ULK1) in Silico Analysis for Biomarker Identiﬁcation: a Vital Component of Autophagy

Gene 562 (2015) 40–49

Contents lists available at ScienceDirect

Gene

journal homepage: www.elsevier.com/locate/gene

Unc-51 like kinase 1 (ULK1) in silico analysis for biomarker identiﬁcation: A vital component of autophagy

Rohit Randhawa a, Manika Sehgal a, Tiratha Raj Singh a, Ajay Duseja b, Harish Changotra a,⁎ a Department of Biotechnology and Bioinformatics, Jaypee University of Information Technology, Waknaghat, Solan 1732 34 Himachal Pradesh, India b Department of Hepatology, Postgraduate Institute of Medical Education and Research, Chandigarh 160 012, India article info abstract

Article history: Autophagy is a degradation pathway involving lysosomal machinery for degradation of damaged organelles like Received 19 October 2014 the endoplasmic reticulum and mitochondria into their building blocks to maintain homeostasis within the cell. Received in revised form 3 February 2015 ULK1, a serine/threonine kinase, is conserved across species, from yeasts to mammals, and plays a central role in Accepted 5 February 2015 autophagy pathway. It receives signals from upstream modulators such as TIP60, mTOR and AMPK and relays Available online 19 February 2015 them to its downstream substrates like Ambra1 and ZIP kinase. The activity of this complex is regulated through – fi fi Keywords: protein protein interactions and post-translational modi cations. Applying in silico analysis we identi ed Autophagy (i) conserved patterns of ULK1 that showed its evolutionary relationship between the species which were closely ULK1 related in a family compared to others. (ii) A total of 23 TFBS distributed throughout ULK1 and nuclear factor Palmitoylation (erythroid-derived) 2 (NFE2) is of utmost significance because of its high importance rate. NEF2 has already Phosphorylation been shown experimentally to play a role in the autophagy pathway. Most of these were of zinc coordinating Haplotype class and we suggest that this information could be utilized to modulate this pathway by modifying interactions – Protein protein interactions of these TFs with ULK1. (iii) CATTT haplotype was prominently found with frequency 0.774 in the studied population and nsSNPs which could have harmful effect on ULK1 protein and these could further be tested. (iv) A total of 83 phosphorylation sites were identified; 26 are already known and 57 are new that include one at tyrosine residue which could further be studied for its involvement in ULK1 regulation and hence autophagy. Furthermore, 4 palmitoylation sites at positions 426, 927, 1003 and 1049 were also found which could further be studied for protein–protein interactions as well as in trafficking. © 2015 Elsevier B.V. All rights reserved.

1. Introduction required for binding to other essential components of autophagy, Atg13 and FIP200, is lacking in ULK3, ULK4 and STK36. Therefore, Autophagy is an evolutionarily conserved degradation pathway in ULK1 and ULK2 are the primary candidate mammalian Atg1 which cytoplasmic portions that include damaged organelles orthologues, essential for induction of autophagy. ULK1 protein expres- and misfolded or aggregated proteins are sequestered in double- sion pattern studies have been done (Kundu et al., 2008). ULK1 knock membrane vesicles called autophagosomes. Then, these contents are out mouse model was viable and did not show any evident develop- delivered to the lysosomes for degradation resulting in removal/ mental defects, in contrast to other core autophagy genes (Atg3, Atg5, recycling of damaged/harmful contents from the cell to maintain the Atg7, Atg9 and Atg16L1) where their deletions led to neonatal lethality. cellular homeostasis. This pathway is dysregulated in many diseases in- ULK1 expression levels were elevated during erythroid maturation but cluding neurodegenerative, inflammatory, muscle, cardiac, infectious, not of ULK2 suggesting that ULK2 was not involved in this process. and neoplastic diseases. There is possibility that modulation of autoph- Moreover, they also showed an important role of ULK1 in selective agy pathway could be helpful in better therapeutic management of clearance of mitochondria and ribosomes in reticulocytes. The reasons these diseases. that ULK1 is not essential for murine survival could be (i) ULK2, which In mammals, autophagy plays an important role in preimplantation shows N50% homology with ULK1 and shows functional redundancy development, survival during neonatal starvation, cell differentiation, and induce autophagy and/or (ii) existence of ULK1 independent erythropoiesis and lymphopoiesis. Autophagy is actively induced in all mechanism of autophagy. Furthermore, Chan and co-workers have neonatal tissues early during development. The five identified Atg1 ho- shown that in HEK293 cells ULK1 was critical for inducing autophagy mologues in mammals include uncoordinated (Unc) 51-like kinase in response to amino acid starvation (Chan et al., 2009). Therefore, the (ULK1) 1 to 4 and STK36. Carboxy-terminal domain (CTD) which is focus of this study was to analyze ULK1 which is a major regulator of autophagy. – ⁎ Corresponding author. ULK1, a serine threonine kinase, is one of the central human E-mail addresses: [email protected], [email protected] (H. Changotra). autophagy-related genes and its chromosomal location is 12q24.3. A

ULK1 gene is 28,517 bp long with 28 exons and is translated to 1050 of site-specific residues and phylogenetic analysis; (2) Regulatory ele- amino acids. ULK1 forms a stable complex with Atg13, FIP200, and ments and over-represented transcription factor binding site (TFBS) Atg101. This complex plays a crucial role in initiation step of autophagy. recognition; (3) Detection of nsSNPs, their phenotypic effects and ULK1 regulates its substrates and is itself regulated by phosphorylation quantitative statistical analysis for genetic parameters; (4) Elucidation events. mTOR1, AMPK and TIP60 are its well known upstream regula- of putative phosphorylation and palmitoylation sites; and (5) Protein– tors. It is hyperphosphorylated in nutrient-rich conditions and dephos- Protein Interaction (PPI) studies. phorylates on starvation. So far, around 30 phosphorylation sites have been identified on ULK1 and most of the kinases responsible for 2.1. Identification of site-specific residues and phylogenetic analysis its phosphorylation and functions of these are still unidentified for ULK1 (Mack et al., 2012). This supports that phosphorylation events play an important role in ULK1 regulation. Recently, decreased expression of The analyses initiated with retrieval of human protein sequence for ULK1 has been shown in breast cancer patients, which was associated ULK1 (GenBank Accession Number: AAC32326) from the National Cen- with cancer progression and low autophagic activity (Tang et al., ter for Biotechnology Information (NCBI) and corresponding protein se- 2012). However, another study showed higher expression of ULK1 in quences for other seven species of families Hominidae (Pan troglodytes; the hepatocellular carcinoma (HCC) patients and furthermore, higher GenBank Accession Number: JAA43195), Bovidae (Bos taurus;GenBank ULK1 expression in HCC patients was associated with low survival rate Accession Number: NP_001192856), Cricetidae (Cricetulus griseus; (Xu et al., 2013). These studies indicate different roles of autophagy in GenBank Accession Number: EGW02429), Pteropodidae (Pteropus different types of cancers and indeed in different diseases. Therefore, alecto; GenBank Accession Number: ELK14239), Muridae (Rattus ULK1 could be used as a prognostic marker for cancer patients. More- norvegicus; GenBank Accession Number: NP_001101811, Mus musculus; over, this gene has been shown to be involved in genetic susceptibility GenBank Accession Number: NP_033495), Pipidae (Xenopus (Silurana) of Crohn's disease (CD). Recent studies have shown the association of tropicalis; GenBank Accession Number: NP_001106388) were also re- three SNPs (rs12303764, rs10902469 and rs7488085) with CD trieved and further deliberated for their evolutionary conservation. (Henckaerts et al., 2011). These variations could be used as prognostic markers in the therapeutic interventions after validation in more num- 2.1.1. Evolutionary conserved and variable regions ber of patients and other populations. The genetic variations leading to different phenotypes were In the present study, we computationally analyzed the ULK1 gene analyzed by observing the variable regions in the multiple sequence for its phylogeny reconstruction which suggests that it is closely related alignment (MSA) generated for the ULK1 gene. The latter was carried fi in a family. We identi ed new TFBS, snSNP with their possible out using multiple sequence comparison by log-expectation (MUSCLE) phenotypic effect on ULK1 protein function, phosphorylation and (Edgar, 2004) and multiple alignment using fast Fourier transform – palmitoylation sites and protein protein interactions. This comprehen- (MAFFT) (Katoh et al., 2002). These programs use log-expectation sive in silico analyses would be helpful to unravel the functions of this scores and fast Fourier transform methods respectively for providing gene and understand autophagy as well as non-autophagy roles of better average accuracy and speed compared to other MSA algorithms. this gene. Programs were used with their default parameters.

2. Material and methods 2.1.2. Evolutionary relationship associated with ULK1 Highly conserved regions play an imperative role in phylogenetic In the study, an extensive examination of the ULK1 gene is carried tree reconstruction. Therefore, the evolutionary relationship among out which is subdivided into seven major sections; (1) Identiﬁcation eight species was elucidated on the basis of sequence similarities by

Fig. 1. Partial representation of Multiple Sequence Alignment of ULK1 gene for 8 different species. This was carried out using multiple sequence comparison by log-expectation (MUSCLE) and multiple alignment using fast Fourier transform (MAFFT) which use log-expectation scores and fast Fourier transform methods, respectively. Programs were used with their default parameters. Human sequence was taken as a reference and is shown at the top of MSA. Areas in boxes represent various conserved regions in MSA. 42 R. Randhawa et al. / Gene 562 (2015) 40–49

Fig. 2. Evolutionary relationship among eight species included in the study. This was done by applying molecular evolutionary genetics analysis 5 (MEGA5). The phylogenetic tree was reconstructed by using maximum parsimony (MP), a character-based method for deducing phylogenetic trees by minimizing the total number of evolutionary steps. The analysis was performed for the veriﬁcation of inferred tree by taking 1000 bootstrap replicates to generate statistically signiﬁcant phylogenetic tree.

applying molecular evolutionary genetics analysis 5 (MEGA5) which 2.2. Identification of regulatory elements and over-represented TFBS helps to estimate the rates of molecular evolution and deduce ancestral affiliations (Tamura et al., 2011). The phylogenetic reconstruction was Identification of regulatory elements like enhancers, silencers and achieved by means of maximum parsimony (MP), a character-based repressors involved in controlling the expression of ULK1 provides use- method for deducing phylogenetic trees by minimizing the total num- ful insights into how the gene is regulated and expressed under the in- ber of evolutionary steps required for explanation of a given set of fluence of these factors. Distant regulatory elements of co-regulated data. The bootstrap analysis was also performed for the verification of genes (DiRE) (Gotea and Ovcharenko, 2008) and oPOSSUM 3 (Kwon inferred tree by taking 1000 bootstrap replicates to generate statistically et al., 2012) were used for detection of the regulatory elements in significant phylogenetic tree. ULK1 and over-represented transcription binding sites, respectively.

Fig. 3. Transcription factor binding sites in ULK1 gene. (a) Distant regulatory elements of co-regulated genes (DiRE) showed that 75% of the total transcription factors were present in UTR and remaining 25% in the intron region. (b) Sequence logos obtained from JASPAR. R. Randhawa et al. / Gene 562 (2015) 40–49 43

Additionally, JASPAR database (Sandelin et al., 2004) was also explored Table 2 for similar kind of datasets to identify various classes, families and List of top ten transcription factors with their rates of occurrence and importance. sequence logos for transcription factors (TFs) and their binding sites. # Transcription factor Occurrence Importance

1 NFE2 25.00% 0.48555 2 CACCCBINDINGFACTOR 25.00% 0.24984 2.3. Identification of nsSNPs, their phenotypic effects and quantitative 3 CHOP 25.00% 0.24961 statistical analyses for genetic parameters 4 MEF3 25.00% 0.24805 5 GLI 25.00% 0.24727 The nsSNPs are the nucleotide changes that result in the altered 6 BARBIE 25.00% 0.24727 7 ARNT 25.00% 0.23340 amino acid in the protein sequence. This altered amino acid may or 8 WT1 25.00% 0.21587 may not affect the function of the protein. The affected protein function 9 BACH2 25.00% 0.20391 in case of modifying nsSNPs is due to change in its (1) structure, (2) sta- 10 STAT 25.00% 0.19805 bility and (3) by influencing functional binding sites. We have used Sorting Intolerant From Tolerant (SIFT) and Polymorphism Phenotyping (PolyPhen) tools to identify nsSNPs which are popular standard tools to analyzed for genetic association were D′ and r2.TheD′ value provides predict intolerant or damaging variants. These are based on sequence the measure of LD between the two blocks and its value closer to zero homology, conservation, structure and SWISS-PROT annotation. For shows a higher amount of historical recombination between the two identification of SNPs, we computationally analyzed ULK1 gene as ex- blocks and r2 gives the correlation coefficient between the two loci perimental methods are complicated, expensive and time consuming. under study. These SNPs were further analyzed for their phenotypic effect in coding sequences. SIFT algorithm was used for identification of genetic variations, leading to diverse phenotypes in ULK1 (Ng and Henikoff, 2003). 2.4. Elucidation of putative phosphorylation and palmitoylation sites The prediction is based on the generated SIFT score and focuses on the in ULK1 phenomenon of protein conservation which states that protein evolution has a strong correlation with protein function. PolyPhen was also Most of the processes occurring in a cell are controlled by signaling used for validating these phenotypic consequences which predicts on cascade dependent on phosphorylation/dephosphorylation. ULK1 is a the basis of sequence, structural, evolutionary annotations and substitu- kinase which itself is controlled by phosphorylation events and phos- tions in the proteins (Ramensky et al., 2002). The profile scores for the phorylates its substrates for modulating their activity (Bononi et al., two amino acid positions (native and mutant) were calculated and 2011; Wu et al., 2014). Hence, detection of phosphorylation sites in assessed for evaluating their phenotypic effects. On the basis of the ULK1 may unravel important functional aspects regarding its involve- scores, PolyPhen categorizes the substitutions into three classes i.e. ‘be- ment in various disorders (Olsen et al., 2010; Wang et al., 2010; Shang nign’, ‘possibly damaging’ and ‘probably damaging’. et al., 2011). The NetPhos algorithm (Blom et al., 1999) was used for The genotype data from CEU (CEPH—Utah Residents with Northern the prediction of phosphorylation sites at serine (S), threonine (T) and and Western European Ancestry) population for the ULK1 gene was re- tyrosine (Y) residues in the ULK1 amino acid sequence. This algorithm trieved from The International HapMap project (Thorisson et al., 2005) utilizes an artificial neural network (ANN) based method which is and was analyzed for various quantitative genetic parameters: linkage trained from PhosphoBase (Kreegipuu et al., 1999), a database of exper- disequilibrium (LD), haplotypes and SNPs. These parameters represent imentally validated phosphorylated proteins. the combination of alleles on neighboring loci on the chromosome Detection of palmitoylation sites is also an important component of being transmitted together and involvement of alleles in a non- this study. The palmitoylation sites were obtained from CSS-PALM random mode of inheritance in the population. This analysis was per- (Zhou et al., 2006), a tool based on the clustering and scoring strategy formed using Haploview (Barrett et al., 2005) and the parameters (CSS) algorithm for the prediction of palmitoylation sites.

Table 1 Regulatory elements, their types and identiﬁed transcription factors.

# Regulatory element Type Score Locus Gene Candidate TFBS (relative positions)

1 chr12:130,946,862–130,947,195 Intron 2.544 chr12:130,905,754–130,979,065 ULK1 CACCCBINDINGFACTOR(120) NFE2(129) BACH2(130) GLI(207) WT1(214) RFX1(307) PR(324) GRE(324) 2 chr12:130,945,120–130,945,301 UTR5 1.221 chr12:130,905,754–130,979,065 ULK1 PAX5(4) NRF1(6) HIC1(9) MTF1(37) ZBRK1(45) 3 chr12:130,972,084–130,972,437 UTR3 2.027 chr12:130,905,754–130,979,065 ULK1 SMAD4(32) MYOGNF1(122) STAT(169) HSF1(205) BARBIE(233) 4 chr12:130,973,189–130,973,466 UTR3 2.784 chr12:130,905,754–130,979,065 ULK1 ARNT(65) CHOP(82) ERR1(151) MEF3(217) PPARG(258) 44 R. Randhawa et al. / Gene 562 (2015) 40–49

Table 3 Over-represented transcription factor binding sites and their annotation.

TF JASPAR ID Class Family Target gene hits Target TFBS hits Z-score Fisher score

Zfx MA0146.1 Zinc-coordinating BetaBetaAlpha-zinc finger 1 5 16.721 0.919 Tcfcp2l1 MA0145.1 Other CP2 1 4 12.964 0.849 Klf4 MA0039.2 Zinc-coordinating BetaBetaAlpha-zinc finger 1 8 11.997 0.596 MIZF MA0131.1 Zinc-coordinating BetaBetaAlpha-zinc finger 1 1 11.047 2.476 Egr1 MA0162.1 Zinc-coordinating BetaBetaAlpha-zinc finger 1 2 9.729 1.358 SP1 MA0079.2 Zinc-coordinating BetaBetaAlpha-zinc finger 1 6 8.687 0.692 RORA_2 MA0072.1 Zinc-coordinating Hormone-nuclear receptor 1 1 8.63 1.936 NR4A2 MA0160.1 Zinc-coordinating Hormone-nuclear receptor 1 5 7.765 0.577 RORA_1 MA0071.1 Zinc-coordinating Hormone-nuclear receptor 1 2 6.65 1.054 NFYA MA0060.1 Other Alpha-Helix NFY CCAAT-binding 1 1 5.88 1.44 Zfp423 MA0116.1 Zinc-coordinating BetaBetaAlpha-zinc finger 1 1 5.353 1.47 ZEB1 MA0103.1 Zinc-coordinating BetaBetaAlpha-zinc finger 1 8 4.153 0.349 Foxa2 MA0047.2 Winged helix–turn–helix Forkhead 1 2 2.758 0.753 MZF1_5-13 MA0057.1 Zinc-coordinating BetaBetaAlpha-zinc finger 1 3 2.147 0.612 E2F1 MA0024.1 Winged helix–turn–helix E2F 1 1 2.033 1.115 GABPA MA0062.2 Winged helix–turn–helix Ets 1 1 1.883 0.973 AP1 MA0099.2 Zipper-type Leucine zipper 1 4 1.584 0.412 Stat3 MA0144.1 Ig-fold Stat 1 1 1.386 0.918 Esrrb MA0141.1 Zinc-coordinating Hormone-nuclear receptor 1 1 1.181 0.89 Myf MA0055.1 Zipper-type Helix–loop–helix 1 1 1.031 0.891

2.5. Protein–Protein Interaction studies for ULK1 the ULK1 gene of eight species. The maximum conserved blocks are present in the kinase domain followed by S/T rich region and CTD. The identification of complex interaction networks involving ULK1 Dots represent conservation with respect to human ULK1 sequence, protein which is often also implicated in different or same pathways while variable characters are shown as amino acids. As discussed has been constructed by Search Tool for Retrieval of Interacting Genes above, these sequence similarities (conserved regions) reflect the evo- and proteins (STRING) version 9.05 (Szklarczyk et al., 2011). The tool lutionary relationship between the species. predicts the interactions between various proteins based on a confi- For the phylogenetic analysis of the ULK1 protein sequence of the dence score and validates the connections using databases, text mining eight different species (selected based on availability of complete vali- and gene fusion support. The interaction studies were performed in var- dated protein sequence data), MEGA5 was used with 1000 bootstrap ious modes and by changing parameters to obtain a robust network replicates using MP method and an evolutionary tree was reconstructed model for ULK1 and its associated interacting partners. The different as shown in Fig. 2. The generated phylogenetic tree helps in clear under- modes include confidence view, evidence view, action mode and the in- standing of evolutionary relationship between the different species. teractive view to infer the most appropriate interactions among nodes Here, as shown by the bootstrap values, Homo sapiens and in the network. P. troglodytes have 100% evolutionary relationship depicting that the ULK1 gene present in these species is quite identical and could have 3. Results and discussions originated from the same ancestors. Similarly, R. norvegicus and M. musculus have 93% similarity and could have the same ancestral ori- 3.1. Phylogenetic analysis shows ULK1 gene is evolutionary conserved and gin as C. griseus, showing bootstrap value of 100. X. (Silurana) tropicalis, is more closely related in a family a species of amphibian family shows a vast difference in its origin and evolution from others, when studied on the basis of protein sequence The conserved regions in protein sequences generally signify the in- of ULK1. From these data, we could conclude that these sequences tegrity and stability of genome which further affect the basic cellular tend to be related closely in a family compared to others. processes. The conserved positions are considered to be involved in important functions, active sites of enzymes and binding sites of the pro- 3.2. Various transcription factor binding sites are distributed throughout tein receptors (del Sol et al., 2006). The MSA generated from MUSCLE the ULK1 gene and MAFFT tools for the ULK1 protein was found to be quite similar and various important conserved patterns were identified. These The regulatory elements in the ULK1 gene were analyzed using DiRE patterns may have important association with diseased states and evo- software with the default value for random set of genes as 5000. This in lutionary relationship among these organisms. Fig. 1a illustrates the silico approach demonstrated a total of four potential regulatory ele- MSA for the eight sequences where highly conserved regions are ments in the ULK1 gene out of which, three are untranslated regions highlighted inside the red blocks whereas Fig. 1b clearly demonstrates (UTR) that correspond to 75% of the total regulatory region in ULK1 the mutations or variations in residues among the regions found in and one intron representing 25% of the total regulatory region as

Table 4 The class-wise categorization of transcription factors detected in ULK1.

Class name TFS Total gene hits TFBS hits

Zinc-coordinating Zfx, Klf4, MIZF, Egr1, SP1, RORA_1, RORA_2, 12 43 NR4A2, Zfp423, ZEB1, Esrrb, MZF1_5-13 Zipper-type AP1, Myf 2 5 Winged helix–turn–helix Foxa2, E2F1, GABPA 3 4 Ig-fold Stat3 1 1 Other alpha-helix NFYA 1 1 Other Tcfcp2l1 1 4 R. Randhawa et al. / Gene 562 (2015) 40–49 45

Fig. 4. Linkage disequilibrium plot, SNPs and haplotypes in ULK1. (a) The LD plot identified 14 kb block in ULK1. SNPs in the block are highlighted. The five important SNPs identified were rs9652059, rs1134574, rs7953348, rs11615995 and rs11616018 with minor allele frequencies of 0.2, 0.085, 0.198, 0.228 and 0.207 respectively. (b) CATTT haplotype was prominently found with frequency of 0.774 in the studied population. shown in Fig. 3a. Among the three UTR regions, two are 3′ UTRs and one of these zinc-coordinating residues gives rise to diverse classes of ZnFs. is 5′ UTR. Presence of regulatory elements at 5′ and 3′ indicates that The interactions of these classes of TFs with the ULK1 gene could be ULK1 is a complex gene and it could be regulated from both ends utilized to modulate its expression which in turn could alter autophagy (Heinrich and Pagtakhan, 2004). Additionally, it is proposed that the pathway. The latter recently has been shown to involve in various neighboring generic locations might play a critical role in ULK1 regula- diseases and its manipulation could be utilized for therapeutic interven- tion and ultimately in autophagy. A total of 23 TFs in UTR and intron re- tions (Beerli et al., 2000; Pabo et al., 2001; Segal et al., 2004). gions were identified using DiRE and are shown along with their locus, positions and score in Table 1. These TFs either bind directly or in the 3.3. Linkage disequilibrium analysis shows that 5 SNPs are linked and in form of complex to the transcriptional regulatory region of ULK1 ULK1, 4 nsSNPs exist that have damaging/harmful effect which could further control its expression. Among these detected transcription factors, CHOP, E2F1, NFE2 and STAT have already been studied In order to analyze various genetic parameters (including linkage for their role in ULK1 expression where CHOP and E2F1 enhance the disequilibrium (LD), haplotypes and SNPs) for the ULK1 gene that autophagy process whereas NFE2 and STAT suppress the process of au- could give an idea about its involvement in predisposition of various tophagy (Deretic et al., 2013; Fullgrabe et al., 2014). All these TFs have diseases, the genotype data for CEU (CEPH—Utah Residents with North- equal occurrence rate of 25% (Table 2). Based on high importance rate ern and Western European Ancestry) obtained from The International (0.48555), nuclear factor (erythroid-derived) 2 (NFE2) is of utmost sig- HapMap Project for ULK1 gene was subjected to extensive statistical nificance when compared to the rest of the TFs (importance rate of analysis. These parameters act as vital biomarkers for the functional b0.24984) found in the analysis (Table 2). Distribution of TFBS all over association with a variety of diseases. The LD analysis revealed an the ULK1 gene suggests multiple points of their action and this informa- important block in the ULK1 gene with five important SNPs having tion could further be utilized to understand its regulation (Whitfield non-random association as represented in Fig. 4a. The five important et al., 2012). These multiple hotspots could be utilized to control the SNPs identified were rs9652059, rs1134574, rs7953348, rs11615995 regulation of ULK1 in an efficient way. Furthermore, to substantiate these results, we used oPOSSUM tool to find over-represented TFBS in the promoter region of ULK1. The identified TFs are shown in Tables 3 Table 6 Information on two loci under consideration with their statistical inference. &4with the sequence logos of TFs represented in Fig. 3b. The recognition of TFBS depends on features that consider the search parameters L1 L2 D′ LOD r2 CIlow CIhi of JASPAR N8bitsandN75% as the threshold of position specificscoring rs9652059 rs1134574 1 7.49 0.363 0.71 1 matrices. Based on standard parameters, we identified 20 TFs binding to rs9652059 rs7953348 1 24.79 1 0.92 1 58 sites (Table 3). On further grouping based on their class, we observed rs9652059 rs11615995 1 21.21 0.903 0.89 1 that major over-represented TFBS (43/58) in ULK1 were of zinc- rs9652059 rs11616018 1 22.98 0.948 0.89 1 – – rs9652059 rs12303764 1 5.2 0.151 0.69 1 coordinating class (Table 4), followed by winged helix turn helix, rs9652059 rs3088051 0.835 1.99 0.06 0.35 0.95 zipper type, alpha helix and Ig-fold. Zinc coordinating class of TFs con- rs1134574 rs7953348 1 6.99 0.839 0.69 1 tains zinc fingers (ZnFs), a widespread protein domain, and the spacing rs1134574 rs11615995 1 6.81 0.324 0.7 1 rs1134574 rs11616018 1 6.78 0.321 0.69 1 rs1134574 rs12303764 1 1.71 0.054 0.31 1 Table 5 rs1134574 rs3088051 0.57 0.28 0.01 0.05 0.89 The linkage disequilibrium table representing important SNPs. rs7953348 rs11615995 1 20.57 0.899 0.88 1 rs7953348 rs11616018 1 22.98 0.948 0.89 1 S. no Name ObsHET PredHET HWpval %Geno MAF Alleles rs7953348 rs12303764 1 5.2 0.151 0.69 1 1 rs9652059 0.267 0.32 0.331 100 0.2 C:T rs7953348 rs3088051 0.833 1.98 0.062 0.35 0.95 2 rs1134574 0.169 0.155 1 98.9 0.085 A:G rs11615995 rs11616018 1 22.33 0.949 0.9 1 3 rs7953348 0.259 0.318 0.2774 96.7 0.198 T:C rs11615995 rs12303764 1 5.48 0.189 0.7 1 4 rs11615995 0.316 0.352 0.6146 94.4 0.228 T:C rs11615995 rs3088051 0.825 1.87 0.065 0.33 0.95 5 rs11616018 0.276 0.328 0.3664 96.7 0.207 T:C rs11616018 rs12303764 1 5.48 0.159 0.7 1 6 rs12303764 0.552 0.471 0.3335 96.7 0.379 T:G rs11616018 rs3088051 0.84 2.13 0.067 0.37 0.96 7 rs3088051 0.424 0.387 0.7676 97.8 0.263 A:G rs12303764 rs3088051 0.616 2.02 0.086 0.27 0.82 46 R. Randhawa et al. / Gene 562 (2015) 40–49

Table 7 3.4. ULK1 comprises novel phosphorylation and palmitoylation sites The identiﬁed coding non-synonymous SNPs having damaging effects.

SIFT prediction PolyPhen prediction Phosphorylation and palmitoylation sites play a crucial role in protein–protein interactions, hence in the functions of a protein (Smotrys SNP ID Amino Tolerance Predicted Probability Predicted acid index impact score impact and Linder, 2004; Watanabe and Osada, 2012). As mentioned earlier, change ULK1 activity is controlled by phosphorylation events and in addition

rs79965940 N148T 0.01 Damaging – Benign it phosphorylates its substrates for modulating their activity (Wang rs61942435 A991V 0 Damaging – Benign et al., 2010; Wu et al., 2014). Most of the processes occurring in a cell rs55824543 T503M 0.01 Damaging 0.224 Benign are controlled by signaling cascade dependent on phosphorylation/de- rs56364352 S298L 0.05 Tolerated 1 Probably phosphorylation. Hence, detection of the phosphorylation sites in damaging ULK1 could be helpful to understand its various functional aspects as well as its involvement in various diseases like bone cancer, cervical ad- and rs11616018 with minor allele frequencies of 0.2, 0.085, 0.198, 0.228 enocarcinoma, gastric cancer and lung cancer (Olsen et al., 2010). For and 0.207 respectively as illustrated in Table 5. These SNPs also had prediction of phosphorylation sites at S, T, and Y amino acids in ULK1, r2 ≥ 0.8 which further validated the higher correlation between the the NetPhos algorithm representing the prediction score ≥ 0.5 was con- loci (Table 6). Furthermore, we identified one haplotype block in the sidered as phosphorylated. The ordered information retrieved from ULK1 gene where 5 SNPs and 4 haplotypes were recognized with differ- NetPhos includes protein ID, phosphorylated AA in the sequence, a ent population frequencies (Fig. 4b). CATTT haplotype was prominently stretch of 9 AAs with the phosphorylated residue at the center and the found with frequency 0.774 in the studied population. The coding score. A complete list of all the phosphorylation sites identified in nsSNPs may have significant impact on function of a protein and such ULK1 is shown in Supplementary Table 1. Some of the predicted phos- SNPs could be damaging or deleterious (Marín-Martín et al., 2014). phorylation sites have already been verified experimentally and the cor- We used SIFT and PolyPhen to analyze the deleterious impact of these responding PubMed IDs are represented in Table 8. Among these nsSNPs on the ULK1 protein. The predicted impact of change in amino phosphorylation sites, the newly identified phosphorylation sites are acid may be tolerated or damaging depending on the tolerance index shown in Fig. 5 with their position in the domains of the ULK1 protein. and probability scores. We identified 4 nsSNPs i.e. rs79965940, We found a new phosphorylation site in ULK1 at amino acid position rs61942435, rs55824543 and rs56364352 which could have harmful 295 of tyrosine (Y) residue in kinase domain with significant score functional effects on the ULK1 protein (Table 7). It is proposed that (0.809) which has not been verified experimentally and could further these SNPs and their damaging or deleterious impact could be exam- be studied to unravel its impact on the protein function. This tyrosine ined in other populations for their possible involvement in various dis- phosphorylation site is present in the kinase domain of the gene and eases associated with ULK1 and autophagy. thus may regulate the phosphotransferase activity of the protein.

Table 8 Experimentally veriﬁed phosphorylation sites in ULK1.

Name Position Context sequence Score Prediction Predicted kinase PubMed IDs

ULK1_HUMAN 225 FQASSPQDL 0.993 *S* – 19807128 ULK1_HUMAN 317 ASPPSLGEM 0.732 *S* PKC∝ 22932492 ULK1_HUMAN 403 GRTPSPSPP 0.924 *S* GSK3β 22932492 ULK1_HUMAN 405 TPSPSPPCS 0.987 *S* GSK3β 21383122 ULK1_HUMAN 411 PCSSSPSPS 0.895 *S* GSK3β, ERK1 22932492 ULK1_HUMAN 460 TPRSSAIRR 0.962 *S* – 22932492 ULK1_HUMAN 465 AIRRSGSTS 0.984 *S* – 16964243 ULK1_HUMAN 467 RRSGSTSPL 0.986 *S* AMPK, PKCδ 19807128 ULK1_HUMAN 469 SGSTSPLGF 0.975 *S* ERK1 21383122 ULK1_HUMAN 477 FARASPSPP 0.989 *S* ERK1 18691976 ULK1_HUMAN 479 RASPSPPAH 0.891 *S* CDC2, CDK5 21383122 18691976 19807128 ULK1_HUMAN 495 ARKMSLGGG 0.969 *S* AMPK 22932492 AKT PKA PKCδ PKCμ ULK1_HUMAN 533 RGGRSPRPG 0.994 *S* CDK5 21383122 ULK1_HUMAN 544 APEHSPRTS 0.986 *S* CDC2 22932492 CDK5 ULK1_HUMAN 556 CRLHSAPNL 0.807 *S* AMPK 21383122 21205641 18669648 18846507 ULK1_HUMAN 694 GRSFSTSRL 0.987 *S* AKT 22932492 ULK1_HUMAN 716 PDPGSTESL 0.894 *S* CK1 22932492 ULK1_HUMAN 719 GSTESLQEK 0.574 *S* CK1 22932492 ULK1_HUMAN 747 AGGTSSPSP 0.551 *S* – 16964243 ULK1_HUMAN 761 GSPPSGSTP 0.575 *S* – 16964243 ULK1_HUMAN 775 TRMFSAGPT 0.983 *S* – 21383122 ULK1_HUMAN 866 ALKGSASEA 0.886 *S* – 19807128 ULK1_HUMAN 1042 ERRLSALLT 0.99 *S* AMPK 19807128 PKA ULK1_HUMAN 456 TQFQTPRSS 0.971 *T* – 18669648 ULK1_HUMAN 636 DFPKTPSSQ 0.671 *T* CDK5 22932492 ULK1_HUMAN 695 RSFSTSRLT 0.975 *T* – 22932492 R. Randhawa et al. / Gene 562 (2015) 40–49 47

Fig. 5. ULK1 domain structure with newly identiﬁed phosphorylation sites. NetPhos algorithm was used for prediction of phosphorylation sites at serine (S), threonine (T) and tyrosine (Y) residues. A total of 58 new phosphorylation sites were identiﬁed out of which 25 are in kinase domain, 29 are in Ser/Thr rich domain and only 4 are in CT domain.

Identification of palmitoylation sites was an important component interacting with ULK1 have been identified in this analysis from the of this study, which adds palmitic acid resulting in an increase of hydro- PPI network generated by STRING as shown in Fig. 7. The main proteins phobicity of proteins and helps in their association to the membranes which were found to interact with ULK1 are gamma-amino butyric acid (Wu et al., 2014). The identification of these sites would help to under- receptor-associated protein (GABARAP), Activating Molecule in Beclin- stand the complex processes such as sub-cellular trafficking between 1-Regulated Autophagy (AMBRA1), Mammalian target of rapamycin the membrane compartments (Rocks et al., 2005) and protein–protein (mTOR), Regulatory-associated protein of mTOR (RPTOR), RB1- interactions (Joyoti, 2004) that occur due to palmitoylation. Many au- inducible coiled-coil protein 1 (RB1CC1), Autophagy-related protein tophagy related proteins have been studied for the palmitoylation 13 (ATG13), Beclin-1 (BECN1) and Synaptic Ras GTPase-activating pro- sites (Mercer et al., 2009). In mammals, ATG101 an important autopha- tein 1 (SYNGAP1) (Cecconi et al., 2007; Chano et al., 2007; Sun et al., gy related protein, interacts with ATG13 (ULK1 interacting protein) 2010; Van Humbeeck et al., 2011; Pagliarini et al., 2012; Tameno et al., which is a component of macroautophagy (Mercer et al., 2009). 2012; Johnson et al., 2013; Koike et al., 2013; Tang et al., 2013; Wirth ATG101 is localized to the isolation membrane or phagophore which et al., 2013; Yang et al., 2013; Yu et al., 2013). These proteins interact surrounds the materials to be degraded in the lysosome. The interaction with ULK1 and have been shown to play important roles in autophagy of ATG101 with the phagophore may be due to the palmitoylation site as well as in associated pathways. Further, exploration of structural ele- present at the third amino acid, i.e. cysteine (Mercer et al., 2009). Iden- ments of these interacting proteins could provide beneficial information tification of these sites would help us in exploring diverse interactions which might provide clues for their association with a myriad of biolog- with ULK1 and their sub-cellular localizations. These palmitoylation ical regulatory processes. sites were obtained from CSS-PALM. In ULK1, we found 4 palmitoylation sites at positions 426, 927, 1003 and 1049 (Fig. 6) which could further 4. Conclusion be investigated for their role in trafficking and protein–protein interactions. These are the putative positions where the palmitic acid may at- Autophagy is a lysosomal degradation pathway for damaged cyto- tach to the protein at cysteine residues for its attachment to the plasmic organelles or cytosolic components of a cell which has recently membrane. been shown to be involved in many diseases. It is in debate for its modulation for therapeutic interventions and better management of 3.5. ULK1 interacts with various proteins diseases in which its dysregulation has been shown. ULK1 mammalian homologue of autophagy related gene-1 (Atg1) plays a central role in The functional property of a protein could be analyzed by exploring autophagy pathway. In this study, we in silico identified TFBS which its interactions with other proteins and their involvement in different were present throughout ULK1 and were of zinc coordinating class; biochemical pathways (Perkins et al., 2010). Several proteins CATTT haplotype (0.774) as prominent; four nsSNPs which could have

Fig. 6. The putative palmitoylation sites in ULK1. Four novel putative palmitoylation sites were identiﬁed applying CSS-PALM. 48 R. Randhawa et al. / Gene 562 (2015) 40–49

Fig. 7. ULK1 interacting partners. STRING (version 9.05) was used to find ULK1 interacting proteins; the scale represents the various search parameters and genes that are mapped through co-expression analysis. harmful effect on ULK1 protein (Table 7); and 87 and 4 phosphorylation Fullgrabe, J., Klionsky, D.J., Joseph, B., 2014. The return of the nucleus: transcriptional and epigenetic control of autophagy. Nat. Rev. Mol. Cell Biol. 15, 65–74. and palmitoylation sites, respectively. We suggest that this information Gotea, V., Ovcharenko, I., 2008. DiRE: identifying distant regulatory elements of co- could be utilized in experimental studies to further gain insights into the expressed genes. Nucleic Acids Res. 36, W133–W139. functions of ULK1. Heinrich, G., Pagtakhan, C.J., 2004. Both 5′ and 3′ flanks regulate Zebrafish brain-derived neurotrophic factor gene expression. BMC Neurosci. 5, 19. Supplementary data to this article can be found online at http://dx. Henckaerts, L., Cleynen, I., Brinar, M., John, J.M., Van Steen, K., Rutgeerts, P., Vermeire, S., doi.org/10.1016/j.gene.2015.02.056. 2011. Genetic variation in the autophagy gene ULK1 and risk of Crohn's disease. Inflamm. Bowel Dis. 17, 1392–1397. fl Johnson, S.C., Rabinovitch, P.S., Kaeberlein, M., 2013. mTOR is a key modulator of ageing Con ict of interest and age-related disease. Nature 493, 338–345. Joyoti, B., 2004. Protein palmitoylation and dynamic modulation of protein function. Curr. The authors declare that they have no conflict of interest. Sci. 87, 6. Katoh, K., Misawa, K., Kuma, K., Miyata, T., 2002. MAFFT: a novel method for rapid multiple sequence alignment based on fast Fourier transform. Nucleic Acids Res. 30, Acknowledgments 3059–3066. Koike, M., Tanida, I., Nanao, T., Tada, N., Iwata, J., Ueno, T., Kominami, E., Uchiyama, Y., 2013. Enrichment of GABARAP relative to LC3 in the axonal initial segments of neu- This work is supported by funding (BT/PR6784/GBD/27/466/2012) rons. PLoS One 8, e63568. from Department of Biotechnology, India to HC. Kreegipuu, A., Blom, N., Brunak, S., 1999. PhosphoBase, a database of phosphorylation sites: release 2.0. Nucleic Acids Res. 27, 237–239. Kundu, M., Lindsten, T., Yang, C.Y., Wu, J., Zhao, F., Zhang, J., Selak, M.A., Ney, P.A., References Thompson, C.B., 2008. Ulk1 plays a critical role in the autophagic clearance of mitochondria and ribosomes during reticulocyte maturation. Blood 112, 1493–1502. Barrett, J.C., Fry, B., Maller, J., Daly, M.J., 2005. Haploview: analysis and visualization of LD Kwon, A.T., Arenillas, D.J., Worsley Hunt, R., Wasserman, W.W., 2012. oPOSSUM-3: ad- and haplotype maps. Bioinformatics 21, 263–265. vanced analysis of regulatory motif over-representation across genes or ChIP-Seq Beerli, R.R., Dreier, B., Barbas III, C.F., 2000. Positive and negative regulation of endogenous datasets. G3 (Bethesda) 2, 987–1002. genes by designed transcription factors. Proc. Natl. Acad. Sci. U. S. A. 97, 1495–1500. Mack, H.I., Zheng, B., Asara, J.M., Thomas, S.M., 2012. AMPK-dependent phosphorylation Blom, N., Gammeltoft, S., Brunak, S., 1999. Sequence and structure-based prediction of eu- of ULK1 regulates ATG9 localization. Autophagy 8, 1197–1214. karyotic protein phosphorylation sites. J. Mol. Biol. 294, 1351–1362. Marín-Martín, F.R., Soler-Rivas, C., Martín-Hernández, R., Rodriguez-Casado, A., 2014. A Bononi, A., Agnoletto, C., De Marchi, E., Marchi, S., Patergnani, S., Bonora, M., Giorgi, C., comprehensive in silico analysis of the functional and structural impact of Missiroli, S., Poletti, F., Rimessi, A., Pinton, P., 2011. Protein kinases and phosphatases nonsynonymous SNPs in the ABCA1 transporter gene. Cholesterol 19. in the control of cell fate. Enzym. Res. 2011, 329098. Mercer, C.A., Kaliappan, A., Dennis, P.B., 2009. A novel, human Atg13 binding protein, Cecconi, F., Di Bartolomeo, S., Nardacci, R., Fuoco, C., Corazzari, M., Giunta, L., Romagnoli, Atg101, interacts with ULK1 and is essential for macroautophagy. Autophagy 5, A., Stoykova, A., Chowdhury, K., Fimia, G.M., Piacentini, M., 2007. A novel role for au- 649–662. tophagy in neurodevelopment. Autophagy 3, 506–508. Ng, P.C., Henikoff, S., 2003. SIFT: predicting amino acid changes that affect protein func- Chan, E.Y., Longatti, A., McKnight, N.C., Tooze, S.A., 2009. Kinase-inactivated ULK proteins tion. Nucleic Acids Res. 31, 3812–3814. inhibit autophagy via their conserved C-terminal domains using an Atg13- Olsen, J.V., Vermeulen, M., Santamaria, A., Kumar, C., Miller, M.L., Jensen, L.J., Gnad, F., Cox, J., independent mechanism. Mol. Cell. Biol. 29, 157–171. Jensen, T.S., Nigg, E.A., Brunak, S., Mann, M., 2010. Quantitative phosphoproteomics re- Chano, T., Okabe, H., Hulette, C.M., 2007. RB1CC1 insufficiency causes neuronal atrophy veals widespread full phosphorylation site occupancy during mitosis. Sci. Signal. 3, ra3. through mTOR signaling alteration and involved in the pathology of Alzheimer's dis- Pabo, C.O., Peisach, E., Grant, R.A., 2001. Design and selection of novel Cys2His2 zinc finger eases. Brain Res. 1168, 97–105. proteins. Annu. Rev. Biochem. 70, 313–340. del Sol, A., Fujihashi, H., Amoros, D., Nussinov, R., 2006. Residue centrality, functionally Pagliarini, V., Wirawan, E., Romagnoli, A., Ciccosanti, F., Lisi, G., Lippens, S., Cecconi, F., important residues, and active site shape: analysis of enzyme and non-enzyme fam- Fimia, G.M., Vandenabeele, P., Corazzari, M., Piacentini, M., 2012. Proteolysis of ilies. Protein Sci. 15, 2120–2128. Ambra1 during apoptosis has a role in the inhibition of the autophagic pro-survival Deretic, V., Saitoh, T., Akira, S., 2013. Autophagy in infection, inflammation and immunity. response. Cell Death Differ. 19, 1495–1504. Nat. Rev. Immunol. 13, 722–737. Perkins, J.R., Diboun, I., Dessailly, B.H., Lees, J.G., Orengo, C., 2010. Transient protein–pro- Edgar, R.C., 2004. MUSCLE: multiple sequence alignment with high accuracy and high tein interactions: structural, functional, and network properties. Structure 18, throughput. Nucleic Acids Res. 32, 1792–1797. 1233–1243. R. Randhawa et al. / Gene 562 (2015) 40–49 49

Ramensky, V., Bork, P., Sunyaev, S., 2002. Human non-synonymous SNPs: server and sur- (mTor) mediates tau protein dyshomeostasis: implication for Alzheimer disease. vey. Nucleic Acids Res. 30, 3894–3900. J. Biol. Chem. 288, 15556–15570. Rocks, O., Peyker, A., Kahms, M., Verveer, P.J., Koerner, C., Lumbierres, M., Kuhlmann, J., Thorisson, G.A., Smith, A.V., Krishnan, L., Stein, L.D., 2005. The International HapMap Pro- Waldmann, H., Wittinghofer, A., Bastiaens, P.I., 2005. An acylation cycle regulates lo- ject Web site. Genome Res. 15, 1592–1593. calization and activity of palmitoylated Ras isoforms. Science 307, 1746–1752. Van Humbeeck, C., Cornelissen, T., Hofkens, H., Mandemakers, W., Gevaert, K., De Sandelin, A., Alkema, W., Engstrom, P., Wasserman, W.W., Lenhard, B., 2004. JASPAR: an Strooper, B., Vandenberghe, W., 2011. Parkin interacts with Ambra1 to induce open-access database for eukaryotic transcription factor binding profiles. Nucleic mitophagy. J. Neurosci. 31, 10249–10261. Acids Res. 32, D91–D94. Wang, Y.T., Tsai, C.F., Hong, T.C., Tsou, C.C., Lin, P.Y., Pan, S.H., Hong, T.M., Yang, P.C., Sung, Segal, D.J., Goncalves, J., Eberhardy, S., Swan, C.H., Torbett, B.E., Li, X., Barbas III, C.F., 2004. T.Y., Hsu, W.L., Chen, Y.J., 2010. An informatics-assisted label-free quantitation strat- Attenuation of HIV-1 replication in primary human cells with a designed zinc finger egy that depicts phosphoproteomic profiles in lung cancer cell invasion. transcription factor. J. Biol. Chem. 279, 14509–14519. J. Proteome Res. 9, 5582–5597. Shang, L., Chen, S., Du, F., Li, S., Zhao, L., Wang, X., 2011. Nutrient starvation elicits an acute Watanabe, N., Osada, H., 2012. Phosphorylation-dependent protein–protein interaction autophagic response mediated by Ulk1 dephosphorylation and its subsequent disso- modules as potential molecular targets for cancer therapy. Curr. Drug Targets 13, ciation from AMPK. Proc. Natl. Acad. Sci. U. S. A. 108, 4788–4793. 1654–1658. Smotrys, J.E., Linder, M.E., 2004. Palmitoylation of intracellular signaling proteins: regula- Whitfield, T.W., Wang, J., Collins, P.J., Partridge, E.C., Aldred, S.F., Trinklein, N.D., Myers, tion and function. Annu. Rev. Biochem. 73, 559–587. R.M., Weng, Z., 2012. Functional analysis of transcription factor binding sites in Sun, C., Southard, C., Witonsky, D.B., Kittler, R., Di Rienzo, A., 2010. Allele-specificdown- human promoters. Genome Biol. 13, R50. regulation of RPTOR expression induced by retinoids contributes to climate adapta- Wirth, M., Joachim, J., Tooze, S.A., 2013. Autophagosome formation—the role of ULK1 and tions. PLoS Genet. 6, e1001178. Beclin1–PI3KC3 complexes in setting the stage. Semin. Cancer Biol. 23, 301–309. Szklarczyk, D., Franceschini, A., Kuhn, M., Simonovic, M., Roth, A., Minguez, P., Doerks, T., Wu, W., Tian, W., Hu, Z., Chen, G., Huang, L., Li, W., Zhang, X., Xue, P., Zhou, C., Liu, L., Zhu, Stark, M., Muller, J., Bork, P., Jensen, L.J., von Mering, C., 2011. The STRING database in Y., Li, L., Zhang, L., Sui, S., Zhao, B., Feng, D., 2014. ULK1 translocates to mitochondria 2011: functional interaction networks of proteins, globally integrated and scored. and phosphorylates FUNDC1 to regulate mitophagy. EMBO Rep. 15, 566–575. Nucleic Acids Res. 39, D561–D568. Xu, H., Yu, H., Zhang, X., Shen, X., Zhang, K., Sheng, H., Dai, S., Gao, H., 2013. UNC51-like Tameno, H., Chano, T., Ikebuchi, K., Ochi, Y., Arai, A., Kishimoto, M., Shimada, T., Hisa, Y., kinase 1 as a potential prognostic biomarker for hepatocellular carcinoma. Int. Okabe, H., 2012. Prognostic significance of RB1-inducible coiled-coil 1 in salivary J. Clin. Exp. Pathol. 6, 711–717. gland cancers. Head Neck 34, 674–680. Yang, H., Rudge, D.G., Koos, J.D., Vaidialingam, B., Yang, H.J., Pavletich, N.P., 2013. mTOR Tamura, K., Peterson, D., Peterson, N., Stecher, G., Nei, M., Kumar, S., 2011. MEGA5: molec- kinase structure, mechanism and regulation. Nature 497, 217–223. ular evolutionary genetics analysis using maximum likelihood, evolutionary distance, Yu, M., Gou, W.F., Zhao, S., Xiao, L.J., Mao, X.Y., Xing, Y.N., Takahashi, H., Takano, Y., Zheng, and maximum parsimony methods. Mol. Biol. Evol. 28, 2731–2739. H.C., 2013. Beclin 1 expression is an independent prognostic factor for gastric carcino- Tang, J., Deng, R., Luo, R.Z., Shen, G.P., Cai, M.Y., Du, Z.M., Jiang, S., Yang, M.T., Fu, J.H., Zhu, mas. Tumour Biol. 34, 1071–1083. X.F., 2012. Low expression of ULK1 is associated with operable breast cancer progres- Zhou, F., Xue, Y., Yao, X., Xu, Y., 2006. CSS-PALM: palmitoylation site prediction with a sion and is an adverse prognostic marker of survival for patients. Breast Cancer Res. clustering and scoring strategy (CSS). Bioinformatics 22, 894–896. Treat. 134, 549–560. Tang, Z., Bereczki, E., Zhang, H., Wang, S., Li, C., Ji, X., Branca, R.M., Lehtio, J., Guan, Z., Filipcik, P., Xu, S., Winblad, B., Pei, J.J., 2013. Mammalian target of rapamycin