell, neighboring cells might become new hosts. Within macrophages, a subpopulation of cytosolic M. marinum sheds its ubiquitinated cell wall possibly to avoid recapture into lysosomes. These bacteria induce the polymerization of actin tails and spread via actin-dense filopodia that can be caught by other cells (Stamm et al., 2003;Collins et al., 2009;Hagedorn et al., 2009). However, and presumably due to the high rates of actin turnover, cytosolic M. marinum does not form persistent actin tails in D. discoideum, but events of cell-to-cell transmission synchronized with actin-based ejection are observed (Hagedorn et al., 2009). D. discoideum RacH, mentioned before to contribute to the maturation of the MCV, also assists M. marinum in its release from the amoeba and the subsequent intercellular dissemination. Accordingly, ejection of M. marinum is not observed in racH-cells (Hagedorn et al., 2009), and the population of infected racH-amoebae almost fully retain all bacteria until 37 hpi (Hagedorn and Soldati, 2007;Hagedorn et al., 2009) (Supplementary Table 3). In agreement with these results, quantitative dissemination assays have confirmed the positive role of RacH in M. marinum cell-tocell spread. In these fluorescence microscopy assays, an unlabeled donor D. discoideum strain (wt or racH-) was infected with red fluorescent M. marinum, and a wt acceptor strain expressing GFP was added to the infection after 12 h. Transmission from racHinfected cells to wt cells was reduced over eight-fold (Hagedorn et al., 2009).

M. MARINUM INDUCES MEMBRANE DAMAGE THAT IS COUNTERACTED BY D. DISCOIDEUM
Rupture of host membranes by bacterial pathogens is achieved by the well-orchestrated secretion of lipases, proteases and poreforming toxins. This is the case of vacuolar bacteria such as Mycobacterium, Shigella, Listeria, or Salmonella, which escape to the host cytosol via the disruption of their vacuole. Although the membranolytic activity of the mycobacteria ESX-1 secretion system is widely accepted, it has been proposed that ESX-1dependent membrane disruptions do not require ESAT-6 activity in physiological conditions in macrophages (Conrad et al., 2017). Instead, the perforations would occur upon direct contact of Mtb and M. marinum ESX-1 with host MCV membranes (Conrad et al., 2017), which would confine the membranolytic activity to the bacterial poles, where the ESX-1 is enriched . Incompatible with this proposed model, M. marinum mutants specifically lacking only ESAT-6 and CFP10 ( CE) in an otherwise intact RD1 locus are defective in hemolysis, cytolysis, and cytotoxicity (Gao et al., 2004). Similarly, the virulent Mtb H37Rv strain with a 12 amino acids deletion in ESAT-6 shows impaired membranolytic activity in the THP-1 human macrophages (Houben et al., 2012). Regarding CFP10, it has been shown that the protein dissociates from ESAT-6 in acidic conditions similar to those encountered within the phagosome, and it does not disrupt artificial liposomes on its own (de Jonge et al., 2007). As a consequence, CFP10 has been proposed to act as a chaperone of ESAT-6, conferring stability or protection against proteases. Despite the fact that the structures of ESAT-6 and CFP10 have been solved (Renshaw et al., 2005), it is not yet well-understood how ESAT-6 induces membrane disruptions.

Perforation and Repair of the MCV Membrane at Early Time Points
M. marinum resides inside a phagosome that shields it against the host intracellular immune responses but limits nutrient access. Maybe as a reaction, the bacterium creates a porous vacuole which might appear ideal to obtain unrestricted access to nutrients. Indeed, very early during infection of D. discoideum, M. marinum induces ESX-1-dependent damages in the MCV membrane, which results in the decoration of cytosol-exposed bacteria with Ub already at 1.5 hpi (Cardenal-Muñoz et al., 2017). This MCV membrane damage leads to a well-orchestrated early autophagic response consisting in: (i) transient downregulation of TORC1 and upregulation of transcription of autophagy genes such as atg1, the two paralogs atg8a and atg8b, and p62, (ii) enhanced formation of autophagosomes, and (iii) recruitment of the xenophagy machinery (Ub, GFP-p62, and Atg8a) to the damaged MCV. These responses to MCV damage are significantly reduced or even abolished when M. marinum lacks the RD1 locus and, more specifically, ESAT-6 and CFP-10 (Cardenal-Muñoz et al., 2017).
However, contrary to the increased bacterial killing expected from an enhanced xenophagic response, M. marinum survives and proliferates within D. discoideum by blocking the autophagic flux and its subsequent degradation within autolysosomes (Cardenal-Muñoz et al., 2017). Once again, the main player involved in this blockage is the mycobacterial ESX-1 secretion system, also shown to block the autophagic flux in human cells infected with Mtb (Romagnoli et al., 2012). Along these lines, even though 40-50% of the early M. marinum MCVs are positive for the autophagosomal marker Atg8a and for markers of acidic compartments [vATPase LysoSensor Green and neutral red (NR) (Hagedorn and Soldati, 2007;Kolonko et al., 2014;Cardenal-Muñoz et al., 2017)], and assuming that the presence of Rab7 on the MCV (Cardenal-Muñoz et al., 2017) should ensure fusion with late endosomes and lysosomes (Zhang et al., 2009), the MCVs containing wt bacteria are devoid of lysosomal enzymes and degradative activity, as demonstrated by the lack of cathepsin D and proteolysis-dependent DQ-Green BSA fluorescence (Hagedorn and Soldati, 2007;Cardenal-Muñoz et al., 2017). This suggests that the ESX-1-induced membrane damage make MCVs leaky for hydrolytic enzymes. This MCV leakage, together with the retrieval of the vATPase, and possibly hydrolases, by WASH-and Arp2/3-dependent polymerization of actin[see section Early M. marinum Infection Phase (∼0-12 hpi) and Figure 2], would affect both autophagic flux and killing of mycobacteria. Interestingly, depletion of hydrolases from the acidic Salmonella-containing compartment (SCV) has already been described, although the mechanism is different. Salmonella subverts the Rab9-dependent retrograde trafficking of mannose-6-phosphate receptors in human cells, which results in fusion of the SCV with lysosomes that are reduced in hydrolytic enzymes (McGourty et al., 2012).
The membrane damage inflicted by Salmonella induces an amino acid starvation response in the host cell that inhibits TOR and induces autophagy and membrane repair, leading to the normalization of the amino acid levels and the reactivation of TOR at the surface of SCVs (Tattoli et al., 2012a,b;Kreibich et al., 2015). Interestingly, it has been recently shown that mammalian TRIM16, a member of the TRIM family, recognizes the endomembrane perturbations induced by Mtb ESX-1 and affects TORC1 activity, inducing autophagy to protect macrophages from damage (Chauhan et al., 2016). The role of autophagy in mediating membrane repair during D. discoideum infection with M. marinum has also been recently proposed (Cardenal-Muñoz et al., 2017). After downregulation of TORC1 and induction of the autophagic response mentioned above, D. discoideum Lst8 (a TORC1 component) and two TORC1 activators, the Rheb GTPase and Lamtor1 [a member of the guanine nucleotide exchange factor (GEF) Ragulator complex] localize to damaged MCVs coinciding with reactivation of TORC1 at 7 hpi (Cardenal-Muñoz et al., 2017) (Figure 2). In addition, EM inspection of infected amoebae lacking the major autophagy initiator Atg1 (ULK in humans) (Mesquita et al., 2015) reveals that most bacteria escape very early to the cytosol (1 hpi), followed by a striking increased load of highly ubiquitinated cytosolic bacteria that cannot be degraded by xenophagy. This suggests that the autophagy pathway of D. discoideum is required for the repair of the MCV membranes, preventing the escape of mycobacteria to the cytosol. It remains unclear why M. marinum accelerates its growth only after 24 hpi in the atg1-amoebae (Cardenal-Muñoz et al., 2017). One plausible explanation is that autophagy would not only recognize damage and repair the MCV, but it could also provide M. marinum with cytoplasmic nutrients such as membranes and lipid droplets (LDs) by the fusion of autophagosomes with the MCV (see section M. marinum Exploits D. discoideum Lipids).

D. discoideum Plasma Membrane Damage and Sealing at Later Time Points
Despite the function of autophagy in membrane repair during early infection, most M. marinum bacilli are able to break completely their MCV and finally access the D. discoideum cytosol (Hagedorn and Soldati, 2007). As previously described, M. marinum perforates the plasma membrane of D. discoideum to egress through an actin-based ejectosome (Hagedorn et al., 2009). The ESX1 secretion system, and more specifically secreted ESAT-6, are required for ejection of the bacteria (Hagedorn et al., 2009). This is demonstrated by restoration of ejection of the RD1 mutant by ectopic expression of M. marinum ESAT-6 in D. discoideum (Hagedorn et al., 2009). Despite this perforation, the integrity of the D. discoideum plasma membrane is continuously maintained during and after ejection by the autophagy machinery (Figure 1). This is necessary to avoid lytic cell death and to improve cell-to-cell transmission of the bacteria (Gerstenmaier et al., 2015). Accordingly, the autophagosomal markers Atg8, GFP-p62, and Ub are present at a pocket formed around the intracellular distal pole of ejecting M. marinum (Gerstenmaier et al., 2015). On the contrary, GFP-Atg18 and GFP-2xFYVE, two reporters of expanding phagophores (Calvo-Garrido et al., 2014;Mesquita et al., 2016), are less or rarely present, respectively (Gerstenmaier et al., 2015). This, together with the fact that the transcription levels of autophagy genes remain unchanged at these late stages of infection (Cardenal-Muñoz et al., 2017), suggest that M. marinum exploits preformed autophagosomes during its ejection.
The core autophagy machinery of D. discoideum (Atg1, Atg5, Atg6, and Atg7) is required for the recruitment of Atg8 to the distal pole of ejecting M. marinum (Supplementary  Table 3). Strikingly, atg1-cells can form the actin-based structure of ejectosomes, but these are not functional for cell-to-cell transmission (Gerstenmaier et al., 2015). In addition, the selective receptor p62 is not required for the recruitment of Atg8 to the ejecting bacterium (Gerstenmaier et al., 2015). This, together with the fact that M. marinum grows normally within p62-cells (Cardenal-Muñoz et al., 2017), suggest that alternative autophagy receptors might exist in D. discoideum. Interesingly, the ESX-1 secretion system, which induces the recruitment of autophagic markers to bacteria escaping the MCV (Cardenal-Muñoz et al., 2017), does not play a role in the localization of Atg8 to bacteria during ejection (Gerstenmaier et al., 2015). This was shown by co-infecting cells with wt and either M. marinum RD1 or the non-pathogenic M. smegmatis (whose ESAT-6 is inactive in membrane disruption). In these cells both bacteria form ejectosomes and recruit Atg8 to their distal pole (Gerstenmaier et al., 2015).

M. MARINUM EXPLOITS D. DISCOIDEUM LIPIDS
Inside their hosts, pathogens are restricted to a limited supply of nutrients. For instance, to drive proliferation, intracellular bacteria need to exploit a suitable energy source (Abu Abu Kwaik and Bumann, 2015). In recent years, it has become evident that Mycobacterium spp. access host lipids (Peyron et al., 2008;Caire-Brandli et al., 2014) to gain energy via β-oxidation and the glyoxylate shunt (McKinney et al., 2000;Munoz-Elias and McKinney, 2005). The importance of lipids for Mtb, but also for other Mycobacterium spp., is supported by the high number of genes (more than 250) harbored in the Mtb genome and required for lipid and fatty acid (FA) metabolism (Cole et al., 1998). The encoded proteins have diverse roles in cell wall synthesis, formation of intracytosolic lipid inclusions (ILIs) and production of energy (reviewed in Dumas and Haanappel, 2017). Strikingly, 100 of these genes exclusively encode proteins involved in the five reactions of β-oxidation (Cole et al., 1998). Interestingly, the chromosome of M. marinum is 50% larger than the one of Mtb (6.6 million base pairs compared to 4.4 million), which is also reflected by a higher number of genes involved in lipid metabolism (Stinear et al., 2008). For instance, the genome of M. marinum encodes 32 fadD acyl-CoA synthase paralog, compared with 23 in Mtb. This larger genome likely reflects the capacity of M. marinum to infect and produce a tuberculosis-like disease in a larger range of environmental hosts compared to Mtb (Stinear et al., 2000).

Lipid Droplets Are Recruited to the MCV during the Early Infection Phase
Despite the increasing evidence pointing to a central role of lipids during Mtb infection, the basic cellular mechanisms by which pathogenic mycobacteria exploit host lipids are still poorly understood. One of the major characteristics of tuberculosis is the appearance of foamy macrophages during chronic infection. The accumulation of LDs that leads to the foamy characteristics of the respective host cells was observed during Mtb infection (reviewed in Russell et al., 2009), but also in infections with M. bovis (D'Avila et al., 2006) and M. leprae (de Mattos et al., 2012). By using D. discoideum as a model for foamy macrophages, it was recently shown that host LDs move to the MCV within minutes after phagocytosis (Figure 3, step 1; Barisch et al., 2015b). The clustering of host LDs at the MCV can be monitored by live microscopy using Bodipy 493/503 as a neutral lipid marker for LDs that were induced by feeding cells with exogenous FAs prior to infection. We propose that host LDs might be translocated into the MCV by a process similar to lipophagy (Figure 3, step  1b), a selective autophagy pathway for LDs (reviewed in Schulze et al., 2017), or by a fusion-like process similar to that observed in a macrophage/M. avium infection model system (Caire-Brandli et al., 2014) (Figure 3, step 2). In analogy to LD-bilayer or LD-LD fusion (Thiam et al., 2013), the phospholipid monolayer of an LD might first fuse with the outer leaflet of the MCV membrane leading to the release of the hydrophobic core lipids into the membrane bilayer of the MCV. In a second step, a new LDlike structure might bud into the lumen of the MCV (for more detailed information, please see (Barisch and Soldati, 2017a).

Host Lipid Distribution at the M. marinum Proliferation Phase
At later infection stages (19 hpi), LD-like structures disappear, but neutral lipids and sterols, which can be detected by Bodipy  (Barisch and Soldati, 2017a); (2) LDs might be imported into the MCV by a mechanism similar to phagosomal fusion; (3) Neutral lipids (and sterols) accumulate within the MCV at late infection stages; (4) Host phospholipids are transferred to the MCV by membrane trafficking (Barisch and Soldati, 2017b); (5) Host triacylglycerols (TAGs) and phospholipids serve as fatty acid (FA) source for bacterial intracytosolic lipid inclusion (ILI) formation (Barisch et al., 2015b); (6) Dgat2-positive LDs aggregate at bacteria poles as soon as the MCV breaks, leading to the coalescence of LDs onto the M. marinum cell wall and the complete surrounding of the bacteria by Dgat2 (Barisch and Soldati, 2017a); (7) The D. discoideum homolog of perilipin reaches M. marinum from the cytosol and targets the bacterial cell wall with the help of amphipathic and hydrophobic domains (Barisch et al., 2015b;Barisch and Soldati, 2017b). 493/503 and filipin staining, are homogenously distributed within the MCV (Figure 3, step 3; Barisch et al., 2015b). Interestingly, these observations are in agreement with the fact that mycobacteria are one of the rare species able to catabolize host sterols and, more precisely, cholesterol (Pandey and Sassetti, 2008;Senaratne et al., 2008). By labeling D. discoideum with Topfluor-Lysophosphatidylcholine (Topfluor-LysoPC), a fluorescent phospholipid precursor, it was observed by live microscopy that host phospholipids accumulate at the membranes of the MCV, most likely by standard membrane trafficking (Figure 3, step 4; Barisch and Soldati, 2017b).
Strikingly, intravacuolar bacteria accumulate considerably more ILIs when D. discoideum cells are treated with exogenous FAs (Figure 3, step 5). Not only the FAs released from host triacylglycerols (TAGs) but also those from phospholipids are used by M. marinum for ILI formation (Figure 3, step 5; Barisch and Soldati, 2017b). This was shown by thin layer chromatography (TLC) using a D. discoideum dgat1/2 double mutant, which is depleted in both diacyglycerol Oacyltransferases Dgat1 and Dgat2. This mutant shuttles excessive lipids into phospholipids, and does not synthesize TAGs and generate LDs (Du et al., 2013;Barisch and Soldati, 2017b). Interestingly, by measuring bacteria growth with the help of high-content microscopy and a luminescence-based assay, it was observed that M. marinum growth was unaffected in the dgat1/2 double mutant (Barisch and Soldati, 2017b) leading to the conclusion that host phospholipids can successfully substitute for host TAGs. The transfer of host lipids to the intracellular pathogen can also be monitored by live-cell microscopy. To this end, host cells are labeled with Topfluor-LysoPC prior to infection. Strikingly, fluorescently-labeled host phospholipids become first enriched at the membrane of the MCV 10 min postinfection, and inside the bacteria at 21 hpi. Consequently, it was proposed that host lipids are transferred to the MCV by membrane trafficking and then further processed by bacterial or host phospholipases to facilitate uptake of FAs into the bacteria (Barisch and Soldati, 2017b).
Recently, the question whether Mtb has enzymes to successfully cleave/hydrolyse host lipids was addressed (Singh et al., 2017). The authors propose that host lipids are cleaved by Msh1, which is upregulated upon hypoxic conditions, and secreted into the host cytosol, where it hydrolyses TAGs. For M. marinum, it has been shown that the formation of ILIs depends on its type VII secretion system ESX-5, presumably thanks to the absorption of nutrients such as FAs through membrane porins or channels that are formed into the outer membrane of M. marinum by this ESX-5 apparatus (Ates et al., 2015). In the case of Mtb, FAs are transported into the bacteria via the mammalian cell entry 1 (Mce1) complex and LucA that interacts with subunits of the complex and coordinates its activity, and which is required for full virulence of Mtb in vivo (Nazarova et al., 2017).
The reason why intracellular mycobacteria accumulate ILIs during infection is so far poorly understood. One possibility is that these bacteria use ILIs as an energy source. For instance, it has been shown that in macrophages infected with M. avium and without any additional fat source, LDs are rapidly depleted from the macrophages, immediately followed by ILIs depletion from the pathogens (Caire-Brandli et al., 2014). Moreover, with the help of an in vitro dormancy model, it was proposed that M. bovis BCG uses the TAGs released from ILIs as an energy source during reactivation from dormancy (Low et al., 2009). In addition, when the synthesis of TAGs is inhibited in Drosophila melanogaster, the accumulation of host LDs induced by M. marinum infection is also prevented, and the mycobacteria-LDs association, as well as the number of intracellular viable M. marinum, are reduced (Pean et al., 2017). Furthermore, ILI-rich mycobacteria have been shown to be more tolerant to rifampicin, isoniazid, ethambutol, and ciprofloxacin (Hammond et al., 2015). Finally, an antioxidant role of LDs (or ILIs) was proposed recently (Bailey et al., 2015).
Importantly, host lipids might not only be used by these bacteria as an energy source but also as building blocks for cell wall lipids. The core of the mycobacterial cell wall consists of mycolic acids that are esterified to arabinogalactan (AGs) polysaccharides, which are in turn covalently attached to a peptidoglycan backbone (reviewed in Brennan, 2003;Jackson, 2014). This core structure serves as anchor for extractible "free lipids" such as PDIMs and phenolic glycolipids (PGLs). One unique feature of pathogenic Mycobacterium spp. is their antibiotic resistance that is conferred by the "mycomembrane, " the impermeable lipid bilayer that is formed by the mycolic acids and the free lipids (Jackson, 2014). During infection, PDIMs are involved in Mtb uptake (Astarie-Dequeker et al., 2009), block of phagosomal maturation and acidification arrest (Pethe et al., 2004;Stewart et al., 2005), in escape to the cytosol and egress from the host cells (Quigley et al., 2017). In addition, PDIMs are essential for the multiplication and persistence of Mtb in the lungs of infected mice (Cox et al., 1999), and Mtb strains inhibited in PDIM synthesis are more susceptible to killing by an early innate host response (Day et al., 2014). Corroborating this, it has been demonstrated that FAs released from host LDs are used by Mtb to synthesize PDIMs (Lee et al., 2013).
PDIMs are also implicated in the virulence of M. marinum. Like Mtb in macrophages, M. marinum synthesizes PDIMs (and TAGs) from FAs released from D. discoideum LDs and phospholipids (Barisch and Soldati, 2017b). To monitor lipid transfer from the host to the pathogen, D. discoideum lipids can be labeled by using fluorescent FAs such as Bodipy 558/568 C12 or Topfluor-LysoPC prior to infection (Barisch and Soldati, 2017b). The incorporation of the fluorescent label into D. discoideum and M. marinum lipids can be detected by TLC and fluorescent lipid species can be identified with the help of fluorescent standard lipids.
A transposon insertion in the M. marinum tesA gene, encoding a putative type II thioesterase, leads to an altered cell wall without PDIMs and PGLs and the subsequent lack of the permeability barrier (Alibaud et al., 2011). As a consequence, the M. marinum tesA mutant is attenuated in D. discoideum and in zebrafish embryos (Alibaud et al., 2011). Inhibitor and gene mutations altering the PDIMs and PGLs of M. marinum affect its virulence in D. discoideum and adult zebrafish, respectively (Yu et al., 2012;Ouertatani-Sakouhi et al., 2017). In addition, Cambier et al., elegantly described how Mtb and M. marinum preferentially recruit and infect permissive macrophages while microbicidal ones are evaded (Cambier et al., 2014). This bypass of the innate immune system was proposed to be the result of cell-surface-associated PDIMs hiding underlying pathogenassociated molecular patterns (PAMPs).

LD Proteins Target Cytosolic M. marinum at Late Infection Phases
At later infection stages (42 hpi), LD proteins such as the D. discoideum homolog of perilipin as well as Dgat2 localize to cytosolic M. marinum (Figure 3, steps 6 and 7; Barisch et al., 2015b;Barisch and Soldati, 2017b). Perilipin belongs to the class II of LD proteins, which bind the LD surface through hydrophobic and amphipatic domains, and circulates between LDs and the cytosol. During infection, perilipin is proposed to partition into the waxy mycobacterial cell wall in a similar way, via its amphipathic and hydrophobic regions (Barisch and Soldati, 2017a). Importantly, by using a luminescencebased assay to measure intracellular M. marinum growth, it became evident that in a D. discoideum perilipin (plnA) mutant, bacteria growth is inhibited starting from 20 hpi, the time at which M. marinum reaches the cytosol (Barisch et al., 2015b). Consequently, it was proposed that perilipin might exert a barrier function, protecting the M. marinum cell wall during the cytosolic phase of the infection from degradative host processes such as neutral lipases.
Dgat2 belongs to the class I of LD proteins, which bind the LD surface through an hydrophobic hairpin motif, and is exclusively localized at LDs. Consequently, Dgat2 is transferred to the waxy cell wall of M. marinum by coalescence with LDs and lateral diffusion (Barisch and Soldati, 2017b

D. discoideum Plaque Assays
As for animals, M. marinum is pathogenic for D. discoideum. Contrary to the food bacterium Klebsiella, the growth of D. discoideum is restricted when plated on M. marinum lawns (Ouertatani-Sakouhi et al., 2017). Researchers have made use of this growth incapacity to perform genetic and chemical screens for the search of virulence genes and anti-mycobacterial compounds. For instance, M. marinum TesA, a putative type II thioesterase required for the synthesis of the cell wall lipids PDIMs and PGLs, was identified in a D. discoideum plaque assay where a M. marinum transposon mutant library was screened for bacterial attenuation. The role of TesA in virulence was further confirmed during infection of zebrafish embryos, and the tesA gene was proposed as a possible genetic target for disruption in human mycobacterial pathogens (Alibaud et al., 2011). In addition, another plaque assay screen of M. marinum transposon mutants allowed Chen and collaborators to discover that two genes, mmar_2318 and mmar_2319, involved in the biosynthesis of lipooligosaccharide (LOS) contribute to M. marinum virulence in D. discoideum and limit bacterial entry in human macrophages (Chen et al., 2015) (Supplementary  Table 2). Chemical screens can also be performed in amoebae (Ouertatani-Sakouhi et al., 2017;Trofimov et al., in press). For instance, drugs targeting the M. marinum cell wall have been recently identified to block/decrease M. marinum virulence by inhibiting bacterial aggregation and permeability, sliding motility and/or intracellular replication within D. discoideum (Ouertatani-Sakouhi et al., 2017). All these findings highlight D. discoideum as an alternative host for the screening of Mycobacterium virulence factors and potential antibacterial compounds.

High throughput Screening
Since the advent of next-generation sequencing (NGS) technologies, numerous methods to investigate the hostpathogen interaction at different levels (i.e., RNA, DNA, and proteins) have been developed. Among them, RNA sequencing (RNA-Seq) is becoming the standard to analyze the transcripts of each actor involved in a bacterial infection, either independently (bacteria or host) or simultaneously (bacteria and host) by dual RNA-Seq (Saliba et al., 2017;Westermann et al., 2017). This method has been used to describe the gene expression profile of M. marinum during exponential and stationary growth (Wang et al., 2013), as well as under stress conditions (Pettersson et al., 2015). In addition, it has allowed to determine the contribution of specific cells (neutrophils and macrophages) to the innate immune response against M. marinum, Mtb, and M. bovis (Schnappinger et al., 2003;Nalpas et al., 2015;Kenyon et al., 2017) and to identify genes such as whiB4, which encodes a protein regulating the PE/PPE genes and is required for M. marinum and Mtb virulence (Chawla et al., 2012;Wu et al., 2017). Moreover, a recent metabolomics study coupled to dual RNA-Seq allowed to reconstruct the dietary map of Mtb during macrophage infection (Zimmermann et al., 2017). Since the RNA-Seq methodology has already been established in D. discoideum (Miranda et al., 2013), it will be of high interest to perform dual RNA-seq in the D. discoideum-M. marinum infection model.
Another NGS method is transposon sequencing (Tn-Seq), which allows the analysis of fitness and genetic interactions in microorganisms by using random transposon insertion libraries in bacteria (van Opijnen et al., 2009;DeJesus et al., , 2017avan Opijnen and Camilli, 2013;Nambi et al., 2015) Thanks to this method, various researchers have identified essential genes for Mtb growth and cholesterol metabolism (Griffin et al., 2011;Zhang et al., 2012; and have investigated the Mtb virulence factors required for survival in human dendritic cells (Mendum et al., 2015). Concerning M. marinum, only one study has been published so far , in which the authors identified ∼300 genes essential for survival of the bacteria in vitro. These genes are mostly shared with Mtb and correspond to 6% of the total coding sequences. The authors investigated the strategies used by M. marinum to exploit different phagocytic cells, from protozoan and vertebrate origin. The behavior of the M. marinum transposon mutant pool was highly similar between the two mammalian cell lines (human THP-1 monocytes and mouse RAW264.7 macrophages) and between the two amoebae D. discoidum and A. castellani, while an intermediate behavior was observed in cells derived from fish (carp CLC leukocytes), probably due to their higher similarity with natural M. marinum hosts. Thus, the authors concluded that M. marinum has both conserved and host-specific virulence determinants. Regarding the conserved virulence factors, transposon interruption of PDIMs, ESX-1, or Mce1/4 leads to high attenuation of M. marinum virulence in multiple hosts. In addition, a new virulence factor of M. marinum was highlighted and validated in vivo, the cpsA gene. This gene encodes a protein of the LytR family of transcriptional regulators and has been recently demonstrated to be required for M. marinum cell wall integrity and virulence in the zebrafish model Weerdenburg et al., 2015). Moreover, amoebae-specific pathways have been pointed out by the authors. For instance, transposon insertions in multiple genes involved in biosynthesis of vitamin B 12 highly increased fitness of M. marinum in Acanthamoeba, while they had no or little impact on the bacteria within the other host cells . To conclude, M. marinum uses specific mechanisms to adapt to different intracellular environments. Some specific requirements for infection of mammalian cells are not applicable for protozoa infection, and vice versa.
Finally, REMI sequencing (REMI-Seq) has recently been developed for D. discoideum by the group of C. Thompson (https://thethompsonlab.wordpress.com/mutant-library/). The REMI method was first described in Saccharomyces (Schiestl and Petes, 1991) and then adapted to D. discoideum (Kuspa and Loomis, 1992;Kuspa, 2006). The technique comprises the integration of linear DNA into eukaryotic genomes for insertional mutagenesis. Combining REMI with NGS, a genomewide collection of tagged mutants can be created to investigate host genes. However, REMI-Seq has not yet been used in the context of mycobacterial infection. Therefore, this tool, combined with other methods described above, are promising and innovative approaches that could help identify genes and pathways involved in mycobacteria-host interaction.

CONCLUDING REMARKS
In this review, we have described the so far known processes involved in the intracellular cycle of M. marinum within the amoeba D. discoideum. The comparison with studies carried out in human cells with Mtb revealed that the infection cycles of both bacteria share numerous similarities. However, many of the virulence mechanisms used by both pathogens to subvert the immune response of their hosts remain to be elucidated. We propose the D. discoideum-M. marinum system as a suitable and powerful model to study infection by tuberculous mycobacteria.
It needs to be noticed that not all the markers listed here were monitored in all the listed strains, and vice versa. Please see main text and supplementary tables for more details. IFA, immunofluorescence assay; TEM, transmission electron microscopy; qPCR, quantitative polymerase chain reaction; SEM, scanning electron microscopy; CLEM, correlative light and electron microscopy.

AUTHOR CONTRIBUTIONS
EC-M, CB, LL, and AL-J wrote the original draft. EC-M and CB drew the figures and prepared the tables. EC-M, CB, LL, and TS reviewed and edited the draft. TS supervised this work.

FUNDING
The TS laboratory is supported by multiple grants from the Swiss National Science Foundation, and TS is a member of iGE3 (www. ige3.unige.ch) as well as of the COST Actions BM1203 EU-ROS and CA15138 TRANSAUTOPHAGY.