ER stress signaling requires RHD3, a functionally conserved ER-shaping GTPase

The unique reticulated architecture of the endoplasmic reticulum (ER) relies on a dynamic remodeling of interconnected sheets and tubules, as well as tubule initiation, growth and fusion (Chen et al., 2013; Sparkes et al., 2011). The extent to which disruption of ER architecture affects the function of this essential organelle is largely unknown.

Optimal ER architecture depends on its lipid composition as well as ER-structuring proteins. For example, genetic disruption of the ER phospholipid biosynthetic pathway in Arabidopsis leads to a drastic modification of the ER shape with replacement of tubules by sheets (Eastmond et al., 2010). Furthermore, ER-tubule-forming proteins (reticulons and DP1/Yop1p) and dynamin-like GTPases named atlastins (ATLs) have been found to have crucial roles in ER structure. Although reticulons are required and sufficient to form an ER tubular network, most likely by stabilizing the high curvature of the tubules (Hu et al., 2008; Park et al., 2010; Voeltz et al., 2006), ATLs mediate the process of ER tubule fusion in metazoans (Hu et al., 2009; McNew et al., 2013; Orso et al., 2009). Yeast and plants do not have sequence homologs of ATLs, but Sey1p in S. cerevisiae and Root Hair Defective 3 (RHD3) in A. thaliana (Hu et al., 2011; Hu et al., 2009) have similar protein signature motifs to ATLs. Similar to ATLs and Sey1p, RHD3 has a role in ER architecture (Chen et al., 2011; Stefano et al., 2012) and facilitates membrane fusion (Zhang et al., 2013).

In animals and plants, defects in ER structure due to loss of ATLs or RHD3 have been implicated in severe growth and developmental phenotypes, including abnormal tissue growth and embryonic lethality (Audhya et al., 2007; Chen et al., 2011; Park et al., 2010), underscoring that the maintenance of an optimal architecture of the ER owing to the presence of these proteins has important implications for the life of the organism. Nevertheless, how a loss of function of ER-shaping proteins translates into growth defects at cell and tissue levels is unknown. The smaller size of the aerial and root tissues in RHD3 loss-of-function mutants compared to wild-type plants (Chen et al., 2011; Hu et al., 2003; Stefano et al., 2012; Wang et al., 1997) is associated to reduced cell elongation (Wang et al., 1997). The evidence that ER export of membrane and soluble fluorescent protein markers is not affected in RHD3 loss-of-function mutants (Chen et al., 2011) implies that the plant phenotype of rhd3 is linked to yet-to-be discovered causes that are unrelated to defects in bulk flow from the ER.

Intriguingly, loss of the reticulon Rtnl1 in Drosophila has been shown to cause elevated levels of ER stress (O'Sullivan et al., 2012), which is a condition that cells generally experience when the ability of the ER to balance protein synthesis demand and capacity is compromised. In conditions of ER stress, a largely conserved cytoprotective signaling pathway, known as the unfolded protein response (UPR), is activated (Liu and Howell, 2010; Ron and Walter, 2007). Compromised UPR leads to serious conditions and even death in animals and plants (Chen and Brandizzi, 2012; Chen and Brandizzi, 2013; Deng et al., 2011; Iwawaki et al., 2009). The evidence that loss of an ER-shaping protein in Drosophila activates the UPR suggests that the homeostasis of ER-shaping proteins influences not only ER morphology but also a crucial cellular response that is related to the function of this organelle.

In this work, we aimed to establish whether loss of proper shape affects functional aspects of organelles by assessing the ability of the ER to respond to stress in genetic backgrounds that have defects in ER network integrity. Among the Arabidopsis mutants tested, we found that loss of RHD3 negatively affects the UPR activation arm mediated by the major ER stress sensor, IRE1. Our data demonstrate a new requirement of RHD3 in cell physiology besides a known role in ER architecture (McNew et al., 2013; Stefano et al., 2012; Zhang et al., 2013), and show that ER network integrity can be correlated to a function of the ER although the phenotype is linked to specific ER architecture mutations.

To establish whether defects in the structure could affect the ER ability to evoke the UPR, we used mutant Arabidopsis backgrounds with marked ER architecture defects (supplementary material Fig. S1). Specifically, we used pah1 pah2 (herein referred as to pah1/2), a double knockout of two phosphohydrolase genes involved in the biosynthesis of phospholipids, in which the ER tubules are converted into sheets (Eastmond et al., 2010). We also used mutants with deformed ER network in which ER tubules are intertwined into large globular structures, gold36/MVP1/ERMO3 (herein referred as to gold36), which is linked to a loss of function mutant of a pseudo-lipase (Agee et al., 2010; Marti et al., 2010; Nakano et al., 2012), and g92/ERMO2 (herein referred as to g92), which is a partial loss of function of the COPII coat component Arabidopsis thaliana (At)SEC24A (Faso et al., 2009; Nakano et al., 2009). In addition, we used mutants of RHD3, specifically a null allele (rhd3-7) (herein referred as to rhd3) as well as a mutant bearing a non-silent missense mutation (gom8) (Stefano et al., 2012), which have long unbranched ER tubular structures, similar to mutations linked to ATLs and Sey1p in metazoans and yeast, respectively (McNew et al., 2013; Orso et al., 2009; Zhang et al., 2013).

To test whether these mutants have defects in the UPR, we analyzed variations in the mRNA abundance of well-established UPR molecular markers in the absence or presence of the ER-stress inducer tunicamycin (Tm) using quantitative RT-PCR (qRT-PCR) (Chen et al., 2014; Chen and Brandizzi, 2012). In marked contrast to all the other backgrounds (Fig. 1A), in the rhd3 mutants the levels of induction for UPR genes were significantly lower compared to wild type in conditions of ER stress (Fig. 1B). Nonetheless, compared to wild type, in rhd3 mutants the basal levels of UPR gene transcripts, as well as the transcript levels of genes encoding either secretory or cytosolic proteins, were similar (supplementary material Fig. S2). Taken together, these findings show that RHD3 is required for the expected increase of UPR gene transcripts in conditions of induced ER stress and that this phenotype is not a general feature of mutants with abnormal ER morphology.

Next, we analyzed the expression of UPR indicators in loss-of-function backgrounds of the other RHD3 isoforms RHD3-L1 and RHD3-L2 (Chen et al., 2011) and found that BIP3 transcript levels were unaffected in the knockouts compared to wild type (supplementary material Fig. S3). Although we cannot exclude that RHD3-like proteins and RHD3 share partially overlapping roles in the UPR, these data indicate that RHD3 loss has a predominant impact on the ability of the plant to respond to ER stress compared to the other RHD3-like genes. The pathway components of the UPR are essential: when the adaptive responses of the cell to stress are insufficient, cells enter apoptotic death. Given that in the absence of RHD3 cells fail to actuate the UPR properly, we suggest that the reported lethality of higher order mutations within the RHD3 family (Zhang et al., 2013) might be at least partially linked to the inability of cells to evoke the UPR efficiently during growth, which imposes physiological stress on the ER.

In animals and plants, IRE1, an ER-associated protein kinase and ribonuclease, functions as a major ER stress sensor and transducer (Chen and Brandizzi, 2012; Tirasophon et al., 2000; Urano et al., 2000). The Arabidopsis genome encodes two sequence homologues of IRE1, (At)IRE1A and (At)IRE1B (Chen and Brandizzi, 2012; Koizumi et al., 2001; Noh et al., 2002). Unlike RHD3 loss, AtIRE1 loss does not affect ER structure in optimal conditions of growth and under ER stress induction (supplementary material Fig. S1B). To understand how RHD3 could affect the UPR, we tested whether RHD3 loss could compromise AtIRE1 signaling in ER stress using a genetic approach. We generated a triple mutant, here named ire1/rhd3, by crossing an atire1a atire1b double mutant (Chen et al., 2014; Chen and Brandizzi, 2012) with rhd3, and tested its sensitivity to a Tm concentration in the medium that inhibits growth of ire1 for comparison to wild-type plants, and rhd3 and ire1 mutants. We found that rhd3 mutants grew similar to wild type, whereas the triple mutant showed sensitivity to Tm similar to the ire1 mutant (Fig. 2A). Next, we measured the induction levels of UPR genes in conditions of ER stress in wild-type plants, rhd3, ire1 and ire1/rhd3 mutants (Fig. 2B). As expected, we found a reduction of UPR transcript genes in ire1 and in rhd3; notably, however, in ire1/rhd3 mutants the reduction of transcript levels of UPR genes was similar to ire1 (Fig. 2B). The oversensitive phenotype of ire1 to Tm treatment compared to wild-type plants and rhd3 mutants highlights that, in contrast to RHD3, AtIRE1 is essential to respond to ER stress. Taken together with the evidence that the induction of UPR genes in the ire1/rhd3 mutant exposed to Tm is comparable to ire1 and significantly different from rhd3, the results also support that RHD3 acts upstream AtIRE1 in ER stress responses.

In all eukaryotes, the IRE1 RNase domain initiates splicing of mRNAs encoding bZIP transcription factors, namely XBP1 in mammalian cells, HAC1 in yeast and bZIP60 in plants (Cox et al., 1997; Deng et al., 2011; Kawahara et al., 1997; Moreno et al., 2012; Plongthongkum et al., 2007; Shen et al., 2001; Sidrauski and Walter, 1997). Similar to the effect of RHD3 deletion on the UPR, loss of bZIP60 splicing interferes with UPR activation (Moreno et al., 2012). Although bZIP60 is downstream of AtIRE1 in the UPR signaling, a bZIP60 loss-of-function mutant does not show the ER stress oversensitive phenotype that is typical of ire1 (Chen and Brandizzi, 2012; Deng et al., 2013) (Fig. 3A), indicating that in addition to the splicing of bZIP60 mRNA, AtIRE1 has other roles that are essential to cope with ER stress. Therefore, the lack of an ER-stress-oversensitive phenotype of RHD3 and bZIP60 mutants (Fig. 3; Deng et al., 2013), and the evidence of a genetic interaction between RHD3 and AtIRE1 for UPR induction (Fig. 2), suggests that RHD3 loss could affect the AtIRE1-splicing of bZIP60 mRNA. To test this, we measured the abundance of spliced bZIP60 (sBZIP60) mRNA as a readout of AtIRE1 activity in the UPR for wild-type plants and rhd3 mutants in normal conditions of growth as well as in the presence of Tm. We found that in the rhd3 alleles, sBZIP60 transcript was detectable but at significantly reduced levels compared to in wild type (Fig. 3B). These data support that the defects in UPR gene induction in RHD3 loss-of-function backgrounds during ER stress is linked to interference with the AtIRE1-mediated splicing of bZIP60 mRNA. Upon UPR activation in yeast and in mammalian cells, IRE1, which is distributed over the ER network in physiological conditions of growth, localizes into dynamic clusters in the ER, which are supposed to function as specialized molecular microenvironments for IRE1 signaling in the UPR (Li et al., 2010). Whether AtIRE1 also undergoes dynamic clustering in the UPR is unknown but it is likely given the conservation of IRE1 (Chen and Brandizzi, 2013). We speculate that the disruption of optimal membrane rearrangements due to loss of RHD3 might reduce the ability of AtIRE1 either to traffic through the ER membrane to form signaling clusters or interact efficiently with its bZIP60 mRNA substrate.

Mutants with defective expression of either RHD3 or AtIRE1 show an obvious phenotype in the elongation of primary root (Chen et al., 2011; Chen and Brandizzi, 2012; Hu et al., 2003; Stefano et al., 2012; Wang et al., 1997) (Fig. 4A), but the underlying mechanisms are unknown. In order to investigate whether RHD3 and AtIRE1 interact in the control of organ growth, we analyzed the roots of the ire1/rhd3 mutant for comparison to wild-type plants, and rhd3 and ire1 mutants. The length of the primary root was shorter in ire1/rhd3 mutants than in the respective rhd3 and ire1 mutants, supporting an additive interaction in root growth (Fig. 4). Further support to this observation was provided by quantitative semi-automated cell segmentation analyses (French et al., 2012) specifically designed for the unbiased identification of the transition zone, which marks the boundaries of the division zone and the elongation zone (Fig. 4C,D). Consistent with the length measurements of the primary root, in rhd3 and ire1 mutants the division zone was shorter than in wild-type plants and even shorter in the ire1/rhd3 mutant (Fig. 4C), further demonstrating that the triple mutant has additive phenotypic defects of the ire1 and rhd3 mutations. Taken together, these data indicate that even if the function of AtIRE1 in UPR signaling largely depends on the cellular availability of RHD3, AtIRE1 and RHD3 have independent roles in physiological organ growth. It cannot be excluded, however, that RHD3 and AtIRE1 might share the control of some crucial components in pathways necessary for root growth. During physiological growth, which requires enhanced production of the building blocks of the cell and cell wall, loss of AtIRE1 is likely to compromise the protein-synthesizing capacity of the ER. By contrast, loss of RHD3 function is known to interfere with the subcellular positioning and movement of the Golgi complex (Chen et al., 2011; Stefano et al., 2014; Stefano et al., 2012). Therefore, AtIRE1 and RHD3 might be controlling convergent pathways in organ growth through functionally parallel routes, whereas loss of AtIRE1 might disrupt the production of proteins necessary for organ growth, and disruption of RHD3 might lead to aberrant subcellular distribution of such proteins due to a spatial disorganization of the crucial sorting organelles, such as the Golgi.

Tm (Sigma, T7765; dissolved in DMSO) was directly added in the medium as follows: 0.5 µg/ml for UPR induction analyses and 0.05 µg/ml for testing sensitivity to ER stress. For the mock control the Tm volume was replaced by the same volume of DMSO (Chen and Brandizzi, 2012).

Total RNA extraction and qRT-PCR were performed in triplicate, as described previously (Chen and Brandizzi, 2012). Relative expression levels were normalized to that of UBQ10. Values are representative averages from three technical replicates. Similar patterns of expression were observed in three independent biological replicates. Primers and AGI numbers are provided in supplementary material Table S1. Col-0 was used as wild-type reference genotype.

To establish the position of the transition zone, the Cell-o-Tape macro (open source ImageJ/Fiji) was used (French et al., 2012). The CellSeT software was used to build the root segmentation based on the Propidium Iodide staining (1 µg/ml), as described previously (Pound et al., 2012).

Statistical analyses included either two-tailed Student's t-test, assuming equal variance, or unpaired Student's t-test (Fig. 4).