Supplemental description
Integration of anatomical gene expression patterns in the HSP70 phylogenetic tree
To determine the functions of rice HSP70 family members, we integrated anatomical expression patterns for 11 tissues/organs comprising 209 hybridizations (NCBI GEO series GSE21398). In addition to thefive HSP70 genes that lacked probes on the Agilent44K array, this integration included anatomical expression profiles for 27 otherHSP70 genes (Fig. S1). Those Agilent44K data were represented by log2 normalized intensity values (Table S2). Anatomical meta-profiling data based on 983 rice Affymetrix microarrays are shown in Figure S2.
Cluster I
The phylogenetic tree indicated close links among Os05g38530, Os01g62290, and Os11g47760. The first two had similar expression patterns in most analyzed samples, indicating possible functional redundancy in a developmental or anatomical role (Fig. S1). Their expression was also down-regulated under darkness (Fig. S1). The expression level of Os01g62290was much higher than Os05g38530, suggesting that the former is dominant. WhereasOs05g38530 was preferentially expressed during early ovary development, Os01g62290 showed greater expression in all other organs and tissues examined here (Fig. S1). Constitutive expression and high transcript levels were noted with Os11g47760 (Fig. S1). Neither Os12g38180nor Os03g16920was clustered with other members in this group; the latter was more highly expressed in the stems, anthers, roots, and ovaries at 7 days after pollination (DAP), as well as in maturing embryos and maturing endosperm, whilethe former was constitutively expressed, usually at a moderate level, but with stronger expression being observed in maturing embryos. Therefore, based on these patterns, we expect that Os03g16920 has a significant role in maturing embryos and the endosperm (Fig. S1).
Os03g60620 and Os03g16860were closely linked in the phylogenetic tree; the former showed a high level of constitutive expression in most organs,with one exception: expression was repressed in mature roots at 76 days after transplanting. Expression byOs03g16860 was also high and constitutive, except for leaf tissues collected during stages when the panicles (inflorescence) and pistils were developing (Fig. S1). Os03g16880 had very low expression in most samples, making it difficult to predict any anatomical function. Tandemly duplicated Os11g08460 and Os11g08440also were very closely linked in the phylogenetic tree, with dominant co-expression being observed in mature anthers. While Os11g08460 showed preferential expression only in mature anthers, Os11g08440was significantly expressed in young panicles, early developing anthers, ovaries, and the endosperm (Fig. S1). This suggested that the two genes have redundancy during mature anther formation, but that only Os11g08440 has unique functions in the developing ovary and endosperm. Because the cytosolic OsctHSP70s, Os01g49430, Os11g08445, and Os11g08470were not available on Agilent44K, we could not obtain their data.
In summary, our anatomical meta-profiling data for this cluster indicated housekeeping functions by the constitutively expressed Os11g47760 and Os03g60620,as well as roles in organ/tissue development for Os05g38530 (early ovary), Os11g08440 and Os11g08460 (anther), and Os03g16920 (maturing embryo and endosperm).
Cluster II
In this cluster, the greatest constitutive expression was found with Os02g02410 (BIP1/OsBIP3), suggesting a dominant anatomical role(Fig. S1). Os03g50250, Os05g30480, Os08g09770, and Os05g35400were preferentially expressed in anthers at the vacuolated pollen stage (Fig. S1), implying strong functional redundancy in pollen development. Among them, transcript levels for Os05g30480were approximately 2- to 4-fold higher in the ovaries, embryos, and endosperm when compared with the other three, indicating that it might have a unique and significant role in organ development. Because Os01g33360was not included on the Agilent44K, we could not obtain any anatomical meta-profiling data.
Cluster III
Os05g23740 and Os12g14070, which encode chloroplast-type HSP70 (OscpHSP70) protein, were closely linked in the phylogenetic tree and had similar expression patterns in most samples. An exception wasseen with Os12g14070, which was more highly expressed in the endosperm (Fig. S1). In most samples, expression by Os05g23740was greater than by Os12g14070,except during the late stages of ovary and endosperm development. This suggested that the former has a dominant function in vegetative tissues while the latter has a more prominent role in more mature ovaries and embryos. A mutation of ArabidopsiscpHsc70-1 and cpHsc70-2, encoding chloroplast HSP70 protein, shows defective phenotypes in plant development and thermo-tolerance in germinating seeds(Su and Li 2008), implying that their roles are similar to that of their rice homologs.
Close links on the tree and similar expression patterns were found between Os02g53420 and Os03g02260, which encode the mitochondria HSP70 (OsmtHSP70) protein. However, Os02g53420was the more dominant.Although over-expression of OsmtHSP70 (Os02g53420) suppresses PCD death in rice(Vitte et al. 2004), a loss-of function study is needed toclarify its functional roles. BecauseOs09g31486 is not part of Agilent44K, we did acquire its anatomical meta-profiling data using affymetrix microarrays and identify the similar expression pattern with that of Os03g02260 (Fig. S2), suggesting its functional redundancy with Os03g02260.
Cluster IV
Three proteins belongto this cluster: Os12g05760, Os05g51360, and Os06g10990. All were expected to be localized in the cytoplasm and ER. Of these, Os05g51360 had very low expression overall. Os12g05760 and Os06g10990 showed similarly highexpression in anthers at the young microspore stage (Fig. S1). Compared with other cluster members, expression of Os06g10990 was higher in the ovary and pistil, thus allowing us to conclude that this gene hasthe predominant role.
Cluster V
As the single member in this cluster, Os03g11910was preferentially expressed in mature anthers, ovaries at 1 DAP, and maturing embryos and endosperm (Fig. S1). This pattern was very similar to that of Os03g16920 in Cluster I. It is possible that these two genes have functional redundancybecause both encode a cytoplasmic HSP70 protein.
Cluster VI
Thefour nuclear HSP70 proteins (OsncHSP70) in this cluster are Os02g48110, Os06g46600, Os01g08560, and Os05g08840. In the phylogenetic tree, Os02g48110was positioned slightly distant from the others, and was highly expressed in early ovaries, maturing embryos, and endosperm. Of the other more closely linked genes,Os05g08840 showed constitutively high expression in most analyzed samples, while the others had very low expression (Fig. S1). Evaluating any functional redundancy among the four OsncHSP70 genes will necessitate related gene-indexed mutants or transgenic approaches.
Evaluation of functional redundancy among rice HSP70 family genes
The multiple HSP70 proteins located within each sub-cellular organelleindicated that those in the same clade were more likely to have functional redundancy. Although this can make it more difficult to identify theirroles via loss-of-function studies, mutations in genes with predicted functional redundancy often reveal visible phenotypes. For example, we previously identified defective phenotypes caused by the mutation of a member of the rice serine hydroxymethyltransferase (SHMT) family, indicating that this member has a predominant role in that family(Jung et al. 2008a; Jung et al. 2008b). Therefore a better strategy for predicting a dominant member within a gene (sub)familywould be to integrate transcriptome data into its phylogenetic tree. No phenotypes have yet been identified that are caused by the mutation of any single rice HSP70 gene except for BIP3/OsBip3 (Os02g02410), whose suppression leads to defects in the endosperm(Wakasa et al. 2011). Results from gain-of-function studies have demonstrated that BIP3/OsBip3 (Os02g02410)encodes theHSP70 protein in the ER and cytosol, thereby modulating XA21-mediated innate immunity, while the mitochondrial HSP70 gene (OsmtHSP70, Os02g53420)regulates heat- and H2O2-induced PCDin rice protoplasts(Park et al. 2010),(Qi et al. 2011).
Because functions are not yet known for the remaining genes, we integrated the Arabidopsis best-hit genes next to the GO terms in the cellular component category in order to determine the functions of the rice HSP70 genes. Of those Arabidopsisgenes, chloroplast HSP70 (cpHSP70-1, At4g24280) has a role in protein importation (Su and Li 2010) while cytoplasmic HSP70 (HSP70B, At1g16030) modulates thermo-sensitivity in response to heat or cold. Overexpression of HSC70-1 (At1g16030) inhibits plant resistance to virulent and avirulent pathogens, even if a loss of function by individual cytosolic HSC70 genes reveals no defense phenotype(Jungkunz et al. 2011). Through comparisons with those Arabidopsis best hits, we might predict that Os03g16920plays a role similar to that of At1g16030in controlling thermo-sensitivity,and we might expect Os12g14070to function like At4g24280in importing chloroplast protein. However, sequence similarity does not always determine functional similarity across species(Chikina and Troyanskaya 2011).Instead, exclusive use of genome-wide transcriptome data provides more accurate and unbiased arrangements (Chikina and Troyanskaya 2011). We believe that, by integrating the expression profiling data of the entire HSP70 family within the phylogenetic context (phylogenomic analysis), it is possible to predict more precise functions for an individual member.
Advantage of phylogenomics analysis and limitation of functional genomics analysis due to functional redundancy
Our anatomical meta-profiling data for cytosolic HSP70 family in cluster I effectively estimated different functions such as abiotic stress responses (Os05g38530, Os01g62290, and Os03g16920), housekeeping functions (Os11g47760 and Os03g60620),early ovary development (Os05g38530), anther(Os11g08440 and Os11g08460), and maturing embryo and endosperm(Os03g16920). Until now, any of these were not functionallyidentified, suggesting the possible functional redundancy among gene family members in this cluster. In real, we did not observe the phenotypic defects in a T-DNA insertion line of an OsctHSP70-1. The stimulatedexpression of Os01g62290 in the mutant was sufficient to compensate the defect of Os05g38530,whileexpression by Os11g47760 was not affected by the Os05g38530mutation, indicating that its functional redundancy was restricted to Os01g62290but not expanded to Os11g47760.
On the other hands, the greatest constitutive expression of Os02g02410 (BIP1/OsBIP3) in cluster II suggestsitsdominant role(Fig. S1).Gene knock-down (KD) analysis of BIP1/OsBIP3 in the XA21 background has shown that the resistance response to Xoois not affected (Park et al. 2010), whereas the same analysis of BIP1/OsBIP3 under the control of an endosperm-preferential promoter can alterseed phenotypes, including the intracellular structure of endosperm cells, content of seed storage proteins, starch accumulation, and grain weight, thus demonstrating a dominant role for BIP1/OsBIP3 in seed development(Wakasa et al. 2011). Moreover, expression of other paralogs (Os03g50250, Os08g09770, and Os05g35400) is up-regulated in a BIP1/OsBIP3KD mutant, indicating that the robustness feature that compensates for this gene defect stimulates the expression of other paralogs(Wakasa et al. 2011). Even though we did not assay Os05g30480 transcripts in the BIP1/OsBIP3 KD, we predict that expression of that gene was stimulated in a manner similar to that of the other paralogs. Because expression in the endosperm was greater for Os05g30480than for the others, its compensating effect would be more significant. However, that prediction requires further testing. Regardless, previous results suggested that this compensatory effectmight be temporary, because the BIP1/OsBIP3 KD line eventually manifested defective phenotypes during seed development(Wakasa et al. 2011).