|Home | About | Journals | Submit | Contact Us | Français|
The field of metabolomics continues to catalog new compounds, but their functional analysis remains technically challenging, and roles beyond metabolism are largely unknown. Unbiased genetic/RNAi screens are powerful tools to identify the in vivo functions of protein-encoding genes, but not of non-proteinaceous compounds such as lipids. They can, however, identify the biosynthetic enzymes – of these compounds- findings that are usually dismissed, as these typically synthesize multiple products. Here, we provide a method using follow-on biosynthetic-pathway screens to identify the endpoint biosynthetic enzyme and thus the compound through which they act. The approach is based on the principle that all subsequently identified downstream biosynthetic enzymes contribute to the synthesis of at least this one end product. We describe how to: systematically target lipid biosynthetic pathways; optimize targeting conditions; take advantage of pathway branchpoints; and validate results by genetic assays and biochemical analyses. This approach extends the power of unbiased genetic/RNAi screens to identify in vivo functions of non-nucleic-acid-based metabolites beyond their metabolic roles.
The role of metabolites in biological functions other than metabolism is likely to be considerably underappreciated given the technical difficulties of functionally characterizing these compounds. For instance, lipids are major components of membranes, but their role in morphogenesis is largely unknown1, 2. A key technical problem is the inaccessibility of nontemplate-derived compounds to specific in vivo targeting techniques such as gene knockouts or RNAi knockdowns that constitute the basis for the functional analysis of proteins. Although mass spectrometry (MS)- and NMR-based metabolomic approaches offer systems-level annotation and characterization of metabolites and their specific moieties, they cannot identify a causal relationship between a specific metabolite and its cellular function3–5. Imaging techniques are not yet of high spatial resolution (e.g. the emerging MS-based techniques); they lack molecular specificity or interfere with function (e.g. of lipids on membranes)6; and the large panel of antibodies available for the detection of proteins, does not exist for lipids. Here we co-opt the successful strategy of nucleic acid targeting to investigate non-nucleic acid-derived metabolites such as lipids.
In a recent genome-wide RNAi tubulogenesis screen we identified a number of lipid-biosynthetic enzymes whose knockdowns generated the same specific polarity phenotype7. To find a putative common lipid compound through which these enzymes might act, we designed biosynthetic pathway screens and successfully identified ceramide glucosyltransferases (CGTs) as the endpoint biosynthetic enzymes, thereby revealing their single product glucosylceramide (GlcCer; the glycosphingolipid [GSL] backbone) as the common polarity-affecting lipid. Distinct biochemical functions of other identified enzymes, complemented by MS, characterized the compound being sought as a hydroxylated GlcCer (GlcCer-OH) with a C17-monomethylbranched-chain fatty acid (mmBCFA) and a saturated long-chain fatty acid (LCFA) of likely C22 length. No other existing technique could have identified this specific complex lipid’s function by selectively removing it, nor would its biosynthetic enzymes have been identified in a forward mutagenesis/RNAi screen, as it required a double paralog knockdown. This analysis demonstrates that this biosynthetic pathway screen approach can be used to query, in an unbiased manner, the in vivo cell biological functions of non-DNA-encoded metabolites beyond their roles as biosynthetic substrates or metabolic products.
The approach consists of a stepwise, systematic, although not comprehensive, targeting of all the molecules of the biosynthetic pathway of which the originally identified enzyme is a part. The same genetic/RNAi technique is used to search for a copy of the original phenotype with the goal of identifying the endpoint biosynthetic enzyme in the pathway, and thereby the underlying product species, and perhaps the exact chemical characteristics of the compound. The approach (1) eliminates potential bias by starting the inquiry with the analysis of the function (phenotype) rather than the molecule of interest (a ‘forward’ rather than a ‘reverse’ molecular analysis); (2) bypasses the problem of interconversion and compensation that creates a particular difficulty for the functional analysis of lipids8, 9 and (3) overcomes the obstacle that one enzyme synthesizes many products and affects different functions. This method can be used not only to functionally characterize metabolic products of biosynthetic genes that have already been identified in time-consuming genome-wide RNAi screens (typically set aside as ‘housekeeping’ genes) but also to design ‘lipid/metabolite’ screens upfront by constructing RNAi libraries from their biosynthetic genes for the functional characterization of their products.
The strength of the approach lies in its potential to identify the mostly unknown non-metabolic (rather than the generally well-known metabolic) cellular roles of the multitude of specific metabolic compounds (rather than their enzymes). However, it also characterizes the biosynthetic functions of the enzymes in vivo, which, in turn, may not only identify novel biosynthetic functions but also expand or revise our knowledge about the biosynthetic pathways themselves, often only established in vitro or in a restricted number of species. Moreover, in combination with techniques such as liquid chromatography (LC)-MS and gas chromatography (GC)-MS, that determine compound structure, this approach will also identify species-specific different or altogether novel compounds, together with their functions. To detect such features and incorporate them into the analysis, biochemical procedures are integrated into this overall genetic screen approach that otherwise relies on the particularly high conservation of metabolic enzymes between different species.
Here, we provide a detailed technical protocol on how to design and carry out such a biosynthetic pathway screen, combined with genetic and biochemical analyses. We demonstrate the approach with lipids, specifically, membrane lipids, and illustrate it with our prior RNAi feeding screen on epithelial polarity in C. elegans (see refs.10–17 and http://www.wormbase.org for the expanding field of lipid biology and metabolomics in C. elegans). The concept of the approach should, however, be equally effective in other genetic model organisms (limited only by their accessibility to specific targeting procedures), and it should serve to identify the functions of other categories of lipids or other non-template-derived compounds, such as, for instance, sugars. With the advent of scalable knockout/knockdown techniques (RNAi and more recently clustered regularly interspaced short palindromic repeats (CRISPR) technology, forward screens, the method proposed here, can also be performed in mammalian cell lines. On the other hand, the use of live genetic screens in model organisms, not feasible in humans, is particularly well suited to the functional dissection of biosynthetic pathways and can serve as a direct means to investigate the non-metabolic functions of their well-conserved human orthologs. As such screens start from the phenotype, not the metabolite, they are not limited by species specificities of this metabolite, such as for instance fatty acid (FA) chain length; but, when combined with biochemical analyses, have the potential to identify the exact chemical composition of the causative metabolic end product.
The biosynthetic-pathway screen approach relies on the copy of a signature phenotype that is generated by interference with one or several of multiple biosynthetic enzymes in a linear pathway (Figs. 1 and and2).2). The depletion of any intermediate (via its biosynthetic enzyme) in a specific pathway should (in principle) produce this phenotype, but inadvertently such screens also test conditions in which these same intermediates are increased, e.g., as accumulating substrates upstream of the catabolizing enzymes that are being tested via their depletion. Thus, seemingly paradoxically, the same phenotype may be present in animals that are depleted of a specific intermediate (by interference with its upstream biosynthetic enzyme) and in animals in which this same intermediate accumulates (secondary to interference with its downstream biosynthetic enzyme). However, in the latter scenario (accumulation of the intermediate), the phenotype is caused by the depletion of further downstream products, not by accumulation of this intermediate. This property of pathway screens, the simultaneous testing of intermediates as substrates and products, constitutes this approach’s strength in uniquely addressing an inherent problem of metabolic analysis: the circularity of biosynthetic pathways and the possiblity of compound interconnectivity (Figs.1 and and22).18 It provides exceptionally strong evidence to demonstrate that it is the endpoint biosynthetic end product, not its upstream intermediates, that mediates the function under study. For instance, the same polarity defects were present in animals with loss and gain of ceramide (Cer) in our lipid bisoynthetic pathway analysis (via depletion of Cer’s upstream and downstream enzymes, respectively), suggesting that the mediating lipid compound sought was a further downstream product, not Cer itself (in this case GlcCer)7.
Although the biosynthetic pathway screen approach addresses the problem of interconversion, it may not always solve it, nor can it overcome the problem of biosynthesis compartimentalization to different subcellular membrane systems. For instance, only the additional absence of the phenotype in a null mutant of its catabolizing enzymes would unequivocally exclude the possible contribution of the next step metabolite to the function that is being examined. This becomes an obvious challenge in cases in which a non-endpoint intermediate is suspected to execute the function of interest, which is indicated, for instance, by the absence of phenotypes in animals depleted of all its downstream metabolizing enzymes. Such a scenario may require the construction of null mutants of all immediate downstream pathway components, which may consist of several enzymes, including their paralogs. Moreover, in this particular case, strict validation of the intermediate by compound rescue of the phenotype would require the construction of a double (multiple) mutant of the immediate upstream (synthesizing) and downstream (catabolizing) enzymes. It is therefore particularly difficult to characterize functions of pathway hub compounds, such as Cer, which is a substrate of three and a product of four different metabolic reactions in C. elegans, most of which are catalyzed by several paralogous enzymes (Fig.2). Finally, the compartimentalization of lipid biosynthesis to different endomembrane and plasma membrane domains may cause restrictions to these biochemical rules, the functional consequences of which cannot currently be predicted. However, none of these problems can be addressed at present by any other available in vivo technique.
Procedures are provided for the scenario that one or several upstream FA biosynthetic enzyme genes have been identified in an unbiased screen as being required for a function of interest on the basis of their C. elegans RNAi phenotypes. As shown in Figure 3, a broad initial pathway screen (I; Steps 10–27) is designed to identify the underlying metabolite category/class and pathway branch; this is followed by an in-depth second screen (II; Steps 34 and 35) of this branch, aiming to identify the specific end product, followed (optionally) by a screen for derivatives of this end product (III; Steps 59–61), with the potential to identify its exact chemical characteristics. The direct identification of enzymes synthesizing complex lipids that are located further downstream in the pathway may make the broad screen (I) dispensible (see Supplementary Fig. 1 for complex lipid classes and their chemical compositions). The ancillary genetic and biochemical analyses serve to validate results, to connect results between different screen tiers and to exclude pathway branches, thereby reducing the number of enzymes requiring screening.
In principle, the identification of only one biosynthetic enzyme in a genetic/RNAi screen merits searching for the specific product that mediates its function. When the identified enzyme generates only one product, no further screening is necessary. More commonly, however, it generates multiple intermediates that require further analysis. The identification of more than one biosynthetic enzyme whose losses produce the same phenotype is particularly reassuring; and the identities of these enzymes may already be informative for the screen design. Biosynthetic pathway screens are particularly useful for the analysis of ‘upstream’ (in relation to their endproduct) biosynthetic enzymes that contribute to the generation of multiple products and thus reveal little information with regard to the identity of the product mediating the investigated function. In our polarity screen, the losses of four lipid biosynthetic enzymes were found to generate the same specific phenotype, which strongly suggested they function through a shared lipid compound7. Three of them were common fatty-acid (FA)-biosynthetic enzymes, one of which was the first enzyme for all FA biosynthesis (pod-2/acetyl-CoA carboxylase(ACC)), itself giving no clue as to the nature of this compound (Fig. 1).
Several points should be checked before embarking on follow-on pathway screens (Fig. 3). First, the specificity and robustness of the phenotype is essential for the analysis that relies on its tracking. Lethality or slow growth, for instance, may be induced by interference with any of various lipids and are thus not useful for tracking a single pathway; nor is a phenotype with low penetrance or expressivity. Second, enzyme knockdowns should be validated in a corresponding germline mutant that is expected to display the same or a related phenotype, particularly if only one enzyme was identified. Third, it is useful to exclude upfront the possibility that the phenotype results not from the loss of a product downstream of the missing enzyme but rather from the accumulation of a substrate upstream of the missing enzyme, a possible alternate cause (various lipid storage diseases, for instance, are caused by toxic substrate accumulation)19. The identification of more than one biosynthetic enzyme, or of first-step biosynthetic enzymes, may already suggest that their (common) loss-of-function phenotype is not caused by an accumulating substrate. This can be ascertained by rescuing the phenotype with exogenously supplemented lipids, provided that the identified enzyme synthesizes a product that can be substituted (an expected scenario, as pathway screens are likely to be used to analyze early acting, upstream biosynthetic enzymes that generate simple FAs). Compound rescue simultaneously rules out the third, less likely, cause of a phenotype: a dual-specific function of the enzyme that is unrelated to its biosynthetic product.
Lipid biosynthetic pathways can be divided into three parts that build on each other (Fig.1): (1) short- and medium chain FA biosynthesis (upstream; top); (2) LCFA biosynthesis (middle); (3) complex lipid biosynthesis (downstream; bottom). Synthesis of both short- and medium-chain FAs and LCFAs involves four-step elongation cycles (condensation, reduction, dehydration and reduction, depicted for LCFA synthesis), with malonyl-CoA, which is synthesized by ACC, as chain extender. Medium-chain FAs and LCFAs can be further modified, for example, through desaturation by desaturases (right) and/or hydroxylation by hydroxylases (not shown). They then serve as precursors for complex lipids (bottom).
Not infrequently the loss of first-step, often rate-limiting, biosynthetic enzymes generate detectable phenotypes, which implicate the entire field of possible downstream products as candidate mediating compounds. The first-tier follow-on pathway RNAi screen searching for phenocopy should assemble an RNAi library that targets genes involved in a wide range of lipid-related processes, including synthesis, transport and regulation, with an emphasis on the (upstream) basic machinery of FA metabolism (the size of libraries typically ranges from ~100 to 300 molecules, scalable depending on ease of phenotypic evaluation7; Table 1 shows a collection of critical C. elegans FA biosynthetic enzymes). This analysis is targeted towards the identification of clues, and it should thus be broad rather than comprehensive.
Any positive hit will be both confirmatory with regard to the initially identified enzyme and newly informative with regard to the further screen design. For example, the identification of an elongase in a first-level follow-on pathway screen for an initial identification of ACC/pod-2 (Fig. 1): (1) supports the functioning of ACC/pod-2 via a downstream lipid compound; (2) identifies an LCFA component of this lipid compound and (3) may reveal – depending on what elongase was identified – whether this LCFA is branched or straight. Of nine C. elegans elongases, ELO-5 and ELO-6 have been characterized as producing branched-chain LCFAs (C15ISO and C17ISO)12, and thus their identification would reveal the presence of a mmBCFA in the compound in question (Fig. 1)12, 20.
It is not necessary to probe every enzyme of the pathway, but predictions can be tested for further confirmation: for example, our first-level follow-on lipid biosynthetic pathway screens identified two polarity-affecting enzymes catalyzing successive steps of fatty acid elongation: the elongase ELO-3 and the 3-ketoacyl-CoA-reductase/KAR LET-767. Probing the enzyme catalyzing the reaction immediately downstream of KAR/LET-767 additionally identified the 3-hydroxyacyl-CoA dehydratase T15B7.2, thereby confirming pathway linearity (Fig. 1 and and33)7.
Identifying the phenotype of interest in animals depleted of a first step branch-specific enzyme will immediately direct follow-on screens to the corresponding branch. To increase confidence in the specific branch’s involvement, findings can be validated by the analysis of the corresponding germline mutants and/or by rescuing the phenotype via a downstream lipid compound. As complex lipids - in contrast to FAs - may not be easily available for rescue, enzyme inhibitors can be used to evaluate whether they copy the phenotype of interest. Many small-molecule inhibitors have been developed to modify lipid biosynthetic enzyme activities21–25(Tables 1, ,2).2). For example, inhibitors of serine palmitoyltransferase include sphingofungins, lipoxamycin and myriocin. In C. elegans, they can generally be delivered by feeding. Inhibitors are often competitive substrates, and thus the same precautions apply for their supplementation as for lipid feeding (steps 5 – 9). Their efficiency at inhibition in vivo varies, thus, although phenocopy will be confirmatory, a negative result will not rule out the targeted enzyme.
Negative RNAi results, although informative, cannot by themselves exclude an enzyme as a pathway component. They could be caused by a number of reasons, including clone- or gene-specific RNAi resistance and redundancies, the latter being a typical scenario for biosynthetic enzyme genes, which often have paralogs (Tables 1 and and2).2). Although the screen design is guided by positive results, negative results can be used for branch exclusion (Fig. 3). In this case, further confirmation is nessessary, with rigorous proof attainable only with germline null mutants; thus, it may require the construction of double, triple or even quadruple mutants. It is time-efficient to construct or obtain such mutants (many already available) if, with their help, large numbers of compounds can be excluded. This will be the case if the identified enzymes operate upstream in the pathway, thereby contributing to the synthesis of multiple compounds. For instance, the exclusion of desaturation has the potential to rule out a large proportion of simple, as well as complex, lipids, inclusive of all glycerolipids and phospholipids (Fig. 1). C. elegans contains three delta-9 desaturases, which are enzymes that carry out the first desaturation of 16:0 and 18:0 to 16:1n7 and 18:1n9 and thus generate mono unsaturated FAs with fat-5 encoding a palmitoyl-CoA desaturase and fat-6 and fat-7 encoding stearoyl-CoA desaturases (Fig. 1). Therefore, if the fat-6(tm331); fat-5(tm420) fat-7(wa36) triple mutant, a severe loss-of-function mutant13, lacks the phenotype of interest, this will suggest that unsaturated FA-containing lipids do not mediate the function of the initially identified biosynthetic enzyme (although it will not, of course, exclude their function in other aspects of the same process). FAT-2, the only delta-12 desaturase in C. elegans, catalyzes the first reaction of polyunsaturated FA (PUFA) biosynthesis10. Lack of the phenotype of interest in the nonredundant strong loss-of-function allele fat-2(wa17), which was demonstrated to be completely devoid of normal PUFAs13, would exclude PUFA-containing lipids, for instance eicosanoids. In our study on polarity, we could exclude phosphoinositides – which were prime candidates given their well-described roles in polarity - on the basis of absence of the polarity phenotype in fat-2(wa17)7. In principle, the same concept could be applied for any other type of compound modification (e.g., hydroxylation).
Pathway branchpoints can furthermore be exploited in this way to exclude entire branches. The enzymes that catalyze the first committed (and rate-limiting) steps of sphingolipid (SL) and glycerolipid/glycerophospholipid biosynthesis (Fig. 2a,b), respectively, are serine palmitoyltransferase and glycerol-3-phosphate acyltransferase or dihydroxyacetone phoshate acyltransferase. C. elegans contains three serine palmitoyltransferases (SPTL-1, 2 and 3) and three glycerol-3-phosphate acyltransferases (ACL-4, 5 and 6). The latter have been shown to acetylize both, dihydroxyacetone phosphate and glycero-3-phosphate, which makes them sufficient targets for branch exclusion (Figure 2b, top)26. The absence of the phenotype of interest in triple sptl-1, 2, 3 or acl-4, 5, 6 null mutants would indicate that the corresponding branches (SLs and glycerolipids/glycerophospholipids, respectively) are not involved in the investigated function. Before attempting to construct such triple-null mutants, it should be considered that they might be sterile or early lethal, limiting their usefulness for the tracking of phenotypes occuring at later stages. Moreover, lethal mutants will require balancers (re-introduction of the gene product by providing maternal lipids) and the concomitant depletion of three or more paralogous enzymes is typically achieved by combining germline mutants with RNAi, thus depleting at least one paralog only incompletely. cgt-1(qa1809); cgt-2(tm1192); cgt-3(RNAi) animals, for instance, retain trace amounts of their product GlcCer27. On the other hand, the depletion of several basic lipid biosynthetic enzymes and their products may generate only larval, rather than embryonic, lethality, and the fat-2(wa17) allele, although lacking all normal PUFAs, is viable10.
In principle, any branchpoint further downstream in a biosynthetic pathway can be excluded in the same manner, although branchpoint exclusion may become difficult due to the ability of many complex lipids to interconvert. For instance, the possiblity of mutual interconversion of phosphatidylethanolamine and phosphatidylserine precludes using the phosphatidic acid-processing phospholipid-biosynthetic enzymes phosphatidate phosphatase and cytidine diphosphate-diacylglycerol synthase for branch exclusion of specific phospholipids (Fig. 2). Note that most phospholipids could also be excluded by definitively demonstrating that interrupting the first steps of unsaturated LCFA biosynthesis does not generate the phenotype (see above).
Demonstrating contiguity between upstream and downstream (branch-specific) enzymes will confirm that the initially identified upstream enzymes exert their effects through this branch and are required for the synthesis of its products, while at the same time defining new functions for specific intermediates and delineating the biosynthetic pathway itself. For instance, our polarity screen revealed that mmBCFAs function in polarity, but it also demonstrated that they function as components of SLs7, which is a finding subsequently confirmed by Zhu et al.28.
Evidence for pathway contiguity can be accrued genetically and biochemically. Genetic interactions between hypomorphs, for instance, are expected to show enhancement, and the existence of such interactions would demonstrate that the enzymes operate in the same function but could not prove that they interact in the same pathway. Nulls mutants, which are necessary to address this question rigorously, may not be available given the early essential requirement of many biosynthetic steps. If aspects of the investigated function can be temporally or spatially separated, this can be used to demonstrate functional dependency of upstream on downstream components. Our reversible polarity phenotype, for instance, allowed the temporal separation of its induction and reversion: pathway contiguity could be definitively demonstrated by showing that reversion of a phenotype (induced by depletion of an upstream, FA-biosynthetic enzyme) required the branch-specific downstream enzyme (here, an SL-biosynthetic enzyme)7. Rescue of the phenotype by downstream lipid compounds is another definitive demonstration of pathway contiguity. However, whereas the supplementation of simple FAs generally suffices to rescue upstream enzyme losses, exogenous complex lipids may not fully replace the biological functions of their endogenous counterparts (constraints for rescue include but are not limited to, uptake, species-specific compound composition and chirality, and spatial-subcellular-and temporal availability).
Finally, MS or other quantitative approaches can be used to demonstrate a concomitant relative decrease of the identified branch-specific enzyme’s lipid products in animals depleted of their upstream FA biosynthetic enzymes. Given the possibility of interconnection and complementation between different lipid compounds, alterations of absolute levels of specific compounds must, however, be interpreted with caution.
Once a specific branch for complex lipid biosynthesis has been identified, it is possible to assemble a more comprehensive RNAi library for this specific branch and add: (1) double and triple RNAi conditions to detect redundant enzymes and (2) RNAi sensitive conditions to detect mild effects (the size of libraries remains similar, although the number of targeted reactions is reduced; Table 2 shows a collection of C. elegans SL-biosynthetic enzymes with paralogs). Strong RNAi conditions can be achieved either by using RNAi sensitive strains, such as eri-1(mg366), rrf-3(pk1426) or eri-1(mg366); lin-15B(n744), or by strengthening double-stranded RNA production via increasing the IPTG concentration in agar plates to induce the bacterial polymerase (see ‘RNAi feeding plates’ in Reagent Setup section). Given the abundance of paralogs (Tables 1, ,2),2), the simultaneous perturbation of a redundant gene group using combinatorial RNAi is key. We identified the double knockdown of two CGTs as the endpoint biosynthetic enzymes in our study, and thereby GlcCer as the furthest downstream polarity-affecting lipid species7. A forward RNAi screen is unable to detect these genes, and a forward genetic screen is also unlikely to do so because of the high number of animals that would need to be screened to identify a double hit.
The endpoint biosynthetic enzyme is validated, and its product is shown to mediate the function of the other identified biosynthetic enzymes through a combination of germline mutant analysis, specific inhibitors, MS analysis, genetic interaction experiments and/or exogenous compound rescue, as described above (pathway contiguity, Fig. 3). Definitive proof is rescue of the phenotype generated by loss of upstream biosynthetic enzymes via the endpoint product. This may not always be possible owing to specific requirements for complex lipids for certain functions (see above, pathway contiguity). For instance, C. elegans uses a C17 branched-chained long-chain base on its SLs, which is not readily available for purchase. If the above suggested experiments leave doubt, such compounds must either be purified from C. elegans or be chemically synthesized. Occasionally, bacteria that produce species-compatible lipids have been used for rescue. For example, a C17iso-sphinganine has been isolated from Spingobacterium spiritivorum to rescue worms thst are deficient in serine palmitoyltransferases28, and mmBCFA-deficient worms can be maintained on Stenotrophomonas maltophilia12 (these compounds are not present in OP50, the common C. elegans bacterial food source). Beyond demonstrating absence (reduction) of the presumed lipid end product, MS analysis of wild-type versus mutant animals depeleted of the endpoint biosynthetic enzyme (and/or of other enzymes whose losses produce the phenotype) will detect species-specific features of pathway products, the potential appearance of novel compounds, and it will contribute to the chemical characterisation of the endproduct. Lipid extraction processes and MS instrument setup conditions differ for different lipids- e.g., for SLs, neutral glycerol lipids and glycerolphospholipids29–34. The procedures that are outlined here are targeted to the analysis of C. elegans SLs, and specifically to complex SL species, such as Cer, GlcCer and sphingomyelin.
Complex lipids themselves may be precursors for functionally distinct derivatives. For example, GlcCer, the endproduct identified in our polarity screen, provides the backbone for the large family of GSLs. A lipid glycosylation screen performed in conjunction with a prior lipid biosynthetic pathway screen may thus characterize the specific sugar modification of the identified lipid, and thereby the exact chemical composition of the compound. We used this approach, together with MS analysis, to identify the specific polarity affecting GSL in our study7. In this particular case, three negative results permitted the exclusion of a requirement for further sugar modifications of the identified lipid endproduct, thereby defining it as GlcCer itself: (1) the failure to identify the phenotype in the subsequent glycosylation RNAi screen; (2) its absence in a presumed null mutant of beta4-mannosyltransferase, the first step sugar-biosynthetic enzyme for the arthro-series of GlcCer-derivatives; and (3) the absence of lactosylceramide and consequently the entire GSL lacto-series in C. elegans (demonstrated by MS analysis), restricting the C. elegans GSL repertoire to the arthro series. To further refine the structural characteristics of such an end product, additional chemical characteristics that may have emerged in the prior biosynthetic pathway screens via the requirement of modifying enzymes for the function under study, can be confirmed by MS analysis. For instance, the requirement of a FA hydroxylase and of specific LCFA biosynthetic enzymes for polarity allowed us to characterize the endpoint polarity-affecting lipid compound as a hydroxylated GlcCer with a C17 branched chain LCB and a LCFA with a length of likely C227.
Fatty acids of different lengths, numbers and positions of double bonds, hydroxyl groups and branches (Sigma, Matreya, Larodan, Nu-Chek Prep Inc.)
Glycerolipids (mono-, di- and tri-glycerols; Sigma, Avanti Polar Lipids, Matreya, Larodan)
Phospholipids with different headgroups, such as phosphocholine, phosphoethanolamine, phosphoserine, phosphoinosotol and phosphoglycerol (Avanti Polar Lipids, Matreya, Larodan)
Sphingolipids with different sphingoid bases and lipid moieties (Sigma, Biomol, Avanti Polar Lipids, Matreya, Larodan)
Lipid biosynthetic enzyme inhibitors: for examples of fatty acid-, glycerolipid- and phospholipid biosynthetic enzyme inhibitors and references, see Table 1, and for inhibitors of sphingolipid metabolic enzymes see Table 2 (Sigma, Biomol, Matreya, Toronto Research Chemicals)
Chloroform (Sigma, cat. C7559)
Ethanol (Sigma, cat. no. E7023)
DMSO (Sigma, cat. no. D5879)
C. elegans mutant alleles are available through the Caenorhabditis Genetics Center (http://www.cbs.umn.edu/CGC/) and the National Bioresource Project (http://www.shigen.lab.nig.ac.jp/c.elegans/index.jsp)
RNAi sensitive strains: rrf-3(pk1426), eri-1(mg366), eri-1(mg366); lin-15B(n744) and eri-1(mg366); lin-35(n745), available from the the Caenorhabditis Genetics Center (http://www.cbs.umn.edu/CGC/)
RNAi feeding libraries; Source Bioscience: Ahringer library (19,762 bacterial clones), http://www.lifesciences.sourcebioscience.com/clone-products/non-mammalian/c-elegans/c-elegans-rnai-library/), Vidal ORFeome library (11,804 bacterial clones, http://www.lifesciences.sourcebioscience.com/clone-products/non-mammalian/c-elegans/c-elegans-orfeome-version-11/)
OP50 strain of Escherichia coli (Caenorhabditis Genetics Center, http://www.cbs.umn.edu/CGC/)
Yeast extract (BD Biosciences, cat. no. 211931)
Peptone (BD Biosciences, cat. no. 211677)
Tryptone (Acros Organics, cat. no. 611845000)
Bacto Agar (BD Biosciences, cat. no. 214040)
NaCl (Sigma, cat. no. S7653)
MgSO4 (Sigma, cat. no. M2773)
CaCl2 (Sigma, cat. no. C3881)
KCl (Sigma, cat. no. P9333)
K2HPO4 (Sigma, cat. no. P3786)
KH2PO4 (Sigma, cat. no. P0662)
NaOH (Sigma, cat. no. S5881)
Na2HPO4 (Sigma, cat. no. S7907)
Ammonium acetate (NH4OAC, Sigma, cat. no. 431311)
Acetonitrile (CAN, Sigma, cat. no. 34976)
Methanol (CH3OH, Sigma, cat. no. 34860)
Acetic acid (CH3COOH, Sigma, cat. no. A6283)
Cholesterol (Sigma, cat. no. C8667)
Ampicillin (Sigma, cat. no. A0116)
Cabenicillin (Fisher Scientific, cat. no. BP2648)
Tetracyclin (Fisher Scientific, cat. no. BP912)
IPTG (US Biological, cat. no. I8500)
Sodium hypochlorite solution (4–6% wt/vol)
! Caution this reagent is hazardous and corrosive. Wear protective gloves and a lab coat.
LC-MS instrument, for instance an ABI 4000 Q trap mass spectrometer interfaced to a Dionex U3000 liquid chromatograph equipped with an Astec NH2 4.5 X 150 mm, 5um column.
CentriVap benchtop vacuum concentrator (Savant)
Freeze-drier (Millrock Technology)
Analytical balance (Ohaus)
Probe sonicator (Fisher Scientific Model 505 Sonic Dismembrator)
Ultrasonic water bath (Fisher Scientific Bransonic B-5200R-1)
13 × 75 mm glass test tubes
Dissecting microscope (e.g a Nikon SMZ-U) for phenotype analysis, with or without epifluorescence, depending on the features of the phenotype of interest (e.g. suitable for live analysis of fluorescently-labeled animals on their agar plates). Additional microscopes may be needed, such as one providing, for instance, differential interference contrast (Nomarski; for tracking cell lineage), or a confocal microscope (for analyzing subcellular morphology, e.g., a Leica TCS SL laser-scanning microscope). For detailed phenotypic analysis, high magnification, i.e., 630X or 1000X, is usually required.
Bench-top centrifuge for pelleting bacteria and worms (e.g., Eppendorf 5810R)
37°C shaking incubator to grow bacteria (New Brunswick)
Multichannel pipettes (20–200 ml) for dispensing media and bacterial suspensions
Thermowell sealing tape (aluminum; Corning Costar 6570)
Breatheasy sealing membrane (USA Scientific 9123–6100)
Petri Dishes 60 × 15 mm (Genesee, cat. no. 32–105)
Petri Dishes 100 × 15 mm (Genesee, cat. no. 32–107)
Petri Dishes 150 × 15 mm (Genesee, cat. no. 32–106)
6-well plates (35 × 18 mm, Thomas Scientific 6902A01)
12-well plates (22 × 18 mm, Thomas Scientific 6902A05)
24-well plates (15.5 × 18 mm, Thomas Scientific 6902A09)
48-well plates (9.8 × 18 mm, Thomas Scientific 6902A13)
96-well plates (for freezing RNAi clones, Fisher Scientific 08-772-2C)
96-deep-well plates (for liquid-culture of RNAi clones, Fisher Scientific 08-772-1B)
OmniTray plates (Nalge Nunc 242811)
96-pin replicator (Nalge Nunc 250520)
Conical tubes (50 ml; Fisher Scientific, cat. no. 14-432-22)
Collection tubes (15 ml; Fisher Scientific, cat. no. 05-527-90)
Reagent reservoirs (Corning Costar, 83–4870)
Dissolve lipid or lipid mixture or enzyme inhibitors in an appropriate solvent. Solvents and concentration, as well as preparations vary with different compounds. Examples of stock solutions for some lipids and inhibitors are listed in Table 3.
Mobile Phase A = (97:2:1 CAN: CH3OH: CH3COOH) + 5mM NH4OAC; mobile phase B = (99:1 CH3OH: CH3COOH) + 5 mM NH4OAC. Mobile phases should be freshly prepared daily.
Add 5 g tryptone, 2.5 g yeast extract and 5 g NaCl per 1 liter of water, and autoclave the mixture. Store at 4 °C for 2–3 months. Use directly for growing OP50 bacteria, and add 0.6–1 ml of 100 mg/ml ampicillin for growing RNAi bacteria.
Add 5 g tryptone, 2.5 g yeast extract, 5 g NaCl, 7.5 g agar per 1 liter of water and autoclave the mixture. Allow the medium to cool to 50–60 °C before adding 1 ml of 100 mg/ml ampicillin and 1 ml of 15 mg/ml tetracyclin stock solutions. Pour the medium into plates and leave them at room temperature until the agar sets. Store the plates at 4 °C for 2–3 months. Use for growing RNAi bacteria.
For one liter, add 3 g NaCl, 2.5 g peptone, 17g agar and 975 ml H2O, and autoclave the mixture. Allow the medium to cool to 50–60 °C, and then add 1 ml of cholesterol (5 mg/ml in ethanol), 0.5 ml of 1 M CaCl2, 1 ml of 1 M MgSO4, 25 ml of 1M Potassium phosphate buffer, pH 6. Pour into plates and leave them at room temperature until the agar sets. Store at 4 °C for up to 1 month. Use for growing worms.
For one liter, add 3 g NaCl, 2.5 g peptone, 17g agar and 975 ml H2O and autoclave the mixture. Allow the medium to cool to 50–60 °C, and then add 1 ml cholesterol (5 mg ml−1 in ethanol), 1 ml of 1 M CaCl2, 1 ml of 1 M MgSO4, 25 ml of 1M potassium phosphate buffer, pH 6, 2.4 ml of 200 mg/ml IPTG (final concentration: 2mM), and 0.5 ml of 50 mg/ml carbenicillin (final concentration: 25ug/ml). Pour into plates and leave them at room temperature until the agar sets. Make plates fresh 1–3 days before using for RNAi knockdown by feeding. Different IPTG concentration will produce milder or stronger RNAi effects and should be empirically determined for the phenotype under study.
Disolve 3 g KH2PO4, 6 g Na2HPO4, 5 g NaCl in 1 liter of water and autoclave the mixture. Add 1 ml of 1 M MgSO4 after cooling to room temperature. Use the buffer for washing and aliquotting worms. This solution may be stored at room temperature for months.
Add 25 ml of 10 N NaOH, 15 ml of household sodium hypochlorite solution and 60 ml water to make 100 ml. Bleach solutions lose activity over time. Freshly prepare the solution before using it for synchronizing worms.
Add 147.01 g of CaCl2 to 500 ml of deionized water and stir until CaCl2 is completely dissolved, and then bring the volume to 1 liter using deionized water, and autoclave the mixture. Allow the solution to cool to room temperature. This solution may be stored indefinitely at room temperature.
Add 246.48 g of MgSO4 to 500 ml of deionized water, and stir it until MgSO4 is completely dissolved, and then bring the volume to 1 liter using deionized water and autoclave the mixture. Allow the solution to cool to room temperature. This solution may be stored indefinitely at room temperature.
Add 30.1 g of K2HPO4 and 118.1 g of KH2PO4 to 500 ml of deionized water, and stir it until K2HPO4 and KH2PO4 are completely dissolved, and then bring the volume to 1 liter using deionized water and autoclave the mixture. Allow the solution to cool to room temperature. This solution may be stored indefinitely at room temperature.
Add 20 g of NaOH to 50 ml of deionized water, and stir it until NaOH is completely dissolved. This solution may be stored indefinitely at room temperature and in plastic containers.
Add 1 g of ampicillin to 10 ml of deionized water to make stock solution. Sterilize the stock solution by filtering with a 0.22 um filter and store 1-ml aliquots in 1.5-ml tubes at − 20 °C for 6–12 months.
Add 500 mg of carbenicillin to 10 ml of deionized water to make stock solution. Sterilize the stock solution by filtering with a 0.22- um filter, and store 1-ml aliquots in 1.5-ml tubes at −20 °C for 6–12 months.
Add 150 mg of tetracyclin to 10 ml of ethanol to make stock solution. Sterilize the stock solution by filtering with a 0.22-um filter, and store 1-ml aliquots in 1.5-ml tubes at − 20 °C for 6–12 months.
Add 2 g of IPTG to 10 ml of deionized water to make stock solution. Sterilize the stock solution by filtering with a 0.22-um filter and store 1-ml aliquots in 1.5-ml tubes at − 20 °C for 6–12 months.
The protocol comprises the following steps: a 2-min column pre-equilibration in 9:1 A:B (vol/vol); sample injection; 2-min wash with 100% A; 12-min linear graddient to 100% B; and 1-min post-run column equilibration. The gradient profile is tabulated below, with a flow rate of 1 ml/min and an injection volume of 5 ul.
PRE-SCREEN CONFIRMATION AND ANALYSIS:
SCREEN I: BROAD BIOSYNTHETIC PATHWAY RNAi SCREEN:
CONFIRMATION AND ANALYSIS OF RESULTS OF SCREEN I (optional)
SCREEN II: COMPREHENSIVE PATHWAY BRANCH RNAi SCREEN
CONFIRMATION AND ANALYSIS OF RESULTS OF SCREEN II
SCREEN III: METABOLIC PATHWAY RNAi SCREEN FOR DERIVATIVES (optional)
CONFIRMATION AND ANALYSIS OF RESULTS OF SCREEN III
In this era of fully sequenced genomes, biosynthetic enzyme genes are frequently identified in genome-wide screens (e.g., RNAi screens) that investigate specific cell biological functions. This information is often excluded from further analysis, because such ‘housekeeping genes’ typically contribute to the synthesis of a vast array of different compounds. Systematic follow-on pathway screens, as described here, are expected to identify the one specific metabolite that mediates their contribution to the investigated function, and to additionally map out the entire pathway leading to its synthesis. Such screens will therefore also identify multiple intermediate enzymes and their respective products as required for the function under study. With sufficient screen depth and the addition of screens for derivatives, biosynthetic pathway screens, together with MS-based biochemical analyses, are able to identify not only the compound species whose loss causes the functional defect, but also its exact chemical characteristics. Biosynthetic pathway screens can also be devised upfront as ‘metabolic RNAi screens’ to investigate the contribution of nonproteinaceious compounds to a specific biological function. The prime purpose of these screens is the discovery of novel, non-metabolic functions of nontemplate-derived metabolites such as lipids (themselves no targets for knockouts or knockdowns), although they also extend the already well-established knowledge about their metabolic functions. It is hoped that such screens will harness the power of unbiased genetic/RNAi screens to open the field of metabolomics for the in vivo characterization of the largely unknown non-metabolic roles of biosynthetic intermediates.
C. elegans strains were provided by G. Ruvkun (Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts, USA), S. Mitani (National Bioresource Project Japan) and the Caenorhabditis Genetics Center (NIH Center for Research Resources). We thank J. Moore (Avanti Polar Lipids) and Edward Membreno for contributions to LC/MS and C. elegans maintenance, respectively. We thank H. Weinstein and R. Kleinman for ongoing support. This work was supported by US National Institute of Health grant GM078653, Massachusetts General Hospital IS Funding and a Mattina R. Proctor Award to V.G.
Note: Any supplementary Information and Source Data files are available in the online version of the paper