Abstract
Juvenile CLN3 disease is a recessively inherited paediatric neurodegenerative disorder, with most patients homozygous for a 1-kb intragenic deletion in CLN3. The btn1 gene is the Schizosaccharomyces pombe orthologue of CLN3. Here, we have extended the use of synthetic genetic array (SGA) analyses to delineate functional signatures for two different disease-causing mutations in addition to complete deletion of btn1. We show that genetic-interaction signatures can differ for mutations in the same gene, which helps to dissect their distinct functional effects. The mutation equivalent to the minor transcript arising from the 1-kb deletion (btn1102–208del) shows a distinct interaction pattern. Taken together, our results imply that the minor 1-kb deletion transcript has three consequences for CLN3: to both lose and retain some inherent functions and to acquire abnormal characteristics. This has particular implications for the therapeutic development of juvenile CLN3 disease. In addition, this proof of concept could be applied to conserved genes for other mendelian disorders or any gene of interest, aiding in the dissection of their functional domains, unpacking the global consequences of disease pathogenesis, and clarifying genotype–phenotype correlations. In doing so, this detail will enhance the goals of personalised medicine to improve treatment outcomes and reduce adverse events.
Similar content being viewed by others
Introduction
Yeast provide useful model systems for many human diseases due their genetic tractability. We have developed yeast models to understand the neuronal ceroid lipofuscinoses (NCL), also known as Batten disease, a group of rare childhood inherited neurodegenerative disorders1. Most are inherited in an autosomal recessive manner, with mutations in 13 known human genes2, some of which are conserved in yeast. These diseases share the characteristic hallmarks of accumulation of autofluorescent material, lipofuscin/ceroid, in the lysosome3. The most prevalent NCL is juvenile CLN3 disease, with most patients carrying an intragenic 1-kb deletion of the gene CLN3 on at least one of the two disease alleles4,5.
In this study, we extended application of the powerful synthetic genetic array technique using Schizosaccharomyces pombe to perform three independent systematic screens for three mutant strains of the orthologue of CLN3, btn1. These strains were (1) btn1 encoding a mutation mimicking one transcriptional effect of the 1-kb deletion (btn1102–208del)6,7, (2) btn1 encoding the equivalent of disease-associated missense mutation p.Asp416Gly (btn1D363G)6, (3), and complete deletion of btn1 (btn1Δ)8.
Our objectives were to explore the genetic interactions of two disease-causing mutations, compare them to complete loss of the gene, identify proteins and key pathways that are involved in Btn1 function, and thereby indicate novel human orthologues that may contribute to CLN3 disease pathology and provide novel therapeutic targets. By dissecting functional regions, we have shown differences in the genetic interaction signatures for three distinct mutations in the same gene. The mutant equivalent to the most common mutation, the ‘1-kb deletion’, does not cause complete loss of function, but produces a complex partially functional Btn1 protein with a unique gain of new characteristics. This approach could be extended to understand functional consequences of disease-associated alleles in other genes with yeast orthologues, in addition to more generally dissecting the functional domains in any protein of interest.
Results
SGA analysis reveals overlapping genetic interactions for strains btn1Δ and btn1 D363G, and many novel genetic interactions for btn1 102–208del
Genetic interactions of three btn1 mutant strains
SGA screening was performed for the ade6 control and three btn1 mutants (btn1∆, btn1D363G and btn1102–208del). Each Bioneer library mutant was crossed with each query strain independently 3 times and pinned in quadruplicate, providing a total of 12 replicate colonies for each double mutant (Fig. 1A). The ade6∆ control showed similar growth distributions across all three independent experiments (Fig. 1B). Principal component analysis revealed good separation of query strains representative of their biological-signatures following normalisation of plate and batch effects (Fig. 1C). Cluster analysis of the btn1 strains show similarities between btn1∆ and btn1D363G and greater separation for btn1102–208del for double mutant colony sizes after normalising for batch effects (Fig. 1D). Distribution of normalised colony sizes show decreased colony fitness close to the loci of query strains due to genetic linkage, genes within ± 500 kb range of the loci were excluded from our data (Fig. 1E).
General overview of SGA analysis of btn1 mutants versus ade6 control. (A) Representative images of the SGA plates for control (ade6) and query mutants (btn1∆, btn1D363G, btn1102–208del), with empty control quadrants shown for ade6 (yellow boxes). (B) Exclusion of small colonies for ade6 control across batches as they represent high variability therefore reducing noise. (C) Principle component biplot of the variance within the SGA data for ade6 control (yellow) and query-mutants btn1∆ (blue), btn1D363G (orange), btn1102–208del (red), with experimental batch indicated. (D) Cluster analysis for each strain and all the genes with their normalised colony size difference against ade6 control with batch effects removed. Interactions are coloured in blue for negative interactions (< − 0.5) and yellow for positive interactions (> 0.5). (E) Gene linkage of normalised fitness score for ade6 control and query mutants btn1∆, btn1D363G, btn1102–208del from one experiment. Vertical dashed line represents ade6 or btn1 gene location, red points represent interaction scores excluded from data since less than 500 kb/500,000 bps from query gene location.
Interactions > 0.05 adjusted p-value across triplicate experiments were considered significant and placed into subsets of positive and negative interactions; the number of interactions is summarised in Table 1 and Fig. 2A (complete lists in supplementary tables S2, S3 and S4). btn1Δ has 76 positive interactions and 68 negative interactions, btn1102–208del has 129 positive interactions and 178 negative interactions, and btn1D363G has 75 positive interactions and 47 negative interactions. There are 84 interactions shared between btn1∆ and btn1D363G but only 36 consistent are robust hits across all three query strains (24 negative and 12 positive interactions). A summary of the differences between strains is presented in S5: supplementary tables.
Venn diagrams and comparative volcano plots for strains btn1∆, btn1102–208del (1-kb) and btn1D363G (D363G) against control ade6. (A) Numbers represent robust hits across three independent SGAs for each query mutant. These include the subsets of genes that are shared between two or all mutants as well as their unique interactions. (B–D) Colony size difference plotted for each gene and each strain against the ade6 control respectively (B) btn1∆, (C) btn1102–208del (D) btn1D363G. Every gene had quad intra-repeats and the experiment was done in triplicate. All genes are plotted on the adjusted p-value on a log scale. The strongest interacting genes are highlighted as negative (red) or positive (green) interactors.
As expected, there is considerable overlap between btn1∆ and btn1D363G for their top interactions (Table 2). In contrast, the strongest interactors of btn1102–208del overlap only two negative and one positive interaction with btn1∆ and btn1D363G (pfa3, kes1 and cpp1, respectively). The remaining strongest interactions for btn1102–208del are unique.
Asp363 in the C terminus of Btn1 is a critical residue for function
The most significant interacting genes for btn1D363G considerably overlap with those of btn1∆, with few interactions unique to each strain (Tables 1 and 2, Fig. 2). This is visualised by the PCA biplot (Figs. 1C, 3A). The shared interactions include 34 negative and 50 positive interactions. This suggests that the C-terminal missense mutation p.Asp363Gly leads to a near non-functional protein similar to complete lack of Btn1, indicating that Asp363 is a key functional residue. The equivalent human mutation, p.Asp416Gly may have the same drastic loss of function effect on CLN39. Although other small differences cannot be ruled out, the main difference between the strains btn1∆ and btn1D363G may be the lack of a transcribed protein in btn1∆ in contrast to production of a non-functional protein in btn1D363G. Indeed, there is a relative increase in transcriptional levels for btn1 in both btn1D363G and btn1102–208del compared to WT (S1: supplementary figure 4C).
Comparative 2D-volcano plots for SGA queries of strains btn1∆ and btn1D363G. (A) Biplot of colony size difference between btn1∆ vs ade6∆ on the x-axis and btn1D363G vs ade6∆ on the y-axis for each gene. Clustering of genes along the diagonal line highlights the similarities between the strains. Upper right quadrant represents shared positive interactions, lower left quadrant represents shared negative interactions. (B) Biplot of colony size difference between btn1∆ vs ade6∆ on the x-axis and btn1D363G vs btn1∆ on the y-axis. Clustering of most genes around the centre is due to the similarities between the two strains. Upper left and lower right quadrants represent residual functionalities. Genes away from the centre/at the extremes represent biological differences between the two strains. Gene points are represented by their max adjusted p-value score, with a logarithmic colour distribution.
Novel functionality is associated with the 1-kb deletion mutant protein
In contrast, the btn1102–208del strain gives an overall global genetic interaction signature that is markedly distinct from the strains btn1Δ and btn1D363G, visualised in the separation between this strain and btn1Δ or btn1D363G in the PCA biplot (Fig. 1C). There are many more interacting genes identified (more than double that for btn1Δ or btn1D363G), and most of these are unique for this strain (Table 1). These interacting genes are therefore not a simple subset of those highlighted when the function of btn1 is lost but primarily comprise a large and novel set of genes, with only 17% of all btn1102–208del hits overlapping with btn1∆ (split 66% for negative and 33% for positive interactions). In contrast, 68% of btn1D363G interactions overlap with btn1Δ. The direction of movement away from btn1∆ and towards ade6∆ in the PCA plot, suggests strongly that btn1102–208del generates a mutant Btn1 protein that may retain some functionality associated with Btn1, while the tandem moving away from both ade6∆ and btn1∆ strains, and little overlap in genetic interactors with either of these strains, suggests the gain of new functionality (Fig. 1C).
SGA analysis highlights multiple biological processes associated with Btn1 function
Btn1 function supports protein translation and trafficking
Considering all positive interactions for btn1∆ and btn1D363G, there is an enrichment of GO terms for ribosomes (Table 3), suggesting that a reduction of translation is beneficial when Btn1 function is lost. However, only one gene encoding a ribosomal protein, rpl2301, is identified in the top five strong interactors. The remaining interactions for btn1∆ are drawn from other cellular physiology pathways, particularly trafficking through the Golgi apparatus and endosomal compartments (apt1, blt1, cpp1, grx5, cfr1 and alg12). There is no particular enrichment in GO terms for negative interactions of btn1∆ and btn1D363G. In contrast, btn1102–208del negative interactions are enriched for mitophagy in yeast while positive interactions are enriched for autophagy and pyruvate metabolism (all terms associated with strains are listed in the S6: supplementary tables).
Palmitoylation becomes an essential function in the absence of full Btn1 functionality
The strongest negative interactor of btn1∆, pfa3, is shared by all three strains (Fig. 2B–D), indicating that loss of function of this gene is detrimental both in the absence of Btn1 function or the presence of mutant Btn1102–208del. pfa3 encodes a palmitoyltransferase that catalyses post-translational attachment of fatty acid palmitate to proteins via a cysteine residue. There can be diverse and severe consequences from dysfunctional palmitoylation and depalmitoylation, and insufficiency of the palmitoyl thioesterase Ppt1, defective in CLN1 disease10, has been observed in cells from CLN3 disease patients and a CLN3 mouse model11,12.
Btn1 connects with the ubiquitin protease system
Although btn1∆ and btn1D363G strains have a similar genetic interaction pattern (Fig. 3), ubi4 is a strong (synthetically lethal) negative interaction for btn1∆ but not for btn1D363G (Fig. 3). The expression of the non-functional Btn1D363G protein is able to rescue the negative interaction between btn1∆ and ubi4∆. ubi4 encodes a polyubiquitin protein precursor required in the response to stress and whose absence affects many pathways including meiosis13,14.
Unpacking the complexity of btn1 102–208del to shed light on juvenile CLN3 disease
Comparison of btn1102–208del to the ade6Δ control alone reveals only a portion of the likely complexity of the 1-kb deletion on CLN3 function. By additionally comparing btn1102–208del to btn1Δ (Fig. 4), we are able to dissect the interactions into specific subsets; this new information reveals the consequence of this mutation and is relevant to understanding the impact of the 1-kb deletion in CLN3 disease. These include shared positive and negative interactions and gain of function of btn1102–208del relative to strains btn1∆ or btn1D363G.
Comparative 2D-volcano plots for SGA queries of strains btn1102–208del against btn1∆ and btn1D363G. (A) Biplot of colony size difference between btn1∆ vs ade6∆ on the x-axis and btn1102–208del vs ade6∆ on the y-axis. Upper right quadrant represents shared positive interactions, lower left quadrant represents shared negative interactions. (B) Biplot of colony size difference between btn1∆ vs ade6∆ on the x-axis and btn1102–208del vs btn1∆ on the y-axis. Upper left and lower right quadrants represent residual functionalities in btn1102–208del relative to btn1∆. Upper right and lower left quadrants represent gain of functionalities in btn1102–208del relative to btn1∆. (C) Biplot of colony size difference between btn1D363G vs ade6∆ on the x-axis and btn1102–208del vs ade6∆ on the y-axis. Upper right quadrant represents shared positive interactions, lower left quadrant represents shared negative interactions. (D) Biplot of colony size difference between btn1D363G vs ade6∆ on the x-axis and btn1102–208del vs btn1D363G on the y-axis. Upper left and lower right quadrants represent residual functionalities in btn1102–208del relative to btn1D363G. Upper right and lower left quadrants represent gain of function relative to btn1D363G. Gene points are represented by their max adjusted p-value score and a logarithmic colour distribution with significant genes annotated. For (A) and (C) genes away from the diagonal represent biological differences between the two strains. For (B) and (D) genes away from the centre/at the extremes represent biological differences between the two strains.
Btn1 102–208del loses functions that are also missing in btn1∆
The consequences driven by the loss of amino acids 102–208 of Btn1 can be compared with those of btn1∆. The same effect on colony size indicates interactions that are shared between btn1102–208del and btn1∆ (Fig. 4A). Therefore, these genetic interactions must be associated with the loss of functionality caused by deletion of amino acids 102–208. Positive interactions shared by btn1102–208del and btn1∆ are enriched for N-glycan biosynthesis (alg9, alg12), and protein farnesylation (cpp1) (S7: supplementary tables for common lost hits for 1-kb). Shared negative interactions are enriched for sterol lipids and trafficking (pfa3 and kes1).
Btn1 102–208del loses functions that are also missing in expressed Btn1 D363G
As above, the consequences driven by btn1102–208del can be compared with those of btn1D363G (Fig. 4C). Both strains express mutant Btn1 proteins with p.Asp363Gly nearly equivalent to complete absence of Btn1 function (see above). As expected, the same effect on colony size is shared by many of interactions of btn1∆ and btn1102–208del and the interactions of btn1D363G and btn1102–208del. This indicates loss of the same functions for Btn102–208del and Btn1D363G.
Btn1 102–208del exhibits functions that are missing in btn1∆
Through characterising differences between btn1102–208del and btn1∆, we observe a broad set of genes that show synthetic sickness in btn1∆, but for which fitness is at least partially restored in the btn1102–208del (Fig. 1D). The converse is true for a set of btn1Δ positive interaction (Fig. 1D). That some of the interactions of btn1Δ are reversed in btn1102–208del indicates that there is a partially functional Btn1 protein in the btn1102–208del mutant. To identify this set of genes we use comparative 2D volcano plots to separate residual restorative functionality and increased sensitivity in the btn1102–208del strain (Fig. 4B). btn1102–208del maintains residual Btn1 functionality, represented by opposing genetic interactions from the btn1∆ perspective. These genes are clustered in the top left and bottom right quadrants of Fig. 4B and show the restorative positive and negative interactions by btn1102–208del (full list of genes in S8: supplementary tables). We postulate that these observations are a consequence of normal functional domains remaining in the Btn1102–208del protein, thus effectively reverting colony fitness back to that of the ade6 control. Similar observations are seen when btn1102–208del was compared with btn1D363G (Fig. 4D).
The gene, kgd1, encoding 2-oxoglutarate dehydrogenase15, has the strongest negative statistical difference between btn1102–208del and btn1∆ (and one of the largest differences when comparing to btn1D363G). Indicative of the complexity of the 1-kb deletion it exemplifies a complete reversal of interactions between btn1102–208del and btn1∆ on protein function.
Expressed Btn1 102–208del gains novel functions that are not present in wildtype Btn1
There are two main approaches to determining gain of functions: (1) interactions that show increased severity in one mutant compared to the other, and (2) interactions which are not present in the ablated query but are present in the other mutant query. A marked feature of our dataset is the complex nature of the loss of amino acids 102–208 on Btn1 protein function, revealed in Fig. 4B, contrasting with the effect of the missense mutation p.Asp363Gly displayed in Fig. 3B. We confirm btn1102–208del increases sensitivity for genetic interactions compared to those of btn1∆ within the bottom left and top right quadrants. For example, our data sets reveal two negative interactions within this subset, (gga22, nce103) and one positive interaction (tfx1) S9: supplementary tables. These represent a potential gain of function by loss of amino acids 102–208. In addition, we identify interactions that are unique to btn1102–208del by using the criteria representing a significant difference in colony size between btn1102–208del and btn1∆ but no difference in colony size for btn1∆ vs ade6∆ interactions (Fig. 5). We can conclude that the Btn1102–208del mutant protein has a clear gain of function represented by both unique positive and negative interactions (summarised in Fig. 5).
Gain of function interactions unique to btn1102–208del. This 2D volcano plot of colony size difference between btn1∆ and ade6∆ (x-axis) against colony size difference between btn1∆ vs btn1102–208del (y-axis) highlights the gain of function interactions unique to btn1102–208del. Points highlighted around the vertical axis represent genes that had no colony difference between btn1∆ vs ade6∆ control, however were significant for btn1102–208del. For simplicity any interaction that did not meet the criteria for unique gain of function was changed to 1 adjusted p-value (yellow). Gene points are represented by their adjusted p-value score for interaction btn1102–208del vs ade6∆, colour represented by a logarithmic colour distribution with significant genes (adjusted p-value < 1e−3) annotated. Full list of genetic interactions in S10: supplementary tables.
Btn1 102–208del links to genes implicated in other diseases
Many of the interactions for the three btn1 strains link to human disease (Table 4). For the unique negative interactions of btn1102–208del, where introduction of this mutation into cells deleted for another gene causes these cells to become synthetically sick, the interacting genes are linked to eye diseases (gal1, aim22, fab1, ath1, rad24), epilepsies (tef103, rad24), inherited inborn mitochondrial (hem25, mss1), neurological (sst2, rav1) immune system (ski3) and lipid metabolism (erg32) disorders. For the unique positive interactions, where introduction of btn1102–208del into cells deleted for another gene causes cells to grow better, these genes are also linked to eye disorders (gfh1, hnt3) as well as lysosomal disorders (npc2, SPBC713.07c, SPBC1683.12), monogenic diseases (hri2, kms1, cds1), neurological (hhp2, SPCC1827.07c), and inborn mitochondrial metabolomic (cyt1, fsf1, SPBC8D2.18c) disorders. This suggests a shared compensatory network between these varied disorders and CLN3 disease (Table 5).
Btn1 102–208del novel functions link with multiple pathways
We investigated the KEGG networks for the btn1102–208del unique gain of function interactions, generating cytoscape networks with ClueGO integration (Fig. 6B). Negative interactions localise in endocytosis (sst2, vps28, SPCC794.11c), mitophagy (fis1, mcs4, pek1), and the MAPK signalling pathway (rad24, mcs4, pek1). Positive interactions involve autophagy (atg3, atg8), pyruvate metabolism (glo1, SPAC1952.09c) and peroxisome function (fap1, SPBC16A3.14). In addition, there is significant enrichment for GO cellular compartment terms that relate to integral components for peroxisomes for the negative interactions (fis1, inp2). Biological processes for positive interactions enrich for regulation of ion transmembrane transport (ncs1, gti1), divalent inorganic cation transport (fet4, SPAC17A2.14), and anion transporter activity (SPCC1827.07c, SPBC1683.12, SPBC1271.09, hut1, SPAC17G6.15c, SPCPB1C11.02); those for negative interactions include protein tyrosine kinase activity (hhp1, hhp2 pek1) and ATP dependent chromatin remodelling (arp5, SPCC16C4.20c, rsc1). Some GO terms have both positive and negative interactions: these relate to cell cycle checkpoint [positive (cds1, rum1, hhp2)/negative (hhp1, rad24)] and kinase activity (positive (ncs1, rum1)/negative (tif51, mug80, mcs4) (Fig. 6A).
Genetic network of unique btn1102–208del interactions. (A) Genetic networks of unique btn1102–208del interactions represented in terms of positive (green) and negative (red) interacting gene and their corresponding biological processes GO term. Grey GO terms represent terms with both interactions, with a scale of red or green dependent on the number of interactions of either side associated with that specific GO term. (B) Genetic networks of unique btn1102–208del interactions represented in terms of positive (green) and negative (red) interacting gene and their corresponding KEGG GO term. Kappa score = 0.4. Generate using ClueGO51.
Discussion
Through a novel application of SGAs we compared genetic-interaction signatures and the global effects of two different disease-causing mutations (btn1102–208del and btn1D363G) with complete loss of btn1. Our analysis reveals that genetic-interaction signatures can be specific for mutations in the same gene, which helps in the dissection of their distinct functional effects.
We had already demonstrated a role for Btn1 in TOR signalling and vacuole homeostasis, and clear involvement in multiple pathways in S. pombe6,8,16,17. This work confirms these previous findings and additionally reveals that Btn1 contributes to trafficking and could be important for ubiquitin–proteasome system (UPS) regulation. We also demonstrate that the residue Asp363 (in H.s CLN3 Asp416) is crucial, as a mutant protein containing p.Asp363Gly is equivalent to complete absence of btn1.
A significant finding is that the mutant protein Btn1102–208del, which is equivalent to a transcript of the most common pathogenic mutation in CLN3, the '1-kb deletion', has lost significant Btn1 functions but is not equivalent to complete absence of all Btn1 functions. Further, this mutant protein, which is missing amino acid residues 102–208 (equivalent to CLN3 residues 154–259), also acquires novel functions, as well as retaining some functions of Btn1.
These findings have important implications in terms of understanding the consequences of the 1-kb deletion on juvenile CLN3 disease pathogenic mechanisms and in designing therapeutic interventions. There is much variation in transcription in the brain18, the site of significant CLN3 disease pathology, and this can have functional consequences19. Variant CLN3 transcripts are known that lack exons, some of which affect phasing, and further novel isoforms arise in the presence of the 1-kb deletion7. Therefore, questions remain as to whether the observed disease pathogenesis is due to the specific partial loss of CLN3 function associated with a particular mutant transcript or includes the acquisition of new characteristics, as implied by this work, and whether the disease severity correlates with the prevalence of particular variant mutant transcripts. This is not known for the CLN3 human disease, but an increased level of btn1 mutant transcripts were observed in both btn1D363G and btn1102–208del compared to wildtype btn1. If CLN3 disease is directly related to increased expression of one or more mutant variant transcripts, it may be that manipulation of their concentration could reduce or exacerbate the disease. Therapeutically, introduction of a full length CLN3 transcript may compensate only for the loss of function. Novel strategies would need to be developed for aspects of disease arising from acquired characteristics. It cannot yet be ruled out that high expression of the mutant transcripts is a compensatory mechanism due to less functional or non-functional transcripts. It is possible that having high levels of the partially-functional 1-kb protein is responsible for the gain of function. Introducing a full-length transcript might remove these gains of functions by repressing transcription of the native 1-kb transcript.
Through the enhanced use of the SGA with S. pombe we have extended the utilisation of this powerful technique beyond complete loss of function. The advantages of conducting in-depth analysis of multiple mutants within a single gene are clear with the delineation of novel distinctions between btn1D363G, btn1102–208del and btn1∆. Here, comparative analyses of the interactions of all three query strains show a trifecta of consequences of the Btn1102–208del protein that can be confidently divided into three distinct categories: partial loss of Btn1 functions, retention of functions and gain of novel functions. There is an advantage in using S. pombe to delineate genetic interactions as there is less functional redundancy in its genome compared to S. cerevisiae20, which will have contributed to cleaner phenotypic consequences and aided interpretation; such functional insights may otherwise have been obscured through the use of other comparative experimental approaches such as transcriptomics or proteomics. Further, this approach avoids the constraints of these methods where it is not always possible to correlate transcript isoforms and protein expression levels with phenotype.
In terms of understanding the fundamental function of Btn1, genes that are positive phenotypic suppressors for both btn1Δ and btn1D363G, but not btn1102–208del, can reflect physical interactions with Btn1. Of interest, we identified sdo1 as a positive interaction. This encodes a protein involved in ribosome maturation and is the orthologue of SBDS which causes Shwachman–Bodian–Diamond syndrome21. It has previously been reported that S. cerevisiae Btn1p protein binds the protein Sdo1p and overexpression of BTN1 could compensate for loss of SDO122.
This work has highlighted that btn1 is essential for the survival of several genetically deficient cells. One of these is ubi4∆, yet crossing with btn1D363G, a mutant strain with a very similar genetic-interaction signature to btn1∆ (thereby inferring that Btn1D363G protein is non-functional), was sufficient to rescue this strain. The polyubiquitin protein precursor ubi4 is key to normal cellular physiological processes and to stress survival14. The multi-unit structure of Ubi4 enables the fast release of free ubiquitin and rapid response to acute stress such as oxidative stress, heat stress or toxicity that activates the unfolded protein response (UPR)13,23,24. It is therefore intriguing that expression of Btn1D363G is able to ameliorate this synthetically sick interaction arising from complete loss of the btn1 gene in btn1∆. Expression of Btn1D363G, even though essentially non-functional, could explain differences observed between the pleiotropic phenotypes of btn1∆ and btn1D363G strains (data unpublished). Further, this unsuspected role of Btn1, that we suggest is secondary to its main functional role, may be mirrored in its human orthologue.
One of the strongest interactors of all the btn1 mutants was pfa3, which encodes a protein-cysteine S-palmitoyltransferase. Palmitoylation is essential for many biological processes such as neuronal development, particularly relevant to the NCLs25,26. Disruption of palmitoylation can lead to neurological defects27. Pfa3 has at least four homologues in humans, the zinc finger DHHC domain-containing palmitoyltransferases, ZDHHC2, ZDHHC20, ZDHHC15, and ZDHHC21, none of which are well characterised27. However, ZDHHC15 is highly expressed in many central nervous system cell types, with neurons exhibiting shorter dendrites when ZDHHC15 is depleted28. Intriguingly, ZDHHC2 was highlighted in a recent translatome-regulatory network analysis as a neuroprotective protein when overexpressed in line with our findings and is down-regulated in incipient Parkinson's disease29.
The full and complex consequences of the ‘1-kb deletion’-equivalent btn1102–208del are revealed for the first time. This conceptual increment in genetic understanding allows inference of the likely consequences of the 1-kb deletion in juvenile CLN3 disease. No existing mouse or zebrafish CLN3 models mimic only the acquisition of novel dominant functionality by 1-kb transcripts; btn1102–208del is the first to mimic the minor transcript seen in a patient’s fibroblasts7. Further, many unique genetic interactions of btn1102–208del correspond to conserved human orthologues, including genes related to eye and neurological diseases, for example, npc2 (Niemann–Pick C disease), linking these diseases with the pathogenesis of CLN3 disease. Our enrichment of unique negative interactions of btn1102–208del suggests that the ‘1-kb deletion’ may be detrimental to cells sensitive to aberrant mitophagy. This is not surprising given the accumulation of subunit c of mitochondrial ATP synthase, a hallmark of CLN3 disease and other NCLs30,31.
Our results also pinpoint potential novel interactions specific to the minor 1 kb transcript. Positive interactions in this subset of genes may be useful in developing specific therapeutic targets that would ameliorate phenotypes solely arising from this mutated transcript. Disease caused by other mutations in CLN3 may not respond to these tailored therapeutics.
Notably, our analysis has identified several genes of interest for further study, such as kgd1 and glo1. Many of these reveal a web of interlocking connections common across multiple neurological diseases32,33. When we compared our results with the orthologues of biomarkers for juvenile CLN3 disease collated from murine microarray data34,35, patients homozygous for the 1-kb deletion36 and SH-SY5Y Human Neuroblastoma cells37 (S13: supplementary table), five genes, pfa3, kes1, myo52, efc25 and cox17, were found to be common.
Conclusion
This proof of concept extension of the SGA described here has revealed novel and unexpectedly complex findings for btn1 with implications for therapeutic development for CLN3 disease. The use of SGA in this manner, focusing on an individual gene with mutations selected for their functional or pathogenic consequences could reveal novel findings for many genes. This approach applied to conserved genes, particularly those related to rare mendelian diseases but also genes associated to more common diseases such as Parkinson’s and Alzheimer’s, will aid the dissection of their functional domains and consequences of disease pathogenesis. Moreover, it will enable further correlations with genotype and enhance the goals of personalised medicine to improve treatment outcomes and reduce adverse events.
Methods
Routine yeast methods
Yeast cells were grown in YES broth (Formedium) and incubated for a period of 24hrs at 30 °C. Strains were also grown on YES Agar (Formedium) plates and stored at 2–8 °C. Query strains generated and used: btn1∆, btn1102–208del and btn1D363G.
Generation of query strains
Schizosaccharomyces pombe query strains were generated as follows using primers in Table 6. For btn1∆ (h-, leu1-31, ura4-D18, ade6-M210, btn1::NatMX), a wildtype strain (ALU-) was used as parental strain and the NatMX cassette was introduced through homologous recombination by transformation of a purified PCR product containing NatMX and flanking regions. Primers were designed (using pFa6a-NatMX plasmid as a PCR template) so that the endogenous btn1 gene was swapped for the PCR-amplified NatMX cassette, containing 50 bp homologous regions flanking btn1. For btn1102–208del, a btn1::btn1102–208del strain (ALU-) was used as parental strain and the NatMX cassette was introduced at the C-terminus via homologous recombination. For btn1D363G query strain, a btn1::btn1D363G parental strain (ALU-) was transformed with a PCR-product containing the NatMX-cassette and 50 bp flanking regions of the C-terminus of btn1. All strains were confirmed through agarose gel size comparison with wildtype btn1-strain. Genomic DNA was extracted from all query strains and the btn1-regions were PCR-amplified with primers upstream of btn1 and in the NatMX cassette. PCR products were separated on an agarose gel and compared to wildtype-btn1. Additionally, strains were confirmed through sequencing: briefly, primers were designed upstream of the mutated btn1 genes and in the NatMX cassette and resulting PCR fragments were sequenced with various sequencing primers to cover the D363G mutation and the 102–208del regions. Transformations of the parental strains were performed using standard protocols using lithium acetate38, followed by selection of positive clones on YES plates containing ClonNat (100 μg/ml).
Yeast RNA extraction
After overnight growth in 15 ml of YES media, cells were broken up using glass beads (0.5 mm diameter) with buffer made of RTL buffer + beta-mercaptoethanol (10 μl in 1 ml of RLT buffer). Following this the cells were vortexed in the cold room at 4 °C for 20 min. To elute the cells contents into a new microfuge tube, the side and bottom of the tube was pierced, followed by centrifugation for 5 s. Spin for 2 min for 10,000 rpm. Transfer 350 μl of lysate was transferred to the RNA column (Qiagen) and the Qiagen protocol (RNeasy mini kit) was followed to purify the RNA in 30 μl Rnase free water.
RNA reverse transcription
Extracted RNA was retro-transcribed using the QuantiTect reverse transcription kit (Qiagen). Reaction tubes were incubated at 42 °C for 15 min and inactivated at 95 °C for 3 min. Samples were used immediately. Regular PCR was performed to verify primers worked efficiently.
RT-qPCR
cDNA obtained from the reverse transcription of RNA extracted was submitted to RT-qPCR to see the levels of btn1 mutant transcripts using primers in Table 7. The housekeeping gene actin is used as a reference. Data was analysed using the 2−∆∆C(T) method39.
Generation of double mutant libraries
Three independent biological replicates of each SGA were performed with the Bioneer V5 deletion collection40 using a ROTOR HDA pinning robot (Singer Instruments) as previously described41. Briefly, the three query strains h-btn1::NatMX ura-D18 leu1-32 ade6-M210, h-btn1102–208del::NatMX ura-D18 leu1-32 ade6-M210, h-btn1D363G::NatMX ura-D18 leu1-32 ade6-M210 and a control query strain, h-ade6::NatMX ura-D18 leu1-32 ade6-M21042 were mated with the library on Edinburgh Minimal Medium without nitrogen (Formedium). Following sporulation for 3 days at 25 °C and spore selection for 3 days at 42 °C, spores were pinned onto YES agar for 2 days at 32 °C to recover. Double mutant haploids were selected by growing cells on YES agar with ClonNat (Jena Bioscience; AB-102XL; 100 μg/ml) and G418 (Formedium; 500 μg/ml) for 2 days at 32 °C. Double mutant libraries were then grown on YES agar in quadruplicate (1536-well format) for 2 days at 32 °C.
Image analysis/processing
Images of colonies were acquired as shown (Fig. 1A). Plates were imaged using an EPSON Expression 1680 Pro transmission scanner. The size of each colony was determined using the gitter package in R43.
Mutants were excluded from the analysis if they failed to grow on the control ade6 plates. This was determined independently for each batch. Colony size was normalised to plate median, row and column median and for spatial effects as described previously44. Library mutants which were located with 500 kb of a query mutation were also excluded in order to remove spurious interactions as a result of linkage. Batch effects were removed using the limma package prior to Principal Component Analysis45. Differential fitness analysis was performed using limma45, instead using the plate normalised colony sizes and including batch as a term in the model. In order to account for differences in variance as a function of colony size, we model variance as a function of colony size in an approach akin to limma-voom46. Given that each library mutant was pinned in quadruplicate on the same plate and are hence not truly independent, each quadruplicate was treated as a technical replicate. The correlation between technical replicates was calculated, akin to calculating the correlation between duplicate spots in microarrays, and a mixed linear model was fitted to the data47. For each contrast of interest, p-values were corrected according to Benjamini–Hochberg48.
Cluster analysis was performed using Cluster3.0 for normalised colony sizes of each strain (http://bonsai.hgc.jp/~mdehoon/software/cluster/software.htm#ctv)49 and visualised with Java-Treeview (http://jtreeview.sourceforge.net)50.
Network analysis
Using Cytoscape version 3.4.0 (http://www.cytoscape.org/), gene lists of the most consistent hits were further analysed using add on macros for ClueGO version 2.3.3 (http://apps.cytoscape.org/apps/cluego) and CluePedia version 1.3.3 (http://apps.cytoscape.org/apps/cluepedia) for GO annotations, biological processes and KEGG pathways51,52. Kappa score of 0.4. Significant genetic interactions were compared to pombase’s slim Monarch Disease Ontology (Mondo) (https://www.pombase.org/browse-curation/disease-slim) for genes annotated to human disease.
Change history
05 July 2021
A Correction to this paper has been published: https://doi.org/10.1038/s41598-021-93446-8
References
Minnis, C. J., Thornton, C. D., Fitzpatrick, L. M. & Mckay, T. R. Cellular models of batten disease. Biochim. Biophys. Acta Mol. Basis Dis. 1866, 165559 (2019).
Mole, S. E. et al. Clinical challenges and future therapeutic approaches for neuronal ceroid lipofuscinosis. Lancet Neurol. 18, 107–116 (2019).
Mole, S. E. & Cotman, S. L. Genetics of the neuronal ceroid lipofuscinoses (Batten disease). Biochim. Biophys. Acta 1852, 2237–2241 (2015).
Mitchison, H. M. et al. Genomic structure and complete nucleotide sequence of the batten disease gene, CLN3. Genomics 40, 346–350 (1997).
Munroe, P. B. et al. Spectrum of mutations in the batten disease gene, CLN3. Am. J. Hum. Genet. 61, 310–316 (1997).
Haines, R. L., Codlin, S. & Mole, S. E. The fission yeast model for the lysosomal storage disorder batten disease predicts disease severity caused by mutations in CLN3. Dis. Model Mech. 2, 84–92 (2009).
Kitzmuller, C., Haines, R. L., Codlin, S., Cutler, D. F. & Mole, S. E. A function retained by the common mutant CLN3 protein is responsible for the late onset of juvenile neuronal ceroid lipofuscinosis. Hum. Mol. Genet. 17, 303–312 (2008).
Gachet, Y., Codlin, S., Hyams, J. S. & Mole, S. E. btn1, the Schizosaccharomycespombe homologue of the human batten disease gene CLN3, regulates vacuole homeostasis. J. Cell Sci. 118, 5525–5536 (2005).
NCL-Database (1998). NCL Resource—A Gateway For Batten Disease [Online]. UCL. https://www.ucl.ac.uk/ncl-disease/ [accessed].
Vesa, J. et al. Mutations in the palmitoyl protein thioesterase gene causing infantile neuronal ceroid lipofuscinosis. Nature 376, 584–587 (1995).
Appu, A. P. et al. Cln3-mutations underlying juvenile neuronal ceroid lipofuscinosis cause significantly reduced levels of palmitoyl-protein thioesterases-1 (PPT1)-protein and PPT1-enzyme activity in the lysosome. J. Inherit. Metab. Dis. 42, 944–954 (2019).
Linder, M. E. & Deschenes, R. J. Palmitoylation: Policing protein stability and traffic. Nat. Rev. Mol. Cell Biol. 8, 74–84 (2007).
Finley, D., Ozkaynak, E. & Varshavsky, A. The yeast polyubiquitin gene is essential for resistance to high temperatures, starvation, and other stresses. Cell 48, 1035–1046 (1987).
Gemayel, R. et al. Variable repeats in the eukaryotic polyubiquitin gene ubi4 modulate proteostasis and stress survival. Nat. Commun. 8, 397 (2017).
Repetto, B. & Tzagoloff, A. Structure and regulation of KGD1, the structural gene for yeast alpha-ketoglutarate dehydrogenase. Mol. Cell Biol. 9, 2695–2705 (1989).
Bond, M. E., Brown, R., Rallis, C., Bahler, J. & Mole, S. E. A central role for tor signalling in a yeast model for juvenile CLN3 disease. Microb. Cell 2, 466–480 (2015).
Codlin, S., Haines, R. L., Burden, J. J. & Mole, S. E. Btn1 affects cytokinesis and cell-wall deposition by independent mechanisms, one of which is linked to dysregulation of vacuole pH. J. Cell Sci. 121, 2860–2870 (2008).
Zhang, D. et al. Incomplete annotation has a disproportionate impact on our understanding of Mendelian and complex neurogenetic disorders. Sci. Adv. 6, 8299 (2020).
Clark, M. B. et al. Long-read sequencing reveals the complex splicing profile of the psychiatric risk gene CACNA1C in human brain. Mol. Psychiatry 25, 37–47 (2020).
Qian, W. & Zhang, J. Genomic evidence for adaptation by gene duplication. Genome Res. 24, 1356–1362 (2014).
Menne, T. F. et al. The Shwachman–Bodian–Diamond syndrome protein mediates translational activation of ribosomes in yeast. Nat. Genet. 39, 486–495 (2007).
Vitiello, S. P., Benedict, J. W., Padilla-Lopez, S. & Pearce, D. A. Interaction between Sdo1p and Btn1p in the Saccharomycescerevisiae model for Batten disease. Hum. Mol. Genet. 19, 931–942 (2010).
Cheng, L., Watt, R. & Piper, P. W. Polyubiquitin gene expression contributes to oxidative stress resistance in respiratory yeast (Saccharomycescerevisiae). Mol. Gen. Genet. 243, 358–362 (1994).
Simon, J. R., Treger, J. M. & Mcentee, K. Multiple independent regulatory pathways control UBI4 expression after heat shock in Saccharomycescerevisiae. Mol. Microbiol. 31, 823–832 (1999).
Bennett, V. Dynamic Plasma Membranes: Portals Between Cells and Physiology (Academic Press, 2016).
Fukata, Y. & Fukata, M. Protein palmitoylation in neuronal development and synaptic plasticity. Nat. Rev. Neurosci. 11, 161–175 (2010).
Greaves, J. & Chamberlain, L. H. DHHC palmitoyl transferases: substrate interactions and (patho)physiology. Trends Biochem. Sci. 36, 245–253 (2011).
Shah, B. S., Shimell, J. J. & Bamji, S. X. Regulation of dendrite morphology and excitatory synapse formation by zDHHC15. J. Cell Sci. 132, jcs230052 (2019).
Brichta, L. et al. Identification of neurodegenerative factors using translatome-regulatory network analysis. Nat. Neurosci. 18, 1325–1333 (2015).
Dawson, G., Kilkus, J., Siakotos, A. N. & Singh, I. Mitochondrial abnormalities in CLN2 and CLN3 forms of batten disease. Mol. Chem. Neuropathol. 29, 227–235 (1996).
Fossale, E. et al. Membrane trafficking and mitochondrial abnormalities precede subunit C deposition in a cerebellar cell model of juvenile neuronal ceroid lipofuscinosis. BMC Neurosci. 5, 57 (2004).
Chen, H. et al. Reductions in the mitochondrial enzyme alpha-ketoglutarate dehydrogenase complex in neurodegenerative disease—Beneficial or detrimental?. J. Neurochem. 139, 823–838 (2016).
Huergo, L. F. & Dixon, R. The emergence of 2-oxoglutarate as a master regulator metabolite. Microbiol. Mol. Biol. Rev. 79, 419–435 (2015).
Elshatory, Y. et al. Early changes in gene expression in two models of batten disease. FEBS Lett. 538, 207–212 (2003).
Luiro, K. et al. Batten disease (JNCL) is linked to disturbances in mitochondrial, cytoskeletal, and synaptic compartments. J. Neurosci. Res. 84, 1124–1138 (2006).
Lebrun, A.-H. et al. Analysis of potential biomarkers and modifier genes affecting the clinical course of CLN3 disease. Mol. Med. (Cambridge, Mass) 17, 1253–1261 (2011).
Scifo, E. et al. Drafting the CLN3 protein interactome in SH-SY5Y human neuroblastoma cells: A label-free quantitative proteomics approach. J. Proteome Res. 12, 2101–2115 (2013).
Murray, J. M., Watson, A. T. & Carr, A. M. Transformation of Schizosaccharomycespombe: Lithium acetate/dimethyl sulfoxide procedure. Cold Spring Harb. Protoc. https://doi.org/10.1101/pdb.prot090951 (2016).
Livak, K. J. & Schmittgen, T. D. Analysis of relative gene expression data using real-time quantitative pcr and the 2(−Delta Delta C(T)) method. Methods 25, 402–408 (2001).
Kim, D. U. et al. Analysis of a genome-wide set of gene deletions in the fission yeast Schizosaccharomycespombe. Nat. Biotechnol. 28, 617–623 (2010).
Baryshnikova, A. et al. Synthetic genetic array (SGA) analysis in Saccharomycescerevisiae and Schizosaccharomycespombe. Methods Enzymol. 470, 145–179 (2010).
Rallis, C., Lopez-Maury, L., Georgescu, T., Pancaldi, V. & Bahler, J. Systematic screen for mutants resistant to TORC1 inhibition in fission yeast reveals genes involved in cellular ageing and growth. Biol. Open 3, 161–171 (2014).
Wagih, O. & Parts, L. Gitter: A robust and accurate method for quantification of colony sizes from plate images. G3 4, 547–552 (2014).
Wagih, O. et al. SGAtools: One-stop analysis and visualization of array-based genetic interaction screens. Nucleic Acids Res. 41, W591–W596 (2013).
Ritchie, M. E. et al. Limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 43, E47 (2015).
Law, C. W., Chen, Y., Shi, W. & Smyth, G. K. Voom: Precision weights unlock linear model analysis tools for RNA-Seq read counts. Genome Biol. 15, R29 (2014).
Smyth, G. K., Michaud, J. & Scott, H. S. Use of within-array replicate spots for assessing differential expression in microarray experiments. Bioinformatics 21, 2067–2075 (2005).
Benjamini, Y., Drai, D., Elmer, G., Kafkafi, N. & Golani, I. Controlling the false discovery rate in behavior genetics research. Behav. Brain Res. 125, 279–284 (2001).
De Hoon, M. J., Imoto, S., Nolan, J. & Miyano, S. Open source clustering software. Bioinformatics 20, 1453–1454 (2004).
Saldanha, A. J. Java Treeview-extensible visualization of microarray data. Bioinformatics 20, 3246–3248 (2004).
Bindea, G. et al. Cluego: A cytoscape plug-in to decipher functionally grouped gene ontology and pathway annotation networks. Bioinformatics 25, 1091–1093 (2009).
Kanehisa, M. & Goto, S. KEGG: Kyoto Encyclopedia of Genes and Genomes. Nucleic Acids Res. 28, 27–30 (2000).
Acknowledgements
This project was supported by the Biotechnology and Biological Sciences Research Council London Interdisciplinary Biosciences Consortium Doctoral Training Partnership (BBRSC grant No: 1632387, CM), USA Children’s Brain Diseases Foundation (SM), the European Union’s Horizon 2020 research and innovation programme (BATCure, grant No 666918, SM, CR), and a Wellcome Trust Senior Investigator Award (095598/Z/11/Z) to JB. SM benefits from MRC funding to the MRC Laboratory for Molecular Cell Biology University Unit at UCL (award code MC_U12266B) towards lab and office space. ST was supported by a Boehringer Ingelheim Fonds PhD Fellowship.
Author information
Authors and Affiliations
Contributions
C.M., J.T., S.M. wrote the main manuscript, C.M., J.T., J.P. and E.T. conducted experimental data procurement and analysis, C.M. and J.T. prepared the main figures. C.M., J.T., J.P., C.R., J.B. and S.M. reviewed the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The original online version of this Article was revised: The original version of this Article contained errors in Figure 1c and accompanying legend, where data points in the bar graph plot appeared incorrectly in green colour. Furthermore, Table 2 contained errors, where genes for the strains btn1∆, btn1D363G and btn1102–208del in the column ‘Top 5 genes’ appeared incorrectly in bold font.
Supplementary Information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Minnis, C.J., Townsend, S., Petschnigg, J. et al. Global network analysis in Schizosaccharomyces pombe reveals three distinct consequences of the common 1-kb deletion causing juvenile CLN3 disease. Sci Rep 11, 6332 (2021). https://doi.org/10.1038/s41598-021-85471-4
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-021-85471-4
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
