Two SPLICS probes for different ranges of ER-mitochondria juxtaposition
To generate a modular fluorescence based sensor of organelle proximity, we decided to capitalize on the ability of two non-fluorescent portions (the GFP moiety and the GFP β-strand 11) of the superfolder GFP variant [23,24,25] to restore a fully fluorescent GFP upon self-assembly. We reasoned that if we targeted each moiety on one of the juxtaposed membranes, the GFP fluorescence would be restored only when the two portions were close enough. We therefore placed the non-fluorescent GFP moiety on the cytosolic face of the OMM (OMM-GFP). To follow short- (≈8-10 nm) and long-range (≈40-50 nm) ER-mitochondria interactions [5], two constructs that differ for the length of the spacer placed between the ER targeting sequence and the β fragment were created by considering the distance of 0.36 nm between two alpha-carbons in a peptide chain: a ER-Short β with a 29 aa spacer and a ER-Long β with a 146 aa spacer (i.e., a maximum of ≈10.4 and 52.5 nm, respectively). These values might clearly be subjected to changes (i.e., reduction) since the amino acid sequences might not always be fully extended. We reasoned that co-expression of ER-Short β with OMM-GFP (SPLICS) and of ER-Long β with OMM-GFP (SPLICS) would result in reconstitution of GFP fluorescence (Fig. 1a). Two additional constructs, a β-tagged FP (Kate-β) and an untargeted GFP were also generated to verify the complementation of the OMM-GFP at the OMM (Fig. 1a, left) and the ER-β at the ER (Fig. 1a, middle), respectively. Expression of SPLICS and SPLICS will result in fluorescence emission specifically at the ER-mitochondria interface (Fig. 1a, right).
The different versions of the SPLICS probes were first tested for their correct localization and topology. A clear mitochondrial network appeared in HeLa cells co-expressing OMM-GFP with Kate-β (Fig. 1b, first panels); similarly, the ER network became fluorescent when ER-β and ER-β were co-expressed with a cytosolic non-fluorescent GFP (Fig. 1b, second and third panel couples). Interestingly, when SPLICS or SPLICS were expressed in HeLa (Fig. 1b, fourth and fifth panel couples) and in HEK293 cells (Supplementary Figure S1), fluorescent individual foci appeared, likely representing the juxtapositions between ER and mitochondria. At a closer inspection, the SPLICS and SPLICS signals retrieved in HeLa cells appeared different for number (see quantification in 3D rendered z-stack images, Fig. 1c).
We therefore verified whether SPLICS really recognized areas of ER-mitochondria juxtaposition. In HeLa cells expressing SPLICS, the fluorescent dots co-localized with endogenous markers of mitochondria (mtHSP60) and ER (calreticulin) (Fig. 1d). Noteworthy, the mitochondrial and ER networks were not completely engaged in the formation of the ER-mitochondria contacts reported by the SPLICS (see merge panels of Fig. 1d), suggesting that SPLICS snapshots the juxtaposition at any given moment even when transiently formed. Immuno-EM with anti-GFP antibody revealed that mitochondria and ER membranes in contact with mitochondria were preferentially marked (arrowheads in Supplementary Figure S2). Despite the non-complemented and complemented OMM-GFP cannot be distinguished by the anti-GFP antibody, it is evident that a consistent number of gold nanoparticles are distributed at the ER-mitochondria interface (inset in Supplementary Figure S2).
To gain further insights on the nature of the reconstituted SPLICs, we evaluated their stability by checking whether the number of SPLICS foci could change after 24, 48, and 72 h post transfection. Supplementary Figure S3 shows that the number of SPLICS is stable during the time course. The number of fluorescent reconstituted foci was also unaffected by the expression level of the probes (Supplementary Figure S4), suggesting that bona fide changes in the SPLICS fluorescent foci likely reflect a variation in ER-mitochondria contact sites number rather than differences in the stability/expression levels of the probes. Additionally, the overall morphology of the ER and mitochondria in cells expressing the SPLICS remained grossly unaltered (Supplementary Figure S5).
To exclude that novel and non-physiological contact sites between ER and mitochondria might be artificially induced by SPLICS expression, ER-mitochondria Ca transfer and mitochondrial Ca uptake were evaluated in HeLa cells expressing SPLICS by aequorin-based measurements. If this was the case, mitochondrial Ca transients generated by stimulation with the InsP-linked agonist histamine should be increased in SPLICS-expressing cells [17]; however, they were superimposable to those of control cells (Fig. 1f and quantification in Fig. 1g). Taken together, these results indicate that SPLICS retains the ability to self-associate only in specific areas where the two organelles are found within the distance imposed by the linker region and that it does not artificially increase tethering and Ca transfer between ER and mitochondria.
We next wished to address if SPLICS could respond to pathophysiological conditions known to affect the extent of ER-mitochondria contacts. We therefore measured short- and long-range ER-mitochondria interactions in conditions where increased ER-mitochondria coupling was reported, such as ER stress and induction of autophagy [5, 26, 27]. In HeLa cells treated with the ER stress inducer tunicamycin, or starved, the number of short-range ER-mitochondria contact sites measured by SPLICS were increased (Fig. 2a, b), in agreement with previous results [5, 26, 27]. The picture in the case of long-range ER-mitochondria interactions measured by the SPLICS was more complex: while tunicamycin significantly decreased the number of SPLICS dots, starvation did not induce any significant change (Fig. 2c, d). Altogether, these results indicate that short and long ER-mitochondria interactions are differentially modulated in response to different stimuli and suggest that the heterogeneity between the two types of contact sites reflects their involvement in specialized cellular pathways.
During starvation, inhibition of the mitochondrial fission GTPase Dynamin-related protein 1 (Drp1) results in mitochondrial elongation, increasing energy conversion and sparing mitochondria from autophagosomal degradation [28, 29]. We therefore wished to verify short and long ER-mitochondria interactions upon Drp1-driven mitochondrial shape changes. We expressed wt or a dominant-negative mutant form of Drp1 (Drp1-K38A) to induce mitochondrial fragmentation or elongation and measured the occurrence of short- and long-range ER-mitochondria juxtaposition with SPLICS. Mitochondrial fragmentation induced by wt Drp1 expression did not change the number of short-range ER-mitochondria interactions (Fig. 3a and b, compare top panels vs. middle panels), in agreement with previous data [30]. Conversely, mitochondrial elongation induced by dominant-negative Drp1 expression resulted in a significant increase in the short-range ER-mitochondria contacts detected by SPLICS (Fig. 3a, top panels vs. lower panel, and Fig. 3b). The SPLICS measured a significant reduction in the number of wide ER-mitochondria interactions in cells expressing wt Drp1 (Fig. 3c top panels vs middle panels, and Fig. 3d). Interestingly, forced mitochondrial elongation induced by Drp1-K38A expression resulted in the labeling of the whole surface of mitochondria by SPLICS fluorescence, suggesting a complete engagement of the mitochondrial network with the ER (Fig. 3c, top panels vs. bottom panels). Due to the filamentous nature of the observed SPLICS staining, the number of ER-mitochondria contacts/cell under this condition could not be reliably quantified; nevertheless, the GFP signal occupied almost completely (about 85%) the mitochondrial surface as measured by Tom20 staining (Supplementary Figure S6). Altogether, these results suggest that unopposed mitochondrial fusion is paralleled by an enhancement of the ER-mitochondria interface that may ensure the supply of lipids required for the sustained mitochondrial morphological changes [26, 28, 29, 31].
We next wished to verify if SPLICS responded to genetic modulation of the ER/mitochondria interaction. To this end, we decided to monitor SPLICS behavior following ablation of Mitofusin 2 (Mfn2), a pro-fusion mitochondria-shaping protein originally identified as a tether between the two organelles [11]. However, whether Mfn2 tethers [12, 14, 32,33,34,35] or separates [13, 22, 36,37,38] ER and mitochondria is still a matter of debate. We reasoned that SPLICS might contribute to clarify the issue by providing an estimate of the contact sites over different ranges of interaction. Acute downregulation of Mfn2 by shRNA in HeLa cells by three independent shRNA (Supplementary Figure S7) increased by ≈40% the number of SPLICS foci (Fig. 4a, b). Conversely, under the same conditions of Mfn2 downregulation the SPLICS detected a significant decrease by ≈30% in the number of ER-mitochondria interactions (Fig. 4c, d). Altogether the short- and long-range SPLICS probes not only respond to changes in known modulators of ER-mitochondria tethering, but they might also prove useful to shed light on the observed discrepancies on the role of Mfn2 at the ER-mitochondria interface.
Mfn2 and the Familial Alzheimer's Disease (FAD)-related protein Presenilin-2 (PS2) have been reported to act in a common route to tune the ER-mitochondria interface [38, 39]. We measured short-range ER-mitochondria interactions in human fibroblasts from an FAD-patient carrying the PS2-N141I mutation, previously shown to enhance ER-mitochondria coupling in an Mfn2-dependent manner [38], and a healthy sex- and age-matched control. The SPLICS signal was more than doubled in human FAD-PS2 fibroblasts compared to controls, thus confirming that endogenous FAD-PS2 increases ER-mitochondria coupling, as already reported, and proving that SPLICS represents a useful tool also in patient-derived samples (Fig. 4e, f). Lastly, we tested SPLICS with an additional well-established tethering machinery, i.e., the VAPB/PTPIP51 complex. Interestingly, we detected an increase in the SPLICS number, in agreement with previous data [7, 8, 40] (Supplementary Figure S8). The long-range interactions monitored by SPLICS were instead decreased (Supplementary Figure S8): this finding certainly deserves additional experiments but again, it might indicate that ER-mitochondria tethering can be heterogeneous and tightly modulated.
Comforted by the ability of SPLICS to provide insights under pharmacological and genetic manipulation of the ER-mitochondria interface, we decided to detect changes in ER-mitochondria tethering during Parkin-mediated mitophagy. In mammalian cells, dysfunctional mitochondria recruit the E3 ubiquitin ligase Parkin to the OMM through PINK1 kinase activity, resulting in the recruitment and activation of the autophagy machinery [41]. Parkin has been shown to act as a positive modulator of ER and mitochondria coupling in HeLa cells by organelle-targeted FPs, and in nigral neurons by transmission EM analysis [42, 43]. Nevertheless, increased ER-mitochondria juxtaposition in patient-derived fibroblasts and in PARK2 knockout MEFs [44] was also reported. Thus, the exact function of Parkin at the ER-mitochondria interface under basal conditions and upon mitophagy is unclear. We generated a bicistronic vector in which Parkin was cloned upstream of a self-cleaving viral 2A peptide (P2A) [45] followed by a plasma membrane-targeted RFP (mCherry-CAAX) to track Parkin-positive cells (Supplementary Figure S9). This construct was co-expressed along with SPLICS in HeLa cells where Parkin is absent or weakly expressed [46, 47]. Parkin overexpression increased SPLICS number (Fig. 5a, b), in agreement with our previous data [42]. Conversely, Parkin overexpression reduced the SPLICS foci (Fig. 5c, d). Treatment with CCCP reduced the number of fluorescent foci measured using both the SPLICS probes, suggesting that activation of PINK1/Parkin-mediated mitophagy loosens all types of ER-mitochondria contacts.
We finally wished to test if SPLICS can measure ER-mitochondria tethering in an in vivo setting. Imaging of subcellular structures in living animals, and even more in neuronal axons, is limited by the thickness and anatomical accessibility of tissues. In vivo detection of organelle contact sites is still a major challenge because of their dynamic nature and the lack of appropriate tools. To verify if SPLICS could overcome these hurdles, we expressed the new probes in D. rerio, specifically in Rohon-Beard (RB) sensory neurons. The correct targeting of the OMM-GFP and the ER-β constructs was first verified after mosaic expression in D. rerio embryos. The OMM-GFP signal reconstituted by complementation with a β-tagged cytosolic protein (DJ-1-β) fully overlapped with a mitochondrial targeted RFP (pTagRFP-mito). Analogously, injection of ER-β and a cytosolic GFP resulted in fluorescence emission that co-localized with an ER marker (pDsRed2-ER) (Supplementary Figure S10), thus demonstrating that the SPLICS fragments are properly expressed, targeted and self-assembled in living zebrafish embryos. To allow tissue specific as well as equimolar expression of SPLICS, we generated an expression vector where OMM-GFP and ER-β are linked by a P2A peptide (SPLICS-P2A), an approach suitable also in zebrafish [48]. SPLICS-P2A was placed under the control of a bidirectional UAS promoter together with a cytosolic DsRed (pT2-DsRed-UAS-SPLICS-P2A) to allow GAL4-driven expression of the UAS promoter (Fig. 6a). The pT2-DsRed-UAS-SPLICS-P2A vector was then microinjected in the zebrafish s1102t:GAL4 transgenic line where GAL4 expression is restricted to RB neurons (Fig. 6b), yielding simultaneous, tissue specific expression of DsRed and SPLICS (Fig. 6c). By imaging the DsRed-positive neurons, we noticed the occurrence of short-range ER mitochondria contacts in both cell body and axons (Fig. 6d-g). We retrieved several SPLICS contacts in the soma of RB neurons; their frequency was comparable to that observed in cultured cells. ER-mitochondria contact sites were also retrieved in RB axons and enriched at axonal varicosities and branching points, possibly representing axon zones with specialized functions where ER-mitochondria crosstalk is important to propagate and regulate Ca signals [49,50,51] (arrowheads in Fig. 6f). The number of short ER-mitochondria interactions was comparable in soma and axons (Fig. 6g), suggesting that these juxtapositions are regulated by similar mechanisms in the two portions of the neuron.