Autophagy is a unique membrane-trafficking process in which newly formed membranes, called phagophores, engulf parts of the cytoplasm and lead to the production of double-membrane autophagosomes that are delivered to lysosomes for degradation(Wild et al., 2011; Rubinsztein et al., 2012). Autophagy is the main cellular process responsible for degrading defective organelles and long-lived proteins. This catabolic pathway has been linked to numerous pathological and physiological conditions (Abada and Elazar, 2014). The physiological role of autophagy was deduced when it was discovered, but the origin of autophagosomal membranes remains unclear(Hamasaki et al., 2013a; Diao et al., 2015).

Autophagy initiates with the emergence of a double-membraned isolation membrane, which encloses organelles and portions of the cytoplasm to form the autophagosome (Hayashi-Nishino et al., 2010; Carlsson and Simonsen, 2015; Sanchez-Wandelmer et al., 2015). Despite much progress in identifying the molecules responsible for autophagosome formation, the origin and the source of the autophagosomal membrane remain unsolved and have been the subject of long-standing debate. Autophagosome formation has recently attracted substantial attention (Abada and Elazar, 2014). Most of the seminal studies on the regulatory mechanisms of autophagosome biogenesis were conducted on Saccharomyces cerevisiae or culture cells in vitro (Hamasaki and Yoshimori, 2010). The membrane origins of autophagosomes may involve multiple sources (Hayashi-Nishino et al., 2009; Puri et al., 2013), including ER exit sites (ERES) (Zoppino et al., 2010; Graef et al., 2013), the ER–Golgi intermediate compartment (ERGIC) (Ge et al., 2013), the Golgi (Geng et al., 2010; Bodemann et al., 2011), the plasma membrane (Ravikumar et al., 2010) and recycling endosomes (Longatti et al., 2012; Knævelsrud et al., 2013; Puri et al., 2013). However, there is little ultrastructural information from the initial stage of autophagosome formation. It is unclear if and where distinct membrane sources fuse during autophagosome biogenesis.

A spermatozoon is a highly differentiated cell developed from a spermatid through spermiogenesis in the convoluted seminiferous tubule of the testis. Spermiogenesis is the intracellular clearance process for male gametes. The decrease in cell volume occurs due to the shedding of unnecessary cytoplasm and organelles, but the process for reptilian gametes remains largely unknown (Zhang et al., 2007). Our previous studies showed that epididymal spermatozoa of Chinese soft-shelled turtles contain a huge cytoplasmic droplet with numerous lipid droplets that serve as energy and nutrition sources, which favors long-term sperm storage (Zhang et al., 2015). There was not an obvious shedding process of unnecessary cytoplasm during spermiogenesis although, the cell size of the maturing spermatozoon was reduced. The cytoplasmic droplet always attached to the mid-piece of the spermatozoon without migrating down the sperm tail. In the present study, a subcellular mechanism for autophagosomal membrane biogenesis was examined in detail during in vivo turtle spermiogenesis. The turtle could be a potential animal model for long-term sperm storage at atmospheric temperature(Chen et al., 2015).

Autophagy plays a vital role in maintaining cellular homeostasis through the continuous turnover of proteins as well as the elimination of defective organelles in lysosomes (Yang and Klionsky, 2010; Bejarano et al., 2014). Unlike other organelles that solidly present in the cell, it is considered that autophagosomes are formed on demand by dynamic membrane rearrangements (Shibutani and Yoshimori, 2014). It is difficult to confirm the membrane sources of autophagosomes because autophagosome formation proceeds rapidly once it is initiated. It is estimated that autophagosome formation takes several minutes in yeast and mammals(Mizushima et al., 2001; Fujita et al., 2008; Geng et al., 2008). Despite tremendous progress made in autophagy, the long-term issues about the source and origin of autophagic membranes remain undiscovered in the past few decades. Recognizing the autophagosome formation site will undoubtedly help to figure out another fascinating question: where does the membrane of autophagosome come from (Hamasaki et al., 2013b)? The testis of the soft-shelled turtle expressed LC3-II, which is the marker protein of the autophagosome, located within spermatids in different stages during spermiogenesis. Numerous autophagosomes were observed inside the spermatids under TEM. A special 1–2 μm diameter structure named the “Chrysanthemum flower center” (CFC) was also found within the spermatid and several double-membrane IM could be produced, elongate and enwrap cytoplasm to develop different autophagosomes from this structure. All the cellular evidence showed that the “Chrysanthemum flower center,” a newly formed structure in the spermatid, may be one important site for IM development and autophagosome formation during cell-size reduction in spermiogenesis. The CFC was more developed in spermatids with compact nuclei than in spermatids with granular nuclei, which corresponds to the late stage and early stage of spermiogenesis, respectively. In late spermiogenesis, the differentiation and remodeling of spermatids into spermatozoa is more active. As far as we know, this study has yielded the first subcellular evidence for an in vivo origin and source of IM and autophagosome formation in higher eukaryotes.

In yeast, the membrane of the phagophore and early autophagic structures originate from a single source named the pre-autophagosomal structure (PAS) (Suzuki and Ohsumi, 2010). In higher eukaryotes, phagophore formationmight take place in different positions in the cell, but the current assumption is that phagophores formate at a complex membranous structure named the omegasome. Omegasomes were initially identified as endoplasmic reticulum (ER)-associated spots (Carlsson and Simonsen, 2015). The latest studies have shown that the ER exit sites (ERES), which are specialized ER regions where long-lived proteins are classified into the secretory system, are critical elements in the formation of isolation membranes. Therefore, ERES are a new component that should be concluded into the depiction of autophagosome biogenesis (Sanchez-Wandelmer et al., 2015). It is hypothesized that further research on ERES function will provide provoking insights into eukaryotic cell biology. Electron tomography showed that endoplasmic reticulum associates with isolation membranes or phagophore in mammalian cells. The ER-IM complexs served as a subregion of ER by forming a cradle girding the IM, and showed that both IMs and ER are connected. However, there is little ultrastructural information on the initial stage of autophagosome formation (Hayashi-Nishino et al., 2010). Through TEM, this study demonstrated that the developed Chrysanthemum flower center occurred in differentiating spermatids in vivo during turtle spermiogenesis. It was further revealed that CFC could produce long IM that enwrapped some cytoplasm to form the autophagosome. As a result, CFC in the spermatid might be equivalent to a phagophore nucleation site, or the omegasome, or the ER-IM complex in cultured mammalian cells.

Light microscopy experiments provide evidences to demonstrate the continuity between the ER and IMs. In addition, the experiments also found that DFCP1, an ER-resident protein, has access to IMs. (Axe et al., 2008). The mechanism by which the isolation membranes keep their unique identity while remaining connected to their donor membranes remains unclear (Kraft and Martens, 2012). Several pieces of evidence have shown a strong relationship between ER and autophagosome formation sites in mammalian culture cells(Hamasaki et al., 2013b). It was found that 70% of IMs were associated with ER (named ER–IM complexes) in Atg4B (Atg4BC74A) mutant cells, which suggests that the ER–IM complexes are an important early stage of AP formation (Hamasaki and Yoshimori, 2010). As a result, ER is probably the major donor of lipid bilayers for forming autophagosome and omegasome, which could represent the means to supply lipids from ER to a growing phagophore (Sanchez-Wandelmer et al., 2015). A large number of branched ER were distributed around the CFC, and their ends inserted into the CFC through short narrow tubules that formed many daisy petal-like structures that comprised the CFC from different orientations. The morphological connection of narrow tubules showed that the CFC unambiguously came from ER, which could develop IM and form autophagosome in the turtle spermatids. Our data could support the third hypothesis of autophagosome biogenesis that autophagosomes are formed neither by de novo membrane formation nor by direct utilization of ER cisternae (Hayashi-Nishino et al., 2009).

Different biogenesis pathways might regulate autophagy concurrently in vivo. Membrane fusion events would ask several autophagic biogenesis complexes to work in a few locations at different stages and could be tied together with diverse membranes. Distinct membrane sources will conduce to biogenesis because with cell's third dimension, many membranes and organelles are close to the phagophore (Abada and Elazar, 2014). It is likely that the Golgi route contributes to autophagosome biogenesis in yeast, although it is not clear if this pathway is important in mammalian systems (Rubinsztein et al., 2012). By tracking Atg9 in Atg9 over-expressing mammalian cells using immune-electron microscopy(IEM), researchers noticed that upon induction of autophagy, a post-Golgi tubulo-vesicular compartment responsive for Atg9 transfers to the vacuole(Mari et al., 2010). Furthermore, the tubulo-vesicular compartment experiences homotypic fusion and reestablishment for purpose of producing isolation membranes (IMs) (Kraft and Martens, 2012). How vesicular precursors integrate into IMs is still unknown. In cultured mammalian cells, Atg9 partly colocalizes to trans-Golgi network and endosomes (Hamasaki et al., 2013b). Additionally, depriving of nutrition induces Atg9 vesicles exit from the trans-Golgi network (TGN) and instantaneous co-localization of Atg9 with LC3 (the autophagosome marker) in fluorescence microscopy experiments. (Young et al., 2006; Orsi et al., 2012). Hamasaki and Yoshimori revealed that Atg9 localization intensively suggests the trans-Golgi network (TGN) and endosome as possible origins of autophagosomal membranes (Hamasaki and Yoshimori, 2010). Using cytotoxic stressors, another study suggested the trans-Golgi network (TGN) as a potential membrane origin in Atg5/Atg7-independent alternative autophagy (Nishida et al., 2009). Double-layer membranes from the trans-Golgi network could wrap some cytoplasm and its vesicles to form autophagosomes in turtle spermatids, which suggests that Golgi complex might be another important autophagy biogenesis site in addition to the CFC. On the trans-face of the Golgi complex of the differentiating spermatids, a large clear vesicle was formed. These vesicles could became flat and cling to the lipid droplet, which bore the characteristics of lipophagy (Liu and Czaja, 2013; Dupont et al., 2014).

Experiments revealed that membrane traffic from early to recycling endosomes is crucial for autophagosome formation and that the recycling compartment is not merely responsible for recycling of plasma membrane receptors but also serves as a station in the early stages of autophagosome biogenesis (Puri et al., 2013). It has been proposed that isolation membranes are formed by the fusion of vesicular carriers that are derived from the plasma membrane (Ravikumar et al., 2010; Moreau et al., 2011) or the Golgi apparatus (Ohashi and Munro, 2010; van der Vaart et al., 2010). Owing to autophagosome formation occurring away from donor membranes, diffusion barrier has no use for separating isolation membranes (IMs) from donor membranes in the vesicular fusion models (Kraft and Martens, 2012). Researchers revealed that the plasma membrane makes contributions to autophagosome formation and it is particularly significant in time of increased autophagy (Rubinsztein et al., 2012). Plasma membrane invagination was frequently observed on the spermatid membrane, which could lead to the formation of different vesicles and endosomes distributed around the CFC and autophagosomes or distribution close to the lipid droplet such as the Golgi vesicle in present study. The size reduced sharply from spermatids to spermatozoa. The plasma membrane's large surface area might act as a massive membrane store that allows cells to experiences cycles of autophagosome synthesis at much higher rates than under basic conditions. And it does not compromise other processes. Recent studies demonstrate the idea that these autophagy-related proteins and membranes that are necessary for autophagosome formation are originated from plasma membrane and that these components coalesce in recycling endosomes that are an early PAS or phagophore intermediate (Sanchez-Wandelmer et al., 2015).

In the present study, we determined that endoplasmic reticulum, vesicles from Golgi complexes, plasma membranes and endosomes might be the sources of autophagosomal membrane (Figure 10). The discovery of ER-derived “Chrysanthemum flower center” (CFC) supports the possibility of the cradle/omegasome model (Hamasaki and Yoshimori, 2010) or de novo membrane formation (Hayashi-Nishino et al., 2009). This finding brings us closer to understanding the origin of autophagosomal membranes. However, details in the formation or constituents of CFC still need further investigation.

The authors have made the following declarations about their contributions: YH and QC Conceived and designed the experiments. YH, HC, TL, XC, QL, YL, LH, and QZ Performed the experiments. PY, TL, and NA analyzed the data. YH and QC Wrote the paper. All authors read and approved the final manuscript.

This research was supported by grants from the National Natural Science Foundation of China (Grant number: 31272521 and 31672505) and Priority Academic Program Development of Jiangsu Higher Education Institutions, China.