With 9.6 million new cases and 1.5 million deaths in 2014, tuberculosis (TB) ranks alongside with HIV as the major cause of deadly infectious diseases (WHO, 2015). It is estimated that about one-third of the world population is infected with the etiological agent of TB, Mycobacterium tuberculosis, but only 5 to 10% of the infected individuals are at risk of developing active disease within a few years. Most individuals remain asymptomatic after the primary infection. In such individuals, bacilli do not divide but may persist for decades in a dormant state until reactivation of bacilli leads to active disease (WHO, 2015).

Infection with M. tuberculosis follows a relatively well-defined sequence of events (Russell, 2001; Russell et al., 2009a). After inhalation of contaminated aerosols, M. tuberculosis reaches the lungs, where the bacilli are phagocytosed by alveolar macrophages. Prognosis of the disease, viz, active or dormant, will depend on the host's ability to contain the bacilli at the site of infection. In most cases, the bacilli induce a local pro-inflammatory response which leads to the recruitment of other immune cells (macrophages, dendritic cells and lymphocytes) from neighboring blood vessels. The latter accumulate around the alveolar macrophages, at the site of infection, to form a granuloma which is the hallmark of TB (Russell, 2007; Peyron et al., 2008; Gideon and Flynn, 2011; Galagan, 2014).

The tuberculous granuloma is a highly elaborate host protective structure, which represents an essential immune mechanism for containing the bacilli inside the lungs without fully eradicating them (Kaufmann, 2001; Russell, 2001). The formation process, the composition and maintenance of this well-organized system have been extensively investigated and reviewed over the last decade (Russell et al., 2009a; Ramakrishnan, 2012; Guirado and Schlesinger, 2013). Briefly, macrophages differentiate into epithelioid cells or become highly vacuolated or even fuse to form multi-nucleated giant cells (Puissegur et al., 2007; Russell et al., 2009a; Feng et al., 2014). Other macrophages, which are the focus of intense research, differentiate into foam cells characterized by the intracytosolic accumulation of neutral lipids in the form of lipid bodies (LB) also known as lipid droplets or lipid vacuoles (Daniel et al., 2004; Deb et al., 2006; Russell et al., 2009a). These specific foamy macrophages (FM) are found in the interface region flanking the central necrotic center of the granuloma (Peyron et al., 2008).

Traditionally, FM were qualified as passive organelles involved in lipid storage. More recently, it became clear that these organelles play a central role in several inflammatory diseases, such as atherosclerosis, or in chronic infectious diseases (Russell et al., 2009b). Different myeloid cell types have been used to generate lipid-loaded foam cells and to provide insights into their contribution in chronic inflammatory conditions. In the context of TB, pathologists have described the presence of FM within granulomatous structures in both experimentally-infected animals and patients (Ridley and Ridley, 1987; Cardona et al., 2000), especially in individuals developing secondary TB (Pagel, 1925; Florey, 1958; Hunter et al., 2007), thereby suggesting that FM may play a central role in mycobacterial persistence and reactivation.

Since M. tuberculosis is facing important physiological changes as TB progresses toward containment of the bacilli within granulomas, it is assumed that this involves adaptive requirements that are distinct from those encountered during the initial phase of infection (Hingley-Wilson et al., 2003; Mckinney and Gomez, 2003; Sassetti and Rubin, 2003). Among these, accumulation of neutral lipids, essentially triacylglycerol (TAG), within intracytosolic lipid inclusions (ILI) (Dhouib et al., 2011), has been proposed to serve as a source of carbon and energy, and represents a hallmark of persistent and non-dividing mycobacteria (Garton et al., 2002; Daniel et al., 2011). Therefore, residing within specialized lipid-rich cells, rather than in conventional macrophages, would be beneficial and even promote M. tuberculosis persistence within granulomatous lesions. This also implies that bacilli must produce lipolytic enzymes and/or activate host hydrolytic enzymes in order to degrade the host neutral lipids into free fatty acids (FFA) and then re-synthesize TAG for subsequent ILI formation. Clearly, the abundance of genes encoding enzymes involved in lipid metabolism (Cole et al., 1998) strongly suggests that M. tuberculosis uses host-derived lipids while persisting in vivo (Höner Zu Bentrup and Russell, 2001).

The purpose of the present review is firstly to recall how host cells become foamy, by accumulating neutral lipids in the form of LB and, second, to describe and compare different foam cell models and approaches that have served to delineate lipid accumulation and consumption processes developed by M. tuberculosis and other mycobacteria within their host cells. Advantages and drawbacks of each model are also discussed.

The presence of neutral lipid storage organelles in the cytoplasm of most eukaryotes (named LB) and prokaryotes (named ILI) has been known for a long time. In all cases, these ubiquitous structures are composed of a core of neutral lipids, such as sterol esters and TAG, surrounded by a monolayer of phospholipids, in which cholesterol ester and proteins are inserted (Figure 1A; Murphy, 2001; Martin and Parton, 2005; Pol et al., 2014). A large number of associated proteins have been discovered in recent studies devoted to the composition and dynamics of LB formation (Brasaemle et al., 2004; Yang et al., 2012). Furthermore, it is now clear that, beyond neutral lipid storage, the functions of LB are multiple, and this knowledge has given rise to new hypotheses concerning the role of these organelles in several physiological processes (Saka and Valdivia, 2012).

The current models of LB biogenesis have been reviewed extensively (Murphy and Vance, 1999; Ohsaki et al., 2009; Saka and Valdivia, 2012). The prevailing model suggests that host LB originate at the endoplasmic reticulum (ER), where fatty acids are used to re-synthesize TAG and sterol esters. The de novo anabolic process leading to TAG synthesis relies on multiple enzymatic steps, the final one being catalyzed by diacylglycerol acyltransferases (DGAT) (Harris et al., 2011; Pol et al., 2014). Among these, DGAT1 and DGAT2 are located in the ER where they can synthesize TAG from diacylglycerol (DAG) and acyl-CoA. TAG and sterol esters then accumulate in the hydrophobic space between the leaflets of the ER bilayer membrane, where they are subsequently engulfed by phospholipids of the cytosolic leaflet and pinched off the ER membrane into the cytosol (Figure 1B; Murphy, 2001; Kassan et al., 2013; Pol et al., 2014).

Over one decade ago, Garton et al. established that M. tuberculosis could accumulate neutral lipids in the form of ILI in vitro and in sputum of TB patients (Garton et al., 2002). Since then, ILI formation has been extensively described and analyzed in several mycobacterial species, including M. tuberculosis (Mckinney et al., 2000; Daniel et al., 2004, 2011; Deb et al., 2006), M. bovis BCG (Low et al., 2009, 2010), M. leprae (Mattos et al., 2010, 2011), M. abscessus (Viljoen et al., 2016), and M. smegmatis (Dhouib et al., 2011). It has been proposed, but not yet proven, that storage of fatty acids in the form of TAG within ILI could serve as a source of carbon and energy during dormancy and reactivation of M. tuberculosis (Peyron et al., 2008; Low et al., 2009; Daniel et al., 2011; Dhouib et al., 2011).

Currently, almost nothing is known concerning the biological mechanism by which the storage of fatty acids may promote mycobacterial intracellular survival. Several hypotheses have been raised over the years, suggesting that ILI do not only serve as carbon source during infection, but could also reduce a wide range of oxido-reductive and metabolic stress (Kumar et al., 2011; Lee et al., 2013) It has also been proposed that the TAG accumulation can function in metabolic coupling to balance changes in metabolism of other M. tuberculosis lipids (Jain et al., 2007).

Accumulation of neutral lipids in the form of ILI has also been demonstrated in vitro or in vivo for a wide range of micro-organisms including Rhodoccoccus, Hepatitis C, and Dengue viruses, or Chlamydia (Saka and Valdivia, 2012; Alvarez, 2016) where they may serve other purposes.

Stimulated by these observations, several investigators have focused their activity on the search of external stimuli that may promote the accumulation of LB in macrophages and in other cells, with the aim of addressing the following key questions: (i) How do mycobacteria gain access to lipids from FM? (ii) How do they translocate the cellular lipids into their own cytoplasm? (iii) How are ILI formed and what is their composition? (iv) What are the lipolytic enzymes (from the host and mycobacteria) involved in these mechanisms? and (v) Is the arrest/resumption of mycobacterial division occurring during persistence and reactivation, respectively, linked to LB and ILI formation/consumption? At least some of these questions have been addressed thanks to the recent development of a panel of experimental models.

Several stimuli were tested for designing foam cell models that rest upon the following observations. Firstly, the granuloma is considered to be a hypoxic environment. The high cell density and poor vascularization seems to be responsible for this low oxygen content which has been shown to promote foam cell formation (Bostrom et al., 2006; Cardoso et al., 2015; Datta et al., 2016). Secondly, most FM are localized in the close vicinity of the caseous center (Cáceres et al., 2009) where decaying cells may provide neutral lipids or phospholipids (Kim et al., 2010). Such external nutrients are known to play a major role on macrophage lipid metabolism (Shashkin et al., 2005) and may lead to a foamy phenotype in vivo. Finally, specific cell wall-associated components, such as mycolic acids from pathogenic mycobacteria, might be involved in the formation of lipid-loaded cells inside the granuloma (Peyron et al., 2008; Russell et al., 2009a).

The different models presented in the present review have allowed to obtain relevant information on lipid metabolism in the context of M. tuberculosis persistence and reactivation. The use of animal models and ex vivo experimental systems have led to breakthrough information regarding cellular and molecular aspects related to (i) the acquisition of a foam cell phenotype where cells can be induced to store fat within LB under specific stimuli; (ii) the transfer of host LB contents to the lumen of the mycobacterium-containing phagosome following fusion events between LB and phagosomes, and (iii) the accumulation of neutral lipids in the form of ILI within mycobacteria.

Indeed, this large number of well-established host-pathogens systems, has allowed, recently the identification of a wide range of macrophages genes which are directly involved into pro-inflammatory response, energetic metabolism, such as glucose and lipid anabolism/catabolism leading to a foamy phenotype (Singh et al., 2015; Bhagyaraj et al., 2016; Holla et al., 2016).

In addition to those host factors, several mycobacterial proteins are likely to be involved in these various processes and identified (Deb et al., 2006; Kim et al., 2010; Low et al., 2010; Daniel et al., 2016).

As pointed out by Höner Zu Bentrup and Russell (2001), many genes of the M. tuberculosis genome encode enzymes involved in lipid metabolism (Cole et al., 1998). Among these, several might play important roles in the above-mentioned processes, particularly in the fusion events between LB and mycobacterium-containing phagosomes and the release of host lipids. As proposed by Caire-Brändli et al. (2014), these events can be viewed as the reverse steps involved in the transfer of lipids from the ER into the cytosol in the form of LB. The phospholipid monolayer leaflet covering the entire LB could merge with the outer leaflet of the phagosome membrane bilayer. This would leave the lipid contents between the two leaflets of the phagosomal membrane bilayer, where they remain separated from the mycobacterial surface by the inner leaflet. Because this remaining phospholipid barrier would present a different molecular arrangement than the otherwise intact phagosome membrane bilayer, it might be more vulnerable to the attack by mycobacterial lipolytic enzymes. Several mycobacteria, like M. avium, M. tuberculosis, and M. bovis BCG, seem to use essentially the neutral lipids contained within LB as a source of carbon during persistence, by secreting their own lipolytic enzymes and by activating host lipolytic enzymes (Daniel et al., 2011; Singh et al., 2012). In this context, M. tuberculosis genes encoding lipid/ester hydrolases, with a wide range of hydrolytic activities (Dedieu et al., 2013), deserve special attention. These enzymes display specific properties, such as the ability to be secreted and the capacity to hydrolyze phospholipid membrane layers. Cytotoxic effects have also been associated with some of these enzymes (Bakala N'goma et al., 2010; Schué et al., 2010). We propose that these enzymes attack the inner leaflet of the phagosome membrane bilayer. As a consequence, the LB content is released inside the phagosomal lumen, in the immediate vicinity of the mycobacterial surface. EM analyses also revealed that TAG from LB are rapidly re-used to form ILI, indicating that these molecules are first hydrolyzed to produce FFA and then imported inside the bacterial cytosol (Caire-Brändli et al., 2014). This implies the contribution of mycobacterial lipases exposed at the cell wall surface. The HSL-homologe LipY (Rv3097c), which hydrolyses long fatty acyl chains, is secreted by a type VII secretion system (ESX-5). LipY remains one of the most studied lipase of M. tuberculosis (Deb et al., 2006; Mishra et al., 2008; Delorme et al., 2012). Deletion of the lipY gene is accompanied by a significant decrease in the amount of ILI, indicating that LipY participates in TAG hydrolysis (Deb et al., 2006), although additional lipolytic activities are likely to be involved (Côtes et al., 2007; Dhouib et al., 2010; Schué et al., 2010). Recent transcriptomic studies in adipocytes infected with M. tuberculosis also favored the potential involvement of various M. tuberculosis lipases, cutinases, and esterases in lipid accumulation/consumption processes (Rastogi et al., 2016).

How the major lipolysis products derived from host TAG (DAG, MAG, and FFA) are formed and how they gain access to the mycobacterial cytosol is not fully understood. Over the past few years, several membrane proteins have been identified as potential actors of nutrient uptake and/or lipid mobilization in mycobacteria (Daniel et al., 2014; Ates et al., 2015; Martinot et al., 2016). Interestingly, a specific class of fatty acid transporter proteins (FATP), conserved in mycobacteria and humans, has been identified as a potential actor in the mobilization and activation of FFA (Hirsch et al., 1998; Daniel et al., 2014) leading to TAG synthesis by bacterial di and triacylglycerol synthases (Daniel et al., 2004; Sirakova et al., 2006; Viljoen et al., 2016).

We anticipate that the use of these various and complementary models and approaches described in the present review will undoubtedly facilitate the discovery of new cellular events and/or mycobacterial factors that will help to further delineate the intricate mechanisms involved in mycobacterial lipid acquisition and assimilation during the intracellular lifestyle of M. tuberculosis.

PS participated in the drafting of the section on lipid accumulation in macrophages, FB on adipocytes and NS on Dictyostelium parts. The introduction section was written by PS and FB. The figures were drawn by PS and electron microscopy pictures were provided by CD and IP. MD was helpful for correcting the manuscript. LK, CD, and SC completed and revised the entire manuscript. All of this work was proposed and supervised by SC.