Genome-wide in silico characterization and stress induced expression analysis of BcL-2 associated athanogene (BAG) family in Musa spp.

Programmed cell death (PCD) is a genetically controlled process for the selective removal of damaged cells. Though understanding about plant PCD has improved over years, the mechanisms are yet to be fully deciphered. Among the several molecular players of PCD in plants, B cell lymphoma 2 (Bcl-2)-associated athanogene (BAG) family of co-chaperones are evolutionary conserved and regulate cell death, growth and development. In this study, we performed a genome-wide in silico analysis of the MusaBAG gene family in a globally important fruit crop banana. Thirteen MusaBAG genes were identified, out of which MusaBAG1, 7 and 8 genes were found to have multiple copies. MusaBAG genes were distributed on seven out of 11 chromosomes in banana. Except for one paralog of MusaBAG8 all the other 12 proteins have characteristic BAG domain. MusaBAG1, 2 and 4 have an additional ubiquitin-like domain whereas MusaBAG5-8 have a calmodulin binding motif. Most of the MusaBAG proteins were predicted to be localized in the nucleus and mitochondria or chloroplast. The in silico cis-regulatory element analysis suggested regulation associated with photoperiodic control, abiotic and biotic stress. The phylogenetic analysis revealed 2 major clusters. Digital gene expression analysis and quantitative real-time RT-PCR depicted the differential expression pattern of MusaBAG genes under abiotic and biotic stress conditions. Further studies are warranted to uncover the role of each of these proteins in growth, PCD and stress responses so as to explore them as candidate genes for engineering transgenic banana plants with improved agronomic traits.

www.nature.com/scientificreports/ whereas AtBAG6 is involved in heat-stress and also inducing PCD 3,7 . Plant BAG genes are regulated under biotic stress and upon pathogenic challenge. BAG4 was induced in the leaf tissue of Medicago truncatula when infected with bacterial pathogen Xylella fastidiosa 3 . Similarly, in the seedlings of Hordeum vulgare, BAG1 was induced when challenged with Blumeria 3 . The transcript levels of AtBAG6 were elevated on treatment with salicylic acid in Arabidopsis. AtBAG6 is proteolytically cleaved by aspartyl proteases to trigger autophagy and induce PCD, conferring fungal disease resistance 8 . AtBAG7 is an endoplasmic reticulum (ER) localized protein, playing a role in unfolded protein response (UPR) signal transduction. Under normal condition it remains associated with the luminal binding protein (AtBiP) and during ER stress condition it dissociates from AtBiP and gets translocated to the nucleus to interact with WRKY-DNA binding protein [WRKY29] 9 . AtBAG5 is a mitochondrial localized protein which when present in abundance induces reactive oxygen species (ROS) production. AtBAG5 sequesters HSC70 protein and therefore the low levels of free HSC70 is involved in ROS alleviation and PCD. AtBAG5 mutants showed delayed leaf senescence 10 . AtBAG1 levels are self-regulated and its levels are important for normal growth and development of the plant. It has been shown to be involved in proteasomal degradation of misfolded and unimported plastid proteins assisted by HSC70-4 11 . A BAG domain containing protein HSG1 in grapevine was induced at 45 °C for 60 min in both leaves and berries. Moreover, when HSG1 was overexpressed in Arabidopsis faster floral transition and triggering of flowering promoter CONSTANS was observed without any morphological changes than the control plant indicating its role in photoperiodism 12 . In a previous study, overexpressing native ubiquitin-like domain-containing protein BAG1 coding sequence in banana plants conferred resistance to Fusarium wilt disease 13 .
In this work, we present the first comprehensive in silico analysis of the BAG gene family from banana (Musa spp. AA genome) and experimentally shown their differential expression levels under biotic and abiotic stress conditions. To gain more insight, we studied the gene structure, organization, protein structural properties, phylogeny, and analyzed the expression profiles of the BAG gene family in banana plants under stress conditions. This conscientious analysis of BAG genes in banana plants provides a strong ground for further functional characterization of each of these genes so as to utilize them for developing elite varieties resilient to abiotic and biotic stress.

Results
Musa database search for BAG domain containing proteins. The Arabidopsis BAG genes (AtBAG1-AtBAG8) were used as query sequences to search for its homologs in the banana genome database. A total of 13 BAG genes were identified in the banana genome which were further categorized into 7 different types (Table 1) based on the AtBAG sequences. There were no homologs of AtBAG3 found in the banana genome. Most of the MusaBAG genes are present on the chromosome number 3, 4, and 7. But some BAG genes were also identified on the chromosomes 1, 2, 9 and 10 ( Fig. 1). Out of the 13 BAG genes MusaBAG1, 7 and 8 genes were found to have multiple copies in the genome of banana. MusaBAG1 has four different copies (paralogous genes) present on three different chromosomes (chr. 3, 4 and 7). Similarly, MusaBAG7 gene has two paralogs each on chromosome 2 and 9. MusaBAG8, exists as two different mRNA transcript variants located on the locus LOC103982527 of chromosome 4. A truncated paralogous gene, also denoted as MusaBAG8 (based on similarity search using AtBAG8), was also found at locus LOC103983382 of the same chromosome. The MusaBAG genes encode for a 200-450 amino acid residue protein, except MusaBAG6, which encodes for an 1119 amino acid residue protein. Some of the characteristics of MusaBAGs, such as length of protein and localization, are similar to BAG genes of Arabidopsis and rice, except for the copy numbers of the gene in the respective genomes. Most of the MusaBAG proteins have an isoelectric point (pI) in the alkaline pH range i.e., between pH of 9-10, whereas MusaBAG5 and 6 have a pI in the acidic range and MusaBAG4 and 8 in neutral. www.nature.com/scientificreports/ sequence motif consists of repeated stretches of highly charged amino acids like lysine (K), glutamate (E) and aspartate (D) (Fig. 3).

Structural analysis.
The structural analysis of the MusaBAG proteins was done based on the predicted structures, obtained by submitting individual protein sequences to I-TASSER server, for template based iterative simulation for prediction of structures. The predicted structures with highest C-score, from the information given by the server, were selected as representative structures for further analysis (Fig. 4). The structures of MusaBAG1 (all the four paralogs), MusaBAG2 and MusaBAG4 were comparable, with respect to the presence and organization of the major domains. These structures have a highly organized BAG domain and ubiquitinlike domain. The N-terminal characteristic 12 amino acid conserved sequence motif present in MusaBAG1, MusaBAG2 and MusaBAG4 (as described in the previous section) was found to be a hairpin loop. The ubiquitinlike domain showed high conservation, with strictly four β-sheets and two α-helices. Amongst the four β-sheets, two were organized as central β-sheets, with one α-helix on each side of the β-sheet. This type of ubiquitin-like domain structure was very similar to the ubiquitin-like domain of yeast ribosomal biogenesis protein Ytm1 14 .
The BAG domain of all the MusaBAG proteins consists of the typical three α-helix bundle structure, similar to as described by Sondermann et al. 15 . The IQ-calmodulin binding motif consists of two small α-helices connected by a hairpin loop, which is in continuation with the first helix of the BAG domain, as observed in MusaBAG5, 6, and 8. In MusaBAG7, this motif was a complete single α-helix, which is continued as the first helix of the BAG domain. The KED sequence motif, which is present in both the paralogs of MusaBAG7, was structurally consisting of multiple α-helices connected by multiple hairpin loops. Structurally, this sequence motif showed interaction pockets similar to EF-hand like proteins, which upon docking showed probable calcium and magnesium binding sites (Supplementary Information S1).
Phylogenetic analysis and classification. Using the neighbor-joining method, phylogenetic analysis of all BAG proteins from banana (MusaBAG) was performed. From this phylogenetic tree, it was observed that there are two major phylogenetic clades of MusaBAG proteins (Fig. 5A). In both the clades MusaBAG1 was present, suggesting that MusaBAG1 originated or derived from the last common ancestral protein.
MusaBAG1_LOC103983243 and MusaBAG1_LOC103978318 were clustered in one clade whereas Musa-BAG1_LOC103990510 and MusaBAG1_LOC103977986 were in the other clade. MusaBAG4 and MusaBAG1_ LOC103978318 were grouped as one terminal monophyletic clade whereas both the MusaBAG8 were grouped as another monophyletic clade. Similarly, both the MusaBAG7 were grouped as one monophyletic clade, which is in turn branched with MusaBAG5. MusaBAG6 and MusaBAG1_LOC103990510 were grouped as one monophyletic clade. Interestingly, MusaBAG2, which is structurally similar to MusaBAG1 and MusaBAG4 was an out-group to MusaBAG1 (LOC103990510 and LOC103977986), MusaBAG6 and MusaBAG7. Another un-rooted circular phylogenetic tree was constructed using neighbor-joining method of all BAG proteins from banana (MusaBAG), Arabidopsis (AtBAG) and rice (OsBAG) along with a BAG1 from mouse (MmBAG1) and BAG5 from human (HsBAG5) (Fig. 5B). Generally, BAG1, 5 and 7 were clustered in one clade whereas BAG4, 6, 8 were clustered in another clade. However, 2 out of the 4 copies of MusaBAG1 proteins were clustered along with MusaBAG8 proteins. HsBAG5 clustered in a distant clade than plant BAG5 proteins. An www.nature.com/scientificreports/ unusual observation from this phylogenetic tree was that BAG2, and 3 from rice and banana are clustered as closely related proteins whereas their homologs from Arabidopsis were clustered in a separate clade. In this phylogenetic tree MusaBAG2 was clustered along with MusaBAG8, AtBAG8 and MusaBAG1 (LOC103983243 and LOC103978318), which was different compared with the phylogenetic analysis done only for MusaBAG sequences. Thus it was difficult to trace back the origin and evolutionary trajectory of MusaBAG2 based on this phylogenetic analysis. Subcellular localization. Subcellular localization of proteins is important to fairly ascertain their function.
Several online tools are available to predict protein localization in the cell; that utilizes protein characteristics such as amino acid composition, signal peptides, homology based prediction and combination of these. The subcellular localization of MusaBAG proteins were predicted using DeepLoc 1.0, PSORT, Yloc, Target P, WoLF PSORT and Plant PLoc servers. According to these predictions all the MusaBAG proteins were destined to the nucleus ( Table 2). MusaBAG1 proteins were predicted to be present both in nucleus and mitochondria. Musa-BAG5, MusaBAG7 (GenBank accession ID-XM_009390167.2) and MusaBAG8 were also shown to be localized in the chloroplast, mitochondria or microbody (peroxisomes). MusaBAG6 was exclusively predicted to be localized in the nucleus by all the 6 servers. MusaBAG4 and MusaBAG7 (GenBank accession ID-XM_009387694.2) were predicted to be localized in the cytoplasm as well.
Putative cis-element in promoter sequences. The conserved cis-regulatory element present in the promoter regions are known to be the binding sites for several transcription factors, that in turn regulate the spatial and temporal transcriptional activity. A large number of such conserved cis-elements were identified in the putative promoter sequences of MusaBAG genes when queried in the PlantCARE and PlantPAN databases. Multiple stress-related, light responsive and hormonal regulated cis-elements were identified in the promoter regions of MusaBAG genes (Table 3). Elements related to defense responses (STRE, WRKY and TC-rich    , B2;ARF' bHLH, CSD, CG-1, EIN3,  GATA;tify, HD-ZIP, WOX, MADF, MYB;SANT,  SBP, SOX;YABBY, TCP, Alpha-amylase, WRKY,  MADS Box, LFY, zF-HD, NF-YA, NF-YB, MF-YC,  LEA_5, ARID, BES1, C2H2, C3H, FAR1, (Fig. 6A). In contrast, no such difference in the expression of MusaBAG genes was observed under osmotic stress condition, when compared with respective root osmotic control. However, only MusaBAG6 mRNA expression showed 1.5 fold overexpression, as compared to the control (Fig. 6B). Based on the Sidak's multiple comparison test of hypothesis (P < 0.05), significant difference was observed in MusaBAG1, 4, 6, 7 and 8 genes under biotic stress conditions and Musa-BAG6 gene under abiotic stress condition, with respect to their controls.

Expression analysis of MusaBAG genes.
Given the fact that plants initiate a protective response during stress conditions so as to maintain cellular homeostasis, genes like MusaBAG seem to play an important role. To ascertain the importance of MusaBAG genes we studied their differential regulation in banana plants exposed to various abiotic and biotic stress conditions. Overall there was a significant difference in the expression pattern of MusaBAG1_LOC103978318, MusaBAG5_LOC103992959, MusaBAG6_LOC104000818 and MusaBAG7_ LOC103973192 under abiotic and biotic stress conditions as compared to the unstressed control (Fig. 7). In case of sodium chloride stress, MusaBAG5_LOC103992959 and MusaBAG6_LOC104000818 were upregulated 3 and 2.5 times, respectively; whereas MusaBAG8_LOC103982527-X2 remains undetected in the treated samples (Fig. 7A). However, only MusaBAG7_ LOC103973192 was found to be upregulated 2 times in PEG induced Table 3. Cis-regulatory elements present in the putative promoter sequences of the MusaBAG genes. www.nature.com/scientificreports/ stress and MusaBAG5_LOC103992959 remains undetected (Fig. 7B). In biotic stress, the Fusarium oxysporum f. sp. cubense race 1 (Foc R1) infected plants showed 3 times upregulation of MusaBAG1_LOC103978318 whereas, MusaBAG5_LOC103992959, MusaBAG6_LOC104000818 and MusaBAG7_LOC103973192 were upregulated around 10 times (Fig. 7C). Further, no amplification was observed for MusaBAG1_LOC103977986 and Musa-BAG8_LOC103982527-X2 under this condition. Furthermore, under this experimental setup no detectable expression was observed for MusaBAG2, MusaBAG8_LOC103983382, MusaBAG8_LOC103982527-X1 in any of the treated or control plants.
Protein-protein interactions. BAG proteins are known to carry out their function by interacting with other proteins. So to identify the potential interactors of MusaBAG proteins the interacting partners for each MusaBAG protein was searched using the STRING database version 11 with medium confidence of 0.400 (Supplementary Fig. S2). The predicted interactome was mainly based on textmining, coexpression and experimental data. All the MusaBAG1 paralogs and MusaBAG2 had common interacting proteins ( Supplementary Fig. S2A). The potential interactors include uncharacterized GPI-anchored protein, putative lipoxygenase homology domain containing protein 1, Tra family proteins and IQ-calmodulin-binding motif family protein. Moreo-   Fig. S2B). The direct interactors of MusaBAG4 were vacuolar protein sorting-associated protein 32 homolog 1, IQ-calmodulin-binding motif family protein and putative ubiquitin associated protein and indirectly interacted with putative ALG-2 interacting protein X. Similar to MusaBAG1 and MusaBAG2, MusaBAG4 also showed interaction with autophagy related proteins through the ubiquitin associated protein. Through textmining, MusaBAG5 was predicted to be interacting with IQ-calmodulin-binding motif family protein and an uncharacterized protein which in turn were predicted to interact with several other proteins such as luminal binding proteins, BAG domain proteins and other uncharacterized proteins ( Supplementary Fig. S2C). MusaBAG6 showed a complex network of proteins forming distinct clusters involved in DNA processing, heat shock proteins and proteinases ( Supplementary Fig. S2D). MusaBAG6 was predicted to be interacting with another BAG domain containing proteins and several heat shock proteins. It was predicted to be interacting with putative replication associated recombination protein A which further interacted with three other protein domains namely, DNA polymerase delta catalytic subunit, DNA helicase and putative protein BREAST CANCER SUSCEPTIBILITY 1 homolog. MusaBAG7 paralogs were predicted to be interacting with luminal binding proteins (sensors of endoplasmic stress), heat shock protein binding proteins and calnexin homolog 1 ( Supplementary Fig. S2E). Moreover, an indirect interaction with other BAG domain proteins was also predicted for MusaBAG7. The interactors for MusaBAG8 were very different from the other MusaBAG proteins. MusaBAG8 interactors were predicted based on textmining and showed lymphoid organ expressed yellow head virus receptor protein, growth regulator related protein, HEAT repeat family proteins, EPS4 proteins and STE_PAK_STE20_MST__like_1-STE kinase homologs to sterile 7 as potential interacting proteins (Supplementary Fig. S2F). Through this network, MusaBAG8 was also predicted to be interacting with other proteins indirectly that included, heat shock proteins, putative calcium binding protein 32 like, zinc finger CCCH domain containing protein 45 and cleavage and polyadenylation specificity factors. Among the MusaBAG proteins, all of them were predicted to be interacting with each other except for MusaBAG8 (Fig. 8). At high confidence (0.700) in this interactome strong interaction was predicted between MusaBAG6, Heat shock 70 kDa protein, putative replication associated recombination protein A and putative 78 kDa glucose-regulated protein homolog (Fig. 8). Through textmining at a high confidence level an interaction between MusaBAG4, MusaBAG7, and MusaBAG6 was predicted.

Discussion
The BAG gene family encodes for a multifunctional group of proteins that regulate diverse pathways such as cell signaling, stress responses (both abiotic and biotic), growth and development 1 . BAG proteins were acknowledged to be the critical regulators of PCD in plants and their importance in maintaining homeostasis was established. Inadequacy in the knowledge of plant PCD as compared to that in animals have always been a limitation in engineering improved traits in crop plants. Since BAG proteins seem to be the core regulators, understanding this network and pathway will aid in modulating plant PCD towards environmental stimuli (both abiotic and biotic). In this study we carried out a genome wide search to identify BAG genes in an important fruit crop, banana www.nature.com/scientificreports/ using Arabidopsis BAG genes as query sequences. Banana is sensitive to extremes of environmental conditions and cannot sustain under drought, high salinity and low temperature 16,17 . Banana plants are also susceptible to many pests and diseases that cause severe yield losses 18 . In all these events PCD seems to play a crucial role, and thus studying BAG genes in banana plants is warranted. A total of 13 MusaBAG genes were identified in the whole genome of banana (Musa acuminata) dispersed on seven out of the eleven chromosomes; namely Musa-BAG1 (4 paralogs), MusaBAG2, MusaBAG4, MusaBAG5, MusaBAG6, MusaBAG7 (2 paralogs) and MusaBAG8 (1 paralog and 2 transcript variants). Multiple members of MusaBAG genes in the banana genome thus reveal the importance and diverse role of these genes in various cellular functions. All plant BAG proteins contain the typical BAG domain located towards the C terminus that interacts with the ATPase domain of the HSP70 protein required for its co-chaperone activity 7 . All the MusaBAG proteins showed the presence of this characteristic BAG domain except MusaBAG8 (LOC103983382). In addition, MusaBAG1, MusaBAG2 and MusaBAG4 contain ubiquitin-like domain for recruitment of proteasome degradation machinery for the degradation of misfolded and unimported chloroplast proteins in the cytosol 11,19 . The ubiquitin-like domain present in MusaBAG1, 2 and 4 is the site of ubiquitination by E1, E2 and E3 ligases that eventually results in degradation either through proteasomal degradation or autophagy 20 . An E3 ubiquitin ligase, enhanced blight and blast resistance 1 (EBR1), interacts with rice OsBAG4 protein having ubiquitin-like domain targeting it for ubiquitination-mediated degradation resulting in susceptibility to Xanthomonas oryzae pv. oryzae and Magnaporthe oryzae 19 . MusaBAG5, MusaBAG6, MusaBAG7 and MusaBAG8 contain CaM-binding motif, named as IQ motif suggesting its role in the calcium-calmodulin signaling pathway. This IQ motif present in the BAG proteins preferentially binds the calcium-free state of CaM. In the presence of calcium, there is release of Hsc70 due to competitive binding between CaM and Hsc70 to the AtBAG5 eventually preventing PCD in Arabidopsis plants 10 . Presence of the KED sequence motif in MusaBAG7 indicates its importance in wound healing and maintaining homeostasis during cytosolic acidification created through MAP kinase and defense gene activation 21 . The electron density map of this motif shows a disordered hand domain-like structure having interaction pockets similar to that of the EF-hand proteins, suggesting possible roles in metal (Ca 2+ and Mg 2+ ) binding 22 . The predicted structures of MusaBAG proteins displayed a typical BAG domain consisting of a three-helix bundle. Helices 2 and 3 contact subdomains IB and IIB of the ATPase mainly utilizing residues Glu212, Asp222, Arg237, and Gln245 of mammalian BAG1 protein 15 . Arg237 is highly conserved in all known BAG proteins, and their individual replacement with alanine resulted in BAG variants with substantially decreased ATPase activity and reduced affinities for the Hsc70 ATPase domain 15,23 . The relatedness among the MusaBAGs and homologs from other species was depicted using a phylogenetic tree and showed similarity between these protein sequences arranged together in different clades. The presence of paralogs on the same chromosome indicates the possibility of a tandem duplication event whereas those on different chromosomes are possibly due to segmental duplication events. However, multiple copies and duplication of genes with conserved or additional domains increases the functional divergence of these proteins and establishes its role in the important function of adaptability under changing environmental conditions.
Putative cis-acting elements related to light, stress, hormone and development were identified in the promoter regions of MusaBAG genes. The presence of these elements also indicate that they could be regulated by abiotic and biotic stress and different hormones. The MusaBAG genes could be regulated under stress conditions due to the presence of conserved stress-responsive cis-elements, such as DRE, ABRE, MYC, HSE, and MYB-binding sites. These genes also seem to be regulated by several transcription factors such as bZIP, WRKY, NAC and bHLH that are known to be triggered during abiotic stress and attack by pests or pathogens [24][25][26][27] . These results suggest that MusaBAG genes are regulated by multiple mechanisms and have functional divergence. Li et al. 8 showed that AtBAG6 is cleaved by caspase-like proteases and triggers autophagy in the host preventing Botrytis cinerea infection and confers basal resistance. Expression of BAG1 (BAG family molecular chaperone regulator 1-like) was also differentially regulated in banana cells upon Fusarium inoculation and toxins under in vitro conditions and further overexpression of this BAG1 in banana plants imparted resistance to Fusarium wilt disease 13 . BAG genes are also differentially regulated under abiotic stress conditions and provide protection against extremes of conditions. Overexpression of AtBAG4 in tobacco plants demonstrated tolerance to abiotic stresses such as UV light, cold, oxidants, and salt treatments 3 . In our study, we observed differential regulation of some of the MusaBAG genes under abiotic and biotic stress conditions. MusaBAG5_LOC103992959 and Musa-BAG6_LOC104000818 were upregulated in sodium chloride induced stress; whereas MusaBAG7_LOC103973192 was found to be upregulated in PEG induced stress. Significant upregulation of MusaBAG1_LOC103978318 MusaBAG5_LOC103992959, MusaBAG6_LOC104000818 and MusaBAG7_LOC103973192 was observed in banana plants infected with Foc R1. The total number of MusaBAG genes upregulated in the Foc R1 resistant cultivar Grand Naine as per the RNAseq data was higher in comparison to the gene expression profile obtained in the susceptible cultivar (Rasthali) indicating its role in resistance response. Noteworthy upregulation seen in the MusaBAG genes in Foc R1 treated Rasthali plants indicate that they are undergoing immense stress and have induced protective mechanisms involving these genes. Under abiotic stress, MusaBAG5, MusaBAG6 and MusaBAG7 genes seems to play a critical role. However, further investigations are required to better understand the role of these MusaBAG genes in early or late and susceptible or resistant responses in banana plants.
Each of these MusaBAG proteins are also compartmentalized in the cell for their specialized function. Most of the MusaBAG proteins were predicted to be located in the nucleus, mitochondria and chloroplast, MusaBAG6 was predicted to be localized in the nucleus. At their respective destinations MusaBAG proteins interact with their partners to carry out their specific functions. MusaBAG1, MusaBAG2 and MusaBAG4 interact with the autophagy-related proteins through the UBQ domain for protein degradation. MusaBAG4 also interacts with the vacuolar protein sorting associated proteins hinting at its role in endosomal sorting complexes required for transport of proteins to the vacuoles 28 www.nature.com/scientificreports/ chaperone activity targeting protein folding and maturation 30 . Overexpression of BiP in transgenic plants resulted in abiotic stress tolerance phenotype 31 . In previous studies, the Arabidopsis homolog of BAG7, has been shown to have interaction with several ER-stress related transcription factors, such as bZIP and WRKY29, and later translocated to the nucleus 25 . In our study, MusaBAG6 was shown to interact with various DNA binding proteins and previous studies demonstrated its role in cell death pathways associated with abiotic stress conditions 3,7,32 .
MusaBAG8 was predicted to be interacting with a unique set of proteins which were uncommon among the other MusaBAG proteins. Thus MusaBAG8 probably has diversified a lot and now is involved in other crucial unknown functions. Extensive functional studies are needed to ascertain the role of each of these BAG genes. It will be of considerable interest to experimentally validate the protein interactors of MusaBAG and to characterize the pathways involved in stress tolerance.
In conclusion, BAG proteins are an important link between molecular chaperones and protein degradation pathways (proteasome or autophagy) that maintain cellular homeostasis. The banana BAG gene family can be explored to understand their role in growth, development, fruit ripening, and fundamental responses to abiotic and biotic stresses. This knowledge can then be extended to other horticulture plants to understand the regulation of BAG proteins and exploit them in crop improvement programs.

Methods
Musa database search for BAG domain proteins. The full-length coding sequences of AtBAGs 6 were retrieved from NCBI database (https:// blast. ncbi. nlm. nih. gov/ Blast. cgi) 33 and were used to identify their respective homologs in the banana genome (Musa acuminata, https:// banana-genome-hub. south green. fr/) 34 denoted as MusaBAGs. The copy number of each of the MusaBAGs was determined by performing a nucleotide BLAST in the banana genome hub (https:// banana-genome-hub. south green. fr/) 34 . Each of the MusaBAG genes were mapped on the 11 chromosomes (A genome) of Musa acuminata DH-Pahang (AA) usingMap2GeneChrom v2.1 tool (http:// mg2c. iask. in) 35 . The full-length coding sequence and protein sequence of MusaBAG genes were also retrieved from the NCBI database (https:// blast. ncbi. nlm. nih. gov/ Blast. cgi) for further analysis 33 . The molecular weight and the isoelectric point of these proteins were analyzed using the ExPASy online tool (https:// web. expasy. org/ compu te_ pi/) 36 . The BAG genes of rice denoted as the OsBAGs were also retrieved from the NCBI database (https:// blast. ncbi. nlm. nih. gov/ Blast. cgi) 33 using the accession number mentioned in previous studies 37 .
Conserved domain analysis. The MusaBAG protein sequences, obtained in the previous step, were analyzed for domains using InterPro online tool (https:// www. ebi. ac. uk/ inter pro/) to annotate the conserved BAG domains 38 . The MusaBAG protein sequences were also aligned with the corresponding AtBAG protein sequences using PROMALS3D multiple sequence and structure alignment server (http:// proda ta. swmed. edu/ proma ls3d/ proma ls3d. php) to annotate the corresponding domains present in the AtBAGs, such as the IQ motif, KED sequence motif, conserved ExRPGG[ML/VV]QxR motif, as described by Yan et al. 6,39 . The annotated BAG domain of MusaBAG proteins were also confirmed by aligning them with the BAG domain of their respective Arabidopsis homologs. The BAG domains of MusaBAGs were aligned to examine their secondary structurebased conservation of amino acid residues along the entire domain, using PROMALS3D multiple sequence and structure alignment server (http:// proda ta. swmed. edu/ proma ls3d/ proma ls3d. php) 39 .
Structural analysis. The structures of MusaBAG proteins were obtained by submitting the sequences to Iterative Threading ASSEmbly Refinement (I-TASSER) server 40 . This server predicts the protein structure in a hierarchical method by iterative template-based fragment assembly simulation, using the structural templates provided by Protein Data Bank (PDB). Considering the confidence score (C-score) of the predicted models, the server provides five simulated structures for each of the MusaBAG protein sequences provided. The structure with the highest C-score value was selected, for each of the MusaBAG proteins, for further analysis.
Phylogenetic analysis and classification. All the 12 MusaBAG proteins of banana were clustered to make a phylogenetic tree using the neighbor joining method of 'amino acid substitution' type using Jones-Taylor-Thornton (JTT) model with 1000 bootstrap replications, uniform rating among sites and pairwise deletion gap treatment in MEGA version X 41 . Similarly, all the 28 BAG protein sequences (12 BAG proteins of banana, 8 BAG proteins of Arabidopsis, 6 BAG proteins from rice and two mammalian BAG proteins) were analyzed for their phylogenetic relationship by keeping the parameters same as the previous analysis and using Jones-Taylor-Thornton (JTT) model in MEGA version X 41 . As two out-groups, the full-length protein sequence of BAG1 of mouse (MmBAG1) (Uniprot ID-Q60739) and BAG5 (HsBAG5) (Uniprot ID-Q9UL15) of human were retrieved from Uniprot database 42 .
Prediction of subcellular localization. The subcellular localization of MusaBAG proteins was determined using web-based tools. These include: DeepLoc 1.0 (http:// www. cbs. dtu. dk/ servi ces/ DeepL oc/ index. php) that predicts the subcellular localization of eukaryotic proteins using Neural Networks algorithm based on the sequence information 43 . Psort (http:// psort1. hgc. jp/ form. html) prediction is based on the amino acid sequence and composition 44 . Plant PLoc (http:// www. csbio. sjtu. edu. cn/ bioinf/ plant-multi/) predicts localization based on the gene ontology information, functional domain information, and sequential evolutionary information 45 whereas YLoc (https:// abi-servi ces. infor matik. uni-tuebi ngen. de/ yloc/ webloc. cgi) is an interpretable server for predicting subcellular localization of proteins 46 . Another software called TargetP 2.0 (http:// www. cbs. dtu. dk/ servi ces/ Targe tP-2. 0/ cite. php) predicts the presence of N-terminal pre-sequences such as signal peptide, mitochondrial transit peptide, chloroplast transit peptide or thylakoid luminal transit peptide but does not provide information on the nuclear or cytoplasm targeting and represents it as other category and provides its scores www.nature.com/scientificreports/ on the likelihood 47 . WoLF PSORT (https:// wolfp sort. hgc. jp/) predicts subcellular localization sites of proteins based on their amino acid sequences and the results are based on the nearest neighbors 48 . The full-length protein sequences of MusaBAG were deposited in these servers and queried for their subcellular localization.
Cis-element in promoter sequences. Two kilobases region upstream of the transcription start site of each MusaBAG coding sequences except for MusaBAG8 (upto 300 bases) was extracted from the whole genome short-gun sequence database of NCBI and used for the analysis of cis-regulatory elements. The cis-regulatory elements within these sequences were determined using online tools such as PlantPAN 3.0 (http:// plant pan. itps. ncku. edu. tw/) and Plant CARE (http:// bioin forma tics. psb. ugent. be/ webto ols/ plant care/ html/).

Digital expression patterns and ESTs.
The RNAseq data for each of the MusaBAGs is available in the banana genome hub (https:// banana-genome-hub. south green. fr/) 34 . This data accounts for mRNA expression levels under biotic stress condition i.e., Fusarium oxysporum f. sp. cubense TR4 (tropical race 4, now Fusarium odoratissimum) infected condition in Musa acuminata DH Pahang variety (AA), and an abiotic stress condition i.e., osmotic stress in Cavendish Grand Naine variety (AAA), subject to availability of RNA sequencing data in the banana genome hub database. For both biotic and abiotic stress conditions, the average maximum RNA level data was collected for the MusaBAG genes and compared with the non-infected or control RNA levels, by plotting them in a grouped bar graph. The significance of difference between non-infected/control versus the test condition (biotic or abiotic stress) was estimated using non parametric t-test (Wilcoxon matched-pairs signed rank test) and the significance of difference between individual MusaBAG gene expression under control and test conditions was evaluated using two-way ANOVA (Sidak's multiple comparisons test). The plotting of graphs and testing of hypotheses for significance of differences was done using GraphPad Prism version 8.0.2 (Graph-Pad Software, San Diego, California USA; www. graph pad. com).
Quantitative real time RT-PCR analysis. Banana cv. Rasthali (Silk, AAB group; susceptible to both Fusarium oxysporum f. sp. cubense Foc race 1 and tropical race 4) were micro-propagated under tissue culture conditions and rooted plants were used in this experiment [49][50][51] . Rasthali plants were obtained from fields and authenticated by the Karnataka Horticulture Department, Govt. of India. The rooted plants were transferred to fresh MS liquid medium (containing 3% sucrose and pH-5.7) and treated with three different stress conditions so as to observe the differential gene expression pattern of MusaBAG genes. For abiotic stress, banana plants were treated with 300 mM sodium chloride or 5% polyethylene glycol (PEG-8000). For biotic stress, each plant was inoculated with 5 × 10 6 conidia of Fusarium oxysporum f. sp. cubense race 1 (Foc R1) 52 . Three independent plants were used for each stress treatment. Plants transferred to MS liquid medium without any treatment were used as control. After 72 h of treatment, total RNA was extracted from the root tissue as previously described 13 .
The concentration and purity of the extracted total RNA was determined using Infinite M-Plex UV-Vis spectrophotometer (Tecan, Switzerland) and integrity was checked by resolving it on a 2% agarose gel. One microgram of the total RNA was used for first strand cDNA synthesis using ProtoScript ® First Strand cDNA Synthesis Kit (New England Biolabs, MA, USA) following manufacturers' protocol. The cDNA was diluted 1:2.5 and used as a template for qRT-PCR analysis. qRT-PCR was carried out on CFX96 Touch Real-Time PCR Detection System (Bio-Rad Laboratories, USA) using 1 μl of the template DNA and SsoFast EvaGreen Supermix (Bio-Rad Laboratories, USA) following manufacturers' instructions. Primers specific for MusaBAG genes were used alongside Musa acuminata elongation factor 1α (MusaEF1α) coding sequence for gene expression normalization and subsequent quantification. C t values were retrieved using CFX Manager™ Software. The C t value for each of the MusaBAG genes was first normalized with the respective C t value of the MusaEF1α gene to obtain ΔC t . Then for the entire population under each stress condition, the ΔC t values were normalized against the control population for each of the MusaBAG genes to obtain ΔΔC t value and expressed as 2 (−ΔΔCt) . The significance of the difference between individual MusaBAG gene expression under control and stress condition was evaluated using two-way ANOVA (Dunnett's multiple comparisons test). The plotting of graphs and testing of hypotheses for significance of differences was performed using GraphPad Prism version 8.0.2 (GraphPad Software, San Diego, California USA; www. graph pad. com).
Protein-protein interactions. The Protein-protein interaction (PPI) networks for MusaBAGs were identified using the STRING search tool (version 11.0) (http:// string-db. org) with default parameters. The interactions were derived from sources such as experimentally determined interactions, curated databases, co-expression, text mining and co-occurrence of the proteins 53 . www.nature.com/scientificreports/