The Stx family includes the prototype Stx from Shigella dysenteriae and related Shiga-like toxins Stx1 and Stx2, produced by enterohemorrhagic Escherichia coli (EHEC) ^[1][45]. Stx1 differs from Stx by only one amino acid residue, whereas Stx2 represents distinct serotypes with ~56% sequence identity compared to Stx ^[1][2][45,46]. Belonging to the AB₅ toxin superfamily, the Stx family is composed of an A chain (32 kDa) and five identical B chains (7.7 kDa each). The A chain is the enzymatic domain, which acts as an N-glycosidase that cleaves the host ribosomal RNA, while the five B chains form a pentameric ring and serve as the receptor binding domain. The A chain is connected to the B chain by inserting its C-terminus into the central pore of the B chain pentamer.

Once Stxs bind to the cellular receptor and enter cells through endocytosis by either clathrin-dependent or -independent pathways, they are sorted into the retrograde trafficking route and enter the trans-Golgi network (TGN). The A chain is processed by the host protease furin and cleaved into the enzymatic part A1 (27.5 kDa) and the B chain connecting part A2 (4.5 kDa). The A1 and A2 remain connected through an intramolecular disulfide bond between cysteines 242 and 261 residues. Stxs are further transported into the lumen of the ER, where the disulfide bond is reduced. The A1 part then crosses the ER membrane and enters the cytoplasm, utilizing the host ER-associated protein degradation (ERAD) machinery. The cytosolic Stxs eventually shut down protein synthesis by digesting ribosomal RNA and causing cell death ^[1][2][45,46].

The Stx B-chain pentamer specifically recognizes the carbohydrate moiety of Gb3 as its receptor ^[1][2][3][4][45,46,47,48]. The crystal structure suggests that each Stx B chain contains three Gb3 binding sites. Thus, one Stx holotoxin could maximally cluster fifteen Gb3 molecules on the cell surface ^[4][48]. Gb3 is the first member of the globo-series GSLs. The synthesis of Gb3 by transferring UDP-galactose onto LacCer is catalyzed by A4GALT (α-1,4-galactosyltransferase, also known as Gb3 synthetase) ^[5][6][49,50]. The expression of Gb3 in humans is highly restricted to the kidney, nervous system, microvascular endothelium, and a subset of germinal center B cells. In contrast, most other cell types do not express detectable levels of Gb3 ^{[7][8][9][10]}[51,52,53,54]. The kidney-enriched Gb3 is responsible for the life-threatening post-diarrheal hemolytic uremic syndrome (D+HUS) induced by EHEC infection ^[2][11][46,55]. On the other hand, Gb3 lysosomal accumulation leads to a type of sphingolipidoses known as Fabry disease, which is the consequence of the loss-of-function of a lysosomal enzyme called α-galactosidase A, which is responsible for the degradation of Gb3 ^[12][56]. Enzyme replacement therapy is the only reliable treatment for Fabry disease nowadays ^[13][14][57,58]. In contrast, substrate reduction therapy is another promising approach involving the inhibition of Gb3 biosynthesis using small-molecule drugs such as ceramide analogs or imino-sugars, which holds promise as an alternative approach ^[15][38].

In 2018, Tian et al. reported the first CRISPR-Cas9-mediated genome-wide screen for Stx1 and Stx2 ^[16][31]. The screen was conducted using the human bladder carcinoma 5637 cell line in a loss-of-function manner. The majority of the top-ranked hits overlapped between the Stx1 and Stx2 screens. Five top-ranked genes were key factors in the established Gb3 biosynthesis pathway: SPTSSA, UGCG, B4GALT5, A4GALT, and SLC35A2. Particularly, SPTSSA is a component of the serine palmitoyltransferase (SPT) complex on the ER membrane, which catalyzes the rate-limiting step in ceramide generation ^[17][37]. Other notable top-ranked hits shared by both screens include UGP2 and SPPL3. UGP2 (UDP-glucose pyrophosphorylase 2) is the key enzyme that produces UDP-glucose, the substrate for GlcCer synthesis. SPPL3 (signal peptide peptidase-like 3) is a Golgi-localized protease implicated in the cleavage and activation of many glycosyltransferases ^[18][19][59,60]. Tian et al. then focused on investigating the other three newly identified factors, LAPTM4A, TMEM165, and TM9SF2 ^[16][31].

LAPTM4A (lysosomal-associated protein transmembrane 4 A) was identified as a top hit in the screens (ranking No. 2 in the Stx1 screen and No. 1 in the Stx2 screen). However, its function had not been well characterized. Tian et al. found that knocking out LAPTM4A phenotypically mimicked knocking out A4GALT (Gb3 synthetase) in four aspects: (1) Both dramatically increased the cell resistance to Stx1 and Stx2, but not Ctx; (2) both abolished Stx cell surface binding but had no effect on Ctx binding; (3) both greatly reduced the expression level of Gb3, as measured by the mass spectrometry-based lipidomic assay; and (4) both induced the accumulation of the Gb3 precursor LacCer. However, the Golgi localization and the expression level of A4GALT were not altered in the LAPTM4A-knockout cells. Additionally, these properties of LAPTM4A are not shared with those of its homolog, LAPTM4B ^[16][31]. LAPTM4A is a small protein with 233 residues and four transmembrane domains, initially reported as an endosomal/lysosomal protein ^[20][21][61,62]. Tian et al. found that LAPTM4A is predominantly localized in the Golgi in multiple cell lines and physically interacts with A4GALT. Through the investigation of membrane topology and comparison of a panel of LAPTM4A/LAPTM4B chimeric proteins, the second lumenal domain of LAPTM4A was demonstrated to be critical for its function ^[16][31]. These results indicate that LAPTM4A is likely involved in the last step of Gb3 biosynthesis by serving as an essential co-factor for A4GALT’s enzymatic activity. However, the molecular basis of the interaction between LAPTM4A and A4GALT remains to be established.

TMEM165 (transmembrane protein 165) encodes a multi-pass transmembrane protein that is Golgi-localized and has been proposed as a transporter for manganese ions (Mn²⁺) ^[22][63]. TMEM165 is critical for maintaining Mn²⁺ hemostasis, and its mutations have been linked to human disorders with defects in glycosylation ^[23][24][64,65], as Mn²⁺ is required for many Golgi-localized glycosyltransferases. Tian et al. found that TMEM165-deficient cells were more resistant and had lower cell surface binding to Stx and Ctx. In contrast to LAPTM4A, TMEM165-deficient cells had lower levels of Gb3 and Gb3 precursors and gangliosides, suggesting that TMEM165 affects the biosynthesis of GSLs globally. Consistent with previous reports, Tian et al. confirmed the Golgi localization of TMEM165. Furthermore, the downsides of TMEM165 deficiency could be rescued by supplementing extra Mn²⁺, and TMEM165-deficient cells showed lower tolerance to Mn²⁺-induced cytotoxicity ^[16][31]. These findings experimentally confirmed the role of TMEM165 in regulating Mn²⁺ homeostasis.

TM9SF2 (transmembrane 9 superfamily member 2) encodes a highly conserved but poorly characterized multi-pass transmembrane protein that has been reported to have endosomal or Golgi localization ^[25][26][66,67]. It has also been associated with multiple glycosylation pathways, including heparan sulfate proteoglycan biosynthesis ^[26][67]. Tian et al. verified the Golgi-localization of TM9SF2 in multiple cell lines and found that TM9SF2-knockout cells had a lower level of surface heparan sulfate. Similar to TMEM165, TM9SF2-knockout cells expressed a lower level of GSLs, indicating that loss of TM9SF2 causes a global disruption in GSL biosynthesis and contributes to the resistance to both Stx and Ctx ^[16][31]. The detailed mechanism of how TM9SF2 is involved in glycosylation, whether similar to TMEM165 ^[10][54], remains to be established.

In 2019, Yamaji et al. reported an independent CRISPR screen for Stx1 in HeLa cells ^[27][68]. The screen once again identified well-established genes involved in GSL biosynthesis and the novel factors LAPTM4A, TMEM165, and TM9SF2. Using radioisotope labeling and thin-layer chromatography, Yamaji et al. demonstrated the requirement for LAPTM4A but not LAPTM4B in the last step of Gb3 synthesis. They further provided direct evidence showing that the enzymatic activity of A4GALT in cell lysates was greatly reduced when LAPTM4A was absent ^[27][68]. Yamaji et al. also showed that TM9SF2 is involved in Gb3 biosynthesis (likely through A4GALT) by its conserved C-terminus across all transmembrane 9 superfamily members (TM9SF1, TM9SF2, TM9SF3, and TM9SF4) ^[27][68]. In 2021, the same group reported a related screen on Vero cells derived from green monkeys, and the knockout was generated by a library targeting the human genome ^[28][69]. Although the degree of gene enrichment was less than the compatible screen in human cells, this screen still identified major players in the GSLs pathway, including LAPTM4A and TM9SF2 ^[28][69]. In 2020, Majumder et al. performed a similar screen on HeLa cells ^[29][70]. In addition to the same set of factors (including LAPTM4A and TM9SF2), a transcription factor, AHR (aryl hydrocarbon receptor), was uniquely identified, which may regulate Gb3 biosynthesis by regulating the expression of several known factors such as SPTSSA ^[29][70].

In 2018, Pacheco et al. conducted a unique screen directly using EHEC. This carries an additional virulence factor besides Stx and is known as the type III secretion system (T3SS) in the human intestinal epithelial HT29 cell line. HT29 cells were chosen because they are sensitive to EHEC co-culture but resistant to purified Stx ^[30][71]. Interestingly, many of the identified genes in this screen were involved in Gb3 biosynthesis (including LAPTM4A and TM9SF2), suggesting the potential role of Gb3 in the virulence of both Stx and T3SS. This finding further emphasizes that targeting factors in the GSL biosynthesis pathway is promising for designing new drugs to combat Stx and EHEC infections ^[30][71].

Another unique screen was reported by Kono et al., which specially focused on the ceramide salvage pathway ^[31][72]. The screen was carried out in HeLa cells with a de novo ceramide synthesis defect by knocking out the key gene SPTLC1 (serine palmitoyltransferase long chain base subunit 1). The SPTLC1-knockout cells are unable to generate 3-Keto-dihydrosphingosine from serine and palmitoyl-CoA. Then, sphingosine-1-phosphate (S1P) was added to the culture medium to activate the salvage pathway and restore the expression of Gb3. The CRISPR screen using Stx2 as killing stress under this condition successfully identified several important genes. These include ceramide synthase CERS2, well-known genes in the Gb3 biosynthesis pathway (UGCG, B4GALT5, A4GALT, and SLC35A2), and three newly identified factors (LAPTM4A, TMEM165, and TM9SF2). Additionally, two phosphatases, PLPP3 (phospholipid phosphatase 3, also known as PPAP2B) and SGPP1 (S1P phosphatase 1), were also identified through this screening process ^[31][72]. Kono et al. found that the cell surface-expressed PLPP3 is important for the uptake of extracellular S1P by dephosphorylating S1P into sphingosine. Then, the cellular sphingosine is rephosphorylated to S1P and further dephosphorylated by SGPP1 for ceramide synthesis ^[31][72].

Ctx is the major virulence factor produced by toxigenic strains of Vibrio cholerae ^[32][73]. Similar to Stxs, Ctx also belongs to the AB₅ toxin superfamily and shows a similar overall architecture. Upon binding to the cell surface and endocytosis, Ctx undergoes retrograde trafficking and releases its enzymatic A chain across the ER membrane ^[33][74]. The cytosol-exposed A chain deactivates the GTP hydrolase activity of the G_S alpha subunit through an ADP-ribosylation reaction and causes the continuous expression of 3′,5′-cyclic AMP (cAMP). This, in turn, triggers a series of consequences and eventually leads to the opening of the cAMP-dependent chloride channel CFTR (cystic fibrosis transmembrane conductance regulator) ^[34][35][36][75,76,77]. This process is responsible for the pathogenic effects caused by cholera infection, such as rapid fluid loss and rice-water stool ^[37][78].

The Ctx B-chain pentamer specifically recognizes the carbohydrate moiety of GM1, particularly GM1a in the a-series of gangliosides, as its receptor ^[38][79]. Structural studies suggest that one Ctx holotoxin could maximally cluster five GM1a molecules on the cell surface ^[39][80]. Thus, the Ctx B-chain pentamer has been widely used as a probe for studying ganglioside biology ^[40][81]. The biosynthesis of GM1a from LacCer requires three steps: (1) adding CMP-sialic acid to LacCer and generating GM3 by ST3GAL5 (ST3 β-galactoside α-2,3-sialyltransferase 5); (2) adding UDP-N-acetylgalactosamine (UDP-GalNAc) to GM3 and generating GM2 by B4GALNT1 (β-1,4-N-acetylgalactosaminyltransferase 1); and (3) adding UDP-galactose to GM2 and generating GM1a by B3GALT4 (β-1,3-galactosyltransferase 4) ^[41][82]. Unlike Gb3, GM1 is widely expressed in human tissues and is involved in multiple essential functions ^[41][82].

In 2011, Guimaraes et al. reported a genetic screen for Ctx using a retroviral insertion-based loss-of-function strategy in human haploid KBM7 cells ^[42][83]. Since Ctx itself cannot sufficiently kill cells, Guimaraes et al. engineered a lethal chimera toxin by fusing the enzymatic domain of diphtheria toxin (DTA) to the Ctx A chain using the sortase ligation method. DTA is also an ADP-ribosylation enzyme, but it specifically modifies a unique residue called diphthamide in eukaryotic elongation factor 2 (eEF-2), thereby blocking protein synthesis and inducing cell death ^[43][44][84,85]. The screen identified the genes involved in diphthamide biosynthesis that are responsible for DTA, as well as a series of genes involved in GM1a biosynthesis (e.g., UGCG, SLC35A2, B3GALT4, and ST3GAL5). ST3GAL5 is the GM3 synthase, and GM3 is the shared precursor for gangliosides from the a-, b-, and c-series but not from the 0-series (e.g., GM1b) ^[45][86]. Guimaraes et al. found that the ST3GAL5-knockout cells were resistant to Ctx. However, Ctx could still bind to a subset of cells (5–10%) at levels comparable to wild-type cells. They speculated that in the absence of ST3GAL5, an alternative ganglioside synthesis pathway for the 0-series can be initiated in a cell cycle-dependent manner. In this case, GM1b acts as an alternative Ctx receptor ^[42][83].

In 2014, Gilbert et al. reported another set of genetic screens for the Ctx-DTA chimera toxin using the emerging CRISPRi and CRISPRa approaches ^[46][26]. The results of the CRISPRi-based loss-of-function screen suggested that downregulating the genes involved in GM1a biosynthesis (e.g., B3GALT4 and ST3GAL5) leads to the protection of Ctx-DTA. In contrast, downregulating the genes involved in the biosynthesis of other gangliosides, including GM1b (e.g., ST3GAL2), resulted in sensitization to Ctx-DTA. The CRISPRa-based gain-of-function screen yielded results consistent with the CRISPRi screen. In addition, upregulating the genes involved in the biosynthesis of lactosides/neolactosides (e.g., B3GNT5) had protective effects. Both CRISPRi and CRISPRa screens suggested that protection against Ctx-DTA was apparently caused by diverting the shared precursor LacCer away from GM1a synthesis to other branches ^[46][26]. These recent genetic screens expand theour understanding of the complex and branched GSL biosynthesis pathways.