Interactions of cytosolic tails in the Jen1 carboxylate transporter are critical for trafficking and transport activity Free

Eukaryotic plasma membrane (PM) transporters play essential roles in cell nutrition, signalling and responses to stress conditions and drugs. Consequently, transporter malfunction has an impact on many aspects of human cell biology and leads to several pathologies, including neurological and cardiovascular disorders, as well as diabetes and cancer (Bowton et al., 2014; Gonzalez-Menendez et al., 2017; Morgan et al., 2011; Sakrikar et al., 2012; Scheffner and Kumar, 2014). Given their importance in sensing the environment and maintaining cellular homeostasis, transporter function depends on complex and fine regulatory mechanisms. However, our knowledge of PM transporter regulation is not yet complete, especially regarding roles of transporter cytoplasmic termini.

Endocytic internalization of PM transporters in response to physiological or stress signals has been extensively studied in the model fungi Saccharomyces cerevisiae and Aspergillus nidulans. (for recent reviews, see Barata-Antunes et al., 2021; Diallinas and Martzoukou, 2019; Kahlhofer et al., 2020). Endocytic internalization of fungal PM transporters requires ubiquitylation at their C- or N-terminal cytosolic regions by HECT-type E3 ubiquitin (Ub) ligases (including Rsp5 in S. cerevisiae and HulA in A. nidulans), which are recruited by adaptor proteins named α-arrestins (Becuwe and Léon, 2014; Becuwe et al., 2012a; Gournas et al., 2010; Karachaliou et al., 2013; Lin et al., 2008; Merhi and André, 2012; Nikko and Pelham, 2009; Talaia et al., 2017). In S. cerevisiae, 14 α-arrestins have been identified, named the Arts (Art1–Art10), Buls (Bul1–Bul3) and Spo23, which all possess PY motif(s) that can interact with WW domains of the Rsp5 Ub ligase, mediating membrane protein turnover (Becuwe et al., 2012a; Lin et al., 2008; Nikko and Pelham, 2009; Novoselova et al., 2012; O'Donnell et al., 2010; Yashiroda et al., 1996). A. nidulans possesses 10 α-arrestins, including ArtA and PalF, which control transporter downregulation and pH sensing, respectively (Barata-Antunes et al., 2021; Karachaliou et al., 2013), whereas in mammals six α-arrestins have been identified, named ARRDC proteins (Alvarez, 2008; Rauch and Martin-Serrano, 2011), but much less is known regarding their role, especially on transporter cellular expression.

Previous studies in S. cerevisiae have proposed that α-arrestins recognize the exposed cytoplasmic N- and C-termini of nutrient transporters (Crapeau et al., 2014; Ghaddar et al., 2014; Gournas et al., 2017; Guiney et al., 2016; Keener and Babst, 2013; Lin et al., 2008; Marchal et al., 1998, 2000; Merhi et al., 2011). These segments contain specific α-arrestin-interacting motifs and Ub target sites. For example, the N-terminus of the general amino acid transporter Gap1 contains a potential Bul1 and Bul2 (Bul1/2) α-arrestin-interacting motif and two Ub target sites that are required for nitrogen-elicited endocytosis of Gap1 (Ghaddar et al., 2014; Merhi et al., 2011). Additionally, under stress conditions, Bul1/2, in combination with Art3 and Art6, promotes Gap1 ubiquitylation and downregulation via the Gap1 C-terminus (Crapeau et al., 2014). In these cases, it is thought that conformational changes during substrate transport makes the N-terminus more accessible to α-arrestins (Ghaddar et al., 2014; Gournas et al., 2017).

In the filamentous fungus A. nidulans, a C-terminus region of the uric acid transporter UapA is essential for ArtA-mediated ubiquitylation, endocytosis and vacuolar degradation in response to ammonium or excess of substrate (Gournas et al., 2010; Karachaliou et al., 2013). Also, in A. nidulans, the FurE nucleobase-allantoin transporter has been shown to possess elements in its N- and C-terminus that are critical for endocytosis and, surprisingly, for substrate specificity. In this case, the authors provided evidence that the N- and the C-terminus interact physically and promote proper transporter function and turnover (Papadaki et al., 2017, 2019). Whether long-range regulatory effects of cytosolic N- and C-termini extend to transporters other than Fur-like proteins or other members of the amino acid–polyamine–organocation (APC) superfamily (Mikros and Diallinas, 2019) remains to be formally shown. These findings raise new questions regarding conserved intrinsic roles of the cytoplasmic termini of transporters, including APC and major facilitator superfamily (MFS) members.

Specifically, and importantly, do transporter termini contribute to transporter function and regulation in ways that are distinct from their known roles in transporter endocytosis and downregulation?

Here, we address roles of the cytosolic terminal regions of Jen1, a yeast monocarboxylate/H⁺ symporter (lactate and pyruvate being its major substrates) that represents the ubiquitous and largest transporter family, the so-called MFS. By functionally analysing N- and C-terminally truncated versions of Jen1, we provide compelling evidence that both cytosolic termini of Jen1 are crucial for the biogenesis, transport kinetics and turnover of Jen1. Using quantitative bi-fluorescence complementation (BiFC) assays, we present further evidence that the two Jen1 termini interact dynamically in a transport-activity-dependent manner, which ultimately regulates Jen1 cell surface expression and activity. Our findings support the idea that cytosolic termini of MFS transporters provide important multi-functional roles, which should be examined as a general emerging concept in transporter regulation.

Jen1 has been used extensively as a model cargo to dissect mechanisms of regulated transporter internalization. Jen1 ubiquitylation, endocytosis and vacuolar degradation are regulated by two α-arrestins (Rod1 and Bul1), in response to distinct stimuli (Becuwe et al., 2012b; Fujita et al., 2018; Hovsepian et al., 2018; Talaia et al., 2017). Recently, a C-terminal region of Jen1 has been reported to be involved in Rod1-mediated endocytosis of the transporter, triggered by glucose (Fujita et al., 2018). However, roles of the Jen1 N- and C-termini in PM targeting, transport activity and endocytic downregulation are incompletely understood.

To further elucidate roles of the cytosolic termini of Jen1 in its regulation, we employed specific N- or C-terminus truncations of Jen1. Prior to these constructions, it was essential to define the limits of the N- and C-terminus of Jen1 based on available structural information. At first, the selection of the number of residues corresponding to the cytosolic N- and C-terminus portions of Jen1, which lacks an experimentally defined structure, was based on standard topology predictions and homology threading modelling (detailed in Table S1). These predictions were used to construct five separate truncated versions of Jen1 expressed from a low-copy centromeric plasmid in a S. cerevisiae jen1Δ strain. All Jen1 versions were expressed as C-terminal GFP fusion proteins (see Materials and Methods). Two truncations lacked 133 or 94 N-terminal amino acids (Jen1ΔNT133 and Jen1ΔNT94), two lacked the 64 or 33 C-terminal amino acids (Jen1ΔCT64 and Jen1ΔCT33), and the fifth lacked both the 94 N-terminal and the 33 C-terminal amino acids (Jen1ΔNT94ΔCT33) (Fig. 1A,B). We also employed the AlphaFold tool for 3D protein structure prediction (https://alphafold.ebi.ac.uk/) to visualize the predicted structure of Jen1 including the N- and C-termini. Fig. 1C shows the predicted structure of Jen1, highlighting the truncated regions in the cytosolic termini. This approach indicated that the large terminal truncations (Jen1ΔNT133 and Jen1ΔCT62) remove helical segments that could be extensions of transmembrane segment (TMS)1 or TMS12, respectively, necessary for Jen1 function. Consistent with this notion, subsequent analyses showed that the truncated Jen1ΔNT133 and Jen1ΔCT62 proteins are non-functional (Fig. S1; Fig. 2).

We analysed the growth and ¹⁴C-lactic acid transport activity of strains expressing either full-length or truncated Jen1 forms (Fig. 2A,B). The two larger terminal deletions of Jen1 (Jen1ΔNT133 and Jen1ΔCT62) abolished Jen1-dependent growth on lactate, as the sole carbon source (Fig. 2A) and led to loss of Jen1 transport activity (Fig. 2B), resembling the negative control lacking Jen1 (jen1Δ cells carrying the empty p416-GPD plasmid). In contrast, the three shorter truncated versions of Jen1 (Jen1ΔNT94, Jen1ΔCT33 and Jen1ΔNT94ΔCT33) were able to confer growth on lactate and ¹⁴C-lactic acid transport activity similar to wild-type Jen1 control cells (Fig. 2A,B). The recorded transport capacities in the mutants were confirmed by measurements of the external medium alkalinisation that occurs upon Jen1-mediated lactate uptake (Talaia et al., 2017), as only full-length Jen1, Jen1ΔNT94, Jen1ΔCT33 and Jen1ΔNT94ΔCT33 led to an increase in the pH of the medium (Fig. S1A).

The subcellular localization of the truncated Jen1 proteins (Fig. 2C; Fig. S1B) was monitored upon a switch to lactate, which promotes Jen1 localization to the PM (Lac 4 h), and after sequential addition of glucose (Lac 4 h+Gluc 2 h), which leads to endocytic turnover. As expected, stable localization of wild-type Jen1 to the PM was observed upon growth in lactate (Lac 4 h), whereas endocytosis and sorting to vacuoles for degradation was observed 2 h after the addition of glucose (Lac 4 h+Gluc 2 h). The Jen1ΔNT133 and Jen1ΔCT62 proteins bearing large truncations showed a significant retention in the ER, as revealed by fluorescent labelling of nuclear ER (nER) rings as well as the discontinuous fluorescence at the cell periphery that is typical of cortical ER (cER) in yeast (Fig. 2C). Co-staining with CMAC indicated that the intracellular rings do not correspond to vacuoles (Fig. S1B). In the case of Jen1ΔCT62, retention in the ER was confirmed by expression in a S. cerevisiae strain lacking six ER-PM tethering proteins. In this strain (named Δtether), contacts between the cER and the PM are significantly decreased, allowing ER-resident proteins to be unambiguously distinguished from those localized to the PM (Manford et al., 2012). In this strain, an ER-resident marker (DsRed-HDEL), clearly colocalized with Jen1ΔCT62–GFP, but not with the functional Jen1ΔCT33–GFP protein (Fig. S2A). Jen1ΔNT133 and Jen1ΔCT62 might be misfolded and retained in the ER, which could explain, at least in part, why these proteins are non-functional. The cytosolic termini of some transporters contain ER export signals, such as di-acidic, hydrophobic and aromatic variable short sequences (for a review, see Mikros and Diallinas, 2019). Interestingly, the longer N- and C-termini of Jen1 harbour similar motifs that might be necessary for COPII-mediated ER exit or for proper folding (¹¹²⁶NPIPE¹³³ and ⁵⁷⁷EYE⁵⁷⁹, respectively; see Fig. S2B). These regions might be worthy of future study by mutational analysis.

The functional truncated Jen1 proteins also displayed interesting localization patterns. Jen1ΔNT94–GFP, although fully functional as demonstrated by growth tests and lactate transport assays (Fig. 2A,B), displayed a clear decrease in PM localization and a corresponding increase in vacuole localization, even in the absence of glucose (Lac 4 h) (Fig. 2C; Fig. S1B). This suggests that Jen1ΔNT94, even upon constitutive expression, undergoes very rapid constitutive internalization from the PM and sorting to the vacuole for degradation. In contrast, Jen1ΔCT33–GFP localized to the PM, with no indication of ER retention after 4 h in lactate (Fig. 2C). Notably, Jen1ΔCT33–GFP displayed strong fluorescence at the PM, consistently stronger than wild-type Jen1-GFP, suggesting that this truncation stabilizes the transporter. Jen1ΔCT33–GFP also showed reduced vacuole localization upon glucose addition when compared to wild-type Jen1–GFP (Fig. 2C; Fig. S1B). These initial results suggest that loss of N-terminal domain might expose residues in the C-terminal domain involved in α-arrestin-dependent endocytosis. In support of this model, the doubly truncated Jen1ΔNT94ΔCT33 version stably localized to the PM, and it was more resistant to glucose-induced endocytosis as compared to both full-length Jen1 and Jen1ΔCT33 (Fig. 2C; Fig. S1B). These data were supported by the quantification of the ratio of the fluorescence at the plasma membrane over the total fluorescence at the Lac 4 h+Gluc 2 h time point (Fig. 2D). The ratio obtained for Jen1ΔNT94ΔCT33 was the highest (0.51), followed by that determined for Jen1ΔCT33 (0.40). As expected, the ratio for the Jen1ΔNT94 (0.29) was similar to that of wild-type Jen1 (0.26). This suggests that loss of the C-terminal 33 residues of Jen1 is not only epistatic to the instability conferred by deleting the N-terminal 94 residues, but also points to the idea that the two termini of Jen1 might interact functionally.

To better understand the effect of deleting the terminal segments on Jen1 function, we investigated whether functional truncations (Jen1ΔNT94, Jen1ΔCT33 and Jen1ΔNT94ΔCT33) affected the transport kinetics of lactate via direct measurements of lactic acid transport (Fig. 3). All Jen1 truncations tested, including Jen1ΔNT94, which proved to be unstable at the PM, displayed higher substrate affinities (lower K_m) compared to wild-type Jen1. Notably, deleting the C-terminal region in Jen1ΔCT33 resulted in a 10-fold increase in substrate affinity. The doubly truncated Jen1ΔNT94ΔCT33 version and Jen1ΔNT94 also had a 2.5- to 3-fold increased affinity for lactate. Thus, transport kinetics of Jen1 truncations reveal that specific segments of the N- and C-termini are critical for substrate binding and transport dynamics, in addition to their role in PM sorting, PM stability and regulated endocytosis. These results unmask previously unappreciated functions of the N- and C-termini.

Glucose-induced Jen1 internalization has been reported to require both Rod1 and Bul1/2 arrestins (Becuwe et al., 2012b; Hovsepian et al., 2018), suggesting that multiple α-arrestins may recruit the ubiquitylation machinery necessary for Jen1 endocytosis. A glucose-responsive degron recognized by Rod1 has been recently identified in the C-terminus of Jen1 (Fujita et al., 2018). However, no binding motif has been identified for the Bul1 and Bul2 arrestin paralogs. Moreover, it is not known whether Rod1 interacts with other regions in Jen1, including the N-terminus.

We investigated the localization and relative PM levels of the functional Jen1 proteins in wild-type cells (ROD1⁺BUL1/2⁺) and in strains lacking the Rod1 and Bul1/2 arrestin proteins (rod1Δ, bul1Δbul2Δ and rod1Δbul1Δbul2Δ) (Fig. 4; Fig. S3). In these assays, GFP-tagged Jen1 versions were expressed from the GAL1 promoter (Gal 5 h), and then glucose was added for 2 or 4 h to repress expression and trigger Jen1 internalization and subsequent degradation in the vacuole (see Materials and Methods). At the times indicated, cells were collected, visualized by fluorescence microscopy (Fig. 4A–D) and relative levels at the PM (as a fraction of the total GFP signal) were determined (Fig. 4E).

Results obtained in the wild-type background (ROD1⁺ BUL1/2⁺) showed that full-length Jen1 is internalized and targeted to the vacuole after glucose addition (Fig. 4A). This was confirmed by Jen1 colocalization with CMAC, a known vacuolar marker (Fig. S3A). Internalization of full-length Jen1 was delayed in a strain lacking Rod1, but it was localized to the vacuole in bul1Δbul2Δ cells following addition of glucose, similar to what is seen in wild-type cells (Fig. 4B,C,E; see also Fig. S3B,C). Internalization of full-length Jen1 was further impaired in rod1Δbul1Δbul2Δ triple deletion mutant cells, as compared to rod1Δ single mutant cells (Fig. 4D,E; see also Fig. S3D). This was further supported by western blot analyses, which revealed that, upon 4 h of glucose, steady-state levels of Jen1 remained relatively high when Rod1 was absent (i.e. in rod1Δ and rod1Δbul1Δbul2Δ cells; Fig. S4B,D,E). Notice also that in rod1Δ cells, there is a relative decrease in Jen1 levels between 2 and 4 h of glucose addition (Fig. S4B,E), suggesting that the absence of Rod1 does not fully rescue Jen1 from turnover in agreement with localization results shown in Fig. 4A–E. Overall, the results suggest that Rod1 is the principal arrestin that mediates glucose-induced Jen1 endocytosis, although it does not totally rescue Jen1 from sorting to the vacuole, whereas Bul1/2 do not seem to be required for endocytosis, under the experimental conditions used. Interestingly, however, glucose-triggered Jen1 endocytosis was significantly blocked in the triple mutant cells (Fig. 4E), revealing that the Bul1/2 proteins may have a compensatory role upon loss of Rod1, in line with previous reports (Becuwe et al., 2012b; Hovsepian et al., 2018). The additive effect of deleting Rod1 and Bul1/2 might further suggest that these arrestins compete for a common cytoplasmic target in Jen1 and that Rod1 has a higher affinity for this target compared to Bul1/2.

We carried out similar experiments using the functional truncated versions of Jen1. Quantitative fluorescence microscopy demonstrated that Jen1ΔNT94 underwent glucose-induced internalization and delivery to vacuoles in wild-type cells (ROD1⁺ BUL1/2⁺) (Fig. 4A,E; Fig. S3). In contrast, glucose-induced internalization of Jen1ΔNT94 was significantly impaired in rod1Δ mutant cells (Fig. 4B,E; Fig. S3), consistent with previous work suggesting that Rod1 binds an endocytic signal in the Jen1 C-terminus (Fujita et al., 2018). Glucose-induced Jen1ΔNT94 internalization was also impaired in bul1Δbul2Δ double mutant cells and rod1Δbul1Δbul2Δ triple mutant cells, although to a lesser extent than in rod1Δ mutant cells (Fig. 4E). Western blot analyses showed that the deletion of the first 94 residues led to a very unstable Jen1 protein in all genetic backgrounds, especially in the presence of glucose (Fig. S4A–E). Notice, for example, that although we were able to detect PM-associated fluorescence of Jen1ΔNT94 in rod1Δbul1Δbul2Δ (Fig. 4E), we could not detect the intact protein in the relative western experiment (Fig. S5D). The reason for this is not clear, but it might be related to the fact that cells observed under the microscope and those used to extract proteins are necessarily treated differently. However, we could still detect stabilization of the Jen1ΔNT94 protein in rod1Δ, in line with the idea that Rod1 exerts its activity via the C-terminus of Jen1.

As expected, glucose-induced internalization of Jen1ΔCT33 was impaired as compared to full-length Jen1 in wild-type cells (Fig. 4A,E). Unexpectedly, however, glucose-induced Jen1ΔCT33 internalization was further impaired in rod1Δ mutant cells as compared to wild-type control cells (Fig. 4E). Glucose-induced Jen1ΔCT33 internalization was also impaired in bul1Δbul2Δ and rod1Δbul1Δbul2Δ mutant cells as compared to wild-type cells (Fig. 4E). The Jen1ΔNT94ΔCT33 protein displayed significant defects in glucose-induced internalization in wild-type, rod1Δ, bul1Δbul2Δ, and rod1Δbul1Δbul2Δ cells alike (Fig. 4E).

Western blot analyses confirmed that Jen1ΔCT33 and Jen1ΔNT94ΔCT33 were very stable Jen1 versions. Notice also that the doubly truncated version Jen1ΔNT94ΔCT33 was more stable than Jen1ΔCT33 in the wild-type or rod1Δ backgrounds, but not in Δbul1Δbul2Δ or rod1Δbul1Δbul2Δ, further revealing that Bul1/2 act via the N-terminus of Jen1 (Fig. S4B,E). Taken together, these results demonstrate that the Jen1 N- and C-termini are both involved in Jen1 downregulation and that they undergo complex interactions with both Rod1 and Bul1/2 arrestins.

Overall, the findings thus far reveal that the Jen1 N- and C-termini function in Jen1 biogenesis in the ER, and its substrate-binding affinity and transport at the PM, as well as in regulated endocytosis. The data also indicate that both of the Jen1 termini undergo complex interactions with α-arrestin family members. Next, we addressed whether the N- and C-termini are sufficient for endocytic uptake. We also sought to monitor dynamic interactions between the N- and C-termini during substrate binding and transport activity.

To further investigate the functional role of Jen1 termini, we undertook construction and functional analysis of chimeric transporters based on the UapA transporter from A. nidulans. These chimeras were designed to carry the cytosolic termini of Jen1 fused with UapA. UapA is an extensively studied uric acid-xanthine/H⁺ symporter (for a review, see Diallinas, 2016), which is regulated by ammonium or substrate-elicited endocytosis in A. nidulans. However, upon functional expression in S. cerevisiae, it does not respond to endocytosis and, instead, stably remains at the PM (Leung et al., 2010). Thus, UapA provides an appropriate molecular marker for investigating, via domain swap experiments, any potentially context-independent functional role of cis-acting elements present in the Jen1 N- and C-terminal regions. Chimeras of UapA and Jen1 (denoted UapA/Jen1) were constructed as illustrated in Fig. 5A. Briefly, the intact UapA coding sequence was fused with amino acid segments 1–94 (i.e. NT94) and/or 584–616 (i.e. CT33) of Jen1 termini, resulting in the chimeric transporters named UapA/Jen1NT94, UapA/Jen1CT33 and UapA/Jen1NT94-CT33. For details of strains see Materials and Methods. We analysed these chimeras by performing uptake transport assays (Fig. 5B), epifluorescence microscopy and western blotting, in different S. cerevisiae strains (Figs 6 and 7).

UapA, UapA/Jen1NT94, UapA/Jen1CT33 and UapA/Jen1NT94-CT33 conferred saturable ³H-xanthine import (Fig. 5B), which strongly suggested that UapA and the chimeras are all translocated to the PM and are transport active. In fact, UapA and the single chimeras UapA/Jen1NT94 and UapA/Jen1CT33 showed very similar K_m and V_max values. The K_m values measured were also similar to the native K_m of UapA measured in A. nidulans (7.0±2.0 μM, mean±s.d.;Alguel et al., 2016). The double chimera UapA/Jen1NT94-CT33 showed a reduced rate of transport, as the relevant K_m and V_max values were increased and reduced, respectively, an indication that this chimera might be partially misfolded when expressed in yeast (Fig. 5B).

The subcellular localization of wild-type UapA and the three chimeras (UapA/Jen1NT94, UapA/Jen1CT33 and UapA/Jen1NT94-CT33) was followed in a standard wild-type S. cerevisiae carrying a jen1 deletion (jen1Δ). UapA and the single chimeras UapA/Jen1NT94 and UapA/Jen1CT33 showed significant PM localization, concomitant with partial retention in the ER, which was most evident in the case of UapA/Jen1NT94 (Fig. 6A, upper row). The double chimera UapA/Jen1NT94-CT33 seemed to be massively retained in ER-like cytosolic structures. The ability of UapA/Jen1NT94 and UapA/Jen1CT33 chimeras to translocate to the PM was also confirmed in the strain lacking the ER–PM tethering proteins (Fig. S5), where the PM-resident proteins can be unambiguously distinguished from those localized to the ER (Manford et al., 2012). Overall, these findings were in accordance with uptake assays, showing that the single UapA/Jen1 chimeras embedded in the PM are also transport competent (Fig. 5B).

After having established that UapA and UapA/Jen1 single chimeras are translocated to the yeast PM and function with proper kinetics, we followed their response to glucose-triggered endocytosis (Fig. 6A, middle and lower rows). In the presence of glucose, the localization profile of UapA and UapA/Jen1NT94 was similar to that obtained without glucose, suggesting that the relevant proteins are insensitive to glucose-triggered endocytosis. Notably, however, the presence of glucose elicited detectable endocytic turnover of UapA/Jen1CT33 (Fig. 6A,B). This was evident after 2 h in the presence of glucose, when fluorescent vacuoles co-stained with CMAC appeared and the PM-associated fluorescence signal was reduced. Thus, the CT33 region of Jen1 seems to confer glucose-dependent turnover of UapA.

We also analysed the localization of UapA and UapA/Jen1 chimeras in rod1Δ and bul1Δbul2Δ strains, in the absence or presence of glucose. In both tested genetic backgrounds wild-type UapA was translocated to the PM with some evidence of moderate ER retention, irrespectively of the presence of glucose, as probably expected (Fig. 7A,B). UapA/Jen1CT33, the chimera shown to respond to endocytosis by glucose in a wild-type background (ROD1⁺BUL1/2⁺) (see Fig. 6), when expressed in rod1Δ strain, was localized to the PM, irrespectively of the presence or absence of glucose (Fig. 7A,B) and did not colocalize with vacuoles. In contrast, UapA/Jen1CT33 PM localization in bul1Δbul2Δ is affected by the presence of glucose similar to the wild-type background. This suggested that glucose-triggered endocytic turnover of this chimera is Rod1 dependent and Bul1/2 independent. The localization of UapA/Jen1NT94 was not affected in rod1Δ and bul1Δbul2Δ strains, or by the absence or presence of glucose (data not shown).

To better understand the response of UapA/Jen1CT33 to glucose-elicited, and apparently Rod1-dependent, endocytosis, we measured the steady-state protein levels of wild-type UapA and UapA/Jen1CT33 by western blotting. As shown in Fig. 7C,D, UapA/Jen1CT33 protein levels were significantly reduced in a Rod1-dependent manner in the presence of glucose, unlike those of wild-type UapA. These results suggest that the C-terminus of Jen1 is sufficient to promote glucose-elicited, Rod1-dependent turnover of UapA, in a context-independent manner. In addition, our findings confirm that Bul1/2 plays no role in glucose-elicited endocytosis of UapA/Jen1CT33.

Our results strongly suggest that N- and C-termini of Jen1 contain elements critical for biogenesis, function and turnover of Jen1. Most notably, the effects of Jen1 cytosolic termini on Jen1 functional expression proved additive and complex. This was highlighted by the significant stabilization of Jen1 when truncated at both termini in comparison to what was found for the singly truncated versions. These findings also suggested that the termini of Jen1 might interact with each other during the conformational changes accompanying transport activity, which is also associated with endocytic turnover. To further investigate this issue, we used a bimolecular fluorescence (BiFC) assay, based on reconstitution of YFP florescence when the two parts of the split YFP epitope are fused in the two termini of Jen1 (YFPn–Jen1–YFPc). Given that reconstitution of YFP might in principle also occur if Jen1 dimerizes, we also constructed a strain co-expressing the two parts of the split YFP epitope fused in distinct Jen1 molecules (i.e. Jen1–YFPn or Jen1–YFPc). These strains and relative controls (i.e. strains expressing Jen1–YFPn or Jen1–YFPc or co-expressing both) were used to investigate whether YFP is reconstituted in cis via interaction of the Jen1 termini, or/and in trans via dimerization of Jen1 molecules (for details of constructs see Materials and Methods). Fig. 8A shows that strong, PM-associated reconstitution of YFP fluorescence occurs solely when the split epitope parts are fused with the termini of Jen1, whereas no fluorescent signal was obtained when these are fused in different Jen1 molecules. This result not only strongly suggests that the two termini of Jen1 come in close contact when attached in the same Jen1 molecule, but also points against tight dimerization of distinct Jen1 molecules, at least under the conditions tested. To further address the mechanism by which the two termini of Jen1 come into contact, we repeated our assay in the presence of substrate. Fig. 8B (left panel) shows that the presence of lactic acid reduced significantly the YFP fluorescence signal over time, but had no effect on the strength of fluorescent signal coming from Jen1–GFP (Fig. 8B, right panel). Noticeably, the relative strength of reconstituted YFP fluorescence in the presence of lactic acid fluctuated over time. These findings show that stable reconstitution of YFP is transport activity dependent, further suggesting that conformational movements accompanying translocation of the substrate might affect the positioning of the two termini in the outward- and inward-facing topologies of Jen1 (see Fig. 8C). The transport-dependent interaction of N- and C-termini of Jen1 is very similar to what has been observed in FurE, which is a member of the evolutionarily, structurally and functionally distinct NCS1/APC superfamily (Papadaki et al., 2019).

In the present work, we investigated possible functional roles of the terminal cytosolic segments of the Jen1 monocarboxylate transporter in S. cerevisiae. Our first approach consisted of designing, genetically constructing and functionally analysing truncated versions of GFP-tagged Jen1, lacking parts of their cytosolic termini, expressed in wild-type or mutant S. cerevisiae strains lacking the arrestins Rod1 or Bul1/2. Our functional analyses included Jen1-mediated growth tests on lactic acid, or the effect on external pH, direct measurements of Jen1 transport kinetics using radiolabelled lactic acid, in vivo imaging of subcellular localization and western blot measurements of protein steady state levels of the truncated Jen1 versions. These constructs were expressed under induction conditions and in response to physiological signals triggering Jen1 endocytosis (i.e. presence of glucose). Subsequently, we generated and analysed functional chimeric transporters made of UapA, a heterologous nucleobase-allantoin transporter of A. nidulans, fused with the terminal regions of Jen1.

Deleting the entire Jen1 terminal regions, which correspond to 133 N-terminal or 62 C-terminal amino acids, as defined by in silico predictions, led to non-functional Jen1 versions, which in most cases were associated with significant cellular mislocalization, mostly ER retention. Alphafold 3D prediction of Jen1 structure was consistent with this result, as these deletions removed part of the transmembrane segments of Jen1. We thus proceeded by analysing shorter truncations, such as those deleting the 94 N-terminal or/and the 33 C-terminal amino acids. Notice that similar truncated versions of Jen1 have been previously analysed (Fujita et al., 2018), which allowed direct comparison of relevant results (see later). The shorter Jen1 truncations proved to be functional based on growth tests and other functional assays, similar to what has been previously reported in Fujita et al. (2018). Based on these Jen1 truncations and relative chimeras with UapA, we came to the following primary observations.

Jen1ΔCT33 is a stable version of Jen1 in all conditions tested that also shows increased capacity for lactate transport, not only because of its higher concentration in the PM, but also due to a 10-fold increased affinity for its substrate. This suggests that the 33-amino-acid C-terminal segment deleted in the truncated version contains not only a degron, as reported by Fujita et al. (2018), but also a functional motif that seems to affect the mechanism of transport of Jen1 in an ‘allosteric’ manner. This conclusion was based on the fact that the cytosolic C-terminus, as predicted by Alphafold, is distant from the proposed substrate translocation trajectory (Soares-Silva et al., 2011). A similar situation of regulation of transport kinetics by genetic modifications of the cytosolic C-terminus has only been reported in the case of FurE, an A. nidulans nucleobase-allantoin transporter (Papadaki et al., 2017, 2019). Additionally, we presented evidence that the 33-amino-acid C-terminal segment of Jen1 contains the major Rod1-interacting motif, as also reported in Fujita et al. (2018), given that its deletion (i.e. ΔCT33) mimicked the absence of glucose-triggered endocytosis of Jen1 observed in a rod1-null mutant. Furthermore, in the absence of Bul1/2, full endocytosis was observed in wild-type Jen1 and Jen1ΔNT94, but not in Jen1ΔCT33. Finally, we showed that the interaction of Rod1 with the 33-amino-acid C-terminal segment of Jen1 is very probably direct and context independent, because its transfer to the endocytosis-insensitive UapA proved sufficient to promote Rod1-dependent downregulation in the presence of glucose. Overall, our results concerning the C-terminal part of Jen1 confirm the conclusions presented in Fujita et al. (2018), but further reveal two novel and important properties of this part of the transporter. First, the C-terminus of Jen1 regulates the transport mechanism from a distance and, second, Rod1 recognizes a motif in the C-terminus of Jen1 without the involvement of other regions of the transporter. To our knowledge, there is no other report showing a context-independent interaction of a transporter motif with α-arrestins.

Jen1ΔNT94 was shown to be normally produced at basal levels, but proved to be a rather unstable version of Jen1, exhibiting rapid turnover upon further induction. Thus, the N-terminal 94 residues segment of Jen1 would be excepted to include elements critical for post-translational stability, evident upon translocation to the PM. Interestingly, Jen1ΔNT94 showed moderately altered transport kinetics (e.g. 2.5-fold increased substrate affinity), which points to a positive ‘distant’ effect on the transport mechanism, albeit weaker than that of the C-terminal segment. Notably also, the N-terminal part proved critical for endocytic downregulation in response to glucose, because when the ROD1 gene was knocked out or the C-terminus of Jen1 was deleted (i.e. no Rod1 binding), the presence of the N-terminus conferred partial endocytosis, whereas its absence led to an increased stability. Our data further support the conclusion that glucose-triggered endocytosis of Jen1 can also be exerted via Bul1/2 binding to the N-terminal segment, as endocytosis without the C-terminal region (Jen1ΔCT33) or without an active ROD1 depends solely on Bul1/2.

A clear conclusion concerning the termini of Jen1 is that both are needed for maximal glucose-triggered endocytosis of Jen1, with the N-terminus interacting with Bul1/2 and the C-terminus with Rod1. The interaction of Rod1 with the C-terminus seems to result in stronger Jen1 endocytosis when Bul1/2 interaction with the N-terminus is blocked. The interaction of Bul1/2 with the N-terminus confers only partial endocytosis when the Rod1 interaction with the C-terminus is genetically abolished. In our opinion, the most interesting novel finding of this work is the evidence supporting the hypothesis that the two Jen1 termini co-operate in regulating the stability and function of Jen1. A first genetic indication supporting this idea came from the doubly truncated Jen1 version, which showed exceptional new properties that were different from those of the singly truncated mutants and the wild-type Jen1. Specifically, Jen1ΔCT33ΔNT94 shows very high PM stability under all conditions and genetic backgrounds tested. Thus, the doubly truncated Jen1 version is practically a ‘new’ monocarboxylate transporter that is endocytosis resistant and that has a 3-fold increased substrate affinity relative to the wild-type Jen1.

To address the molecular basis underlying the additive functional roles of Jen1 termini, we employed a BiFC assay, which showed a dynamic and transport-dependent interaction of the two termini of Jen1. Using the same assay, we also obtained strong evidence that Jen1 does not form dimers, at least in the conditions tested, which proved fortuitous for more rigorously interpreting the positive BiFC signals obtained when the two parts of split YFP were cloned in the same Jen1 molecule. The only other previously reported case where BiFC assays showed that cytosolic termini interact to control the function and turnover of a transporter is that of FurE in A. nidulans (Papadaki et al., 2017, 2019). In this case, interactions of the two termini affected the stability, trafficking, function and endocytosis of FurE and, surprisingly, also its substrate specificity. Preliminary molecular dynamic analysis has provided some hints on how cytosolic termini might have affected FurE functioning from a distance by modifying the opening and closing of outer and inner gates of the transporter (Papadaki et al., 2019). In the present work, Jen1 truncations did not seem to affect substrate specificity, but, interestingly, all functional Jen1 truncations showed increased (2.5- to 10.0-fold) affinities for lactic acid transport, despite retaining wild-type V_max values (see Fig. 3). The alteration in K_m values reveals a modification in the capacity of Jen1 truncations to bind native substrates. In other words, similar to FurE, changes on the cytosolic termini of Jen1, distantly positioned from the substrate-binding site, affect the mechanism of substrate selection and transport. How this is achieved in the case of Jen1 remains elusive, but constitutes an interesting point to be addressed in the future.

The present work on Jen1 shows that transporter cytosolic termini can be exploited to rationally modify transporter function, which will be valuable, not only for addressing basic mechanisms of solute transport, but also for serving as tools in biotechnological applications (Kim et al., 2014; Knychala et al., 2022; Tanaka et al., 2017; for a review, see Barata-Antunes et al., 2021). The generality of this concept is supported by the work on FurE and Jen1, representing the two major transport superfamilies, APC and MFS, but also several other reports directly or indirectly supporting the emergence of transporter termini as important functional elements (Mikros and Diallinas, 2019).

All the yeast strains used in this work are listed in Table S2. The strains jen1Δ, rod1Δ, bul1Δbul2Δ and rod1Δbul1Δbul2Δ were derived from the 23344c wild-type strain (our laboratory collection). For BiFC analysis, a jen1Δ strain derived from W303-1A was used (our laboratory collection). Yeast cells were grown in a synthetic minimal medium with 0.67% (w/v) yeast nitrogen base (Difco), supplemented to meet the auxotrophic requirements (YNB medium) or in yeast extract (1%, w/v) and peptone (1%, w/v) (YP medium). Solid medium was prepared by adding agar (2% w/v) to the respective liquid medium. Carbon sources utilized were glucose (2%, w/v), lactic acid (0.5%, v/v, pH 5.0), galactose (2%, w/v) or glycerol (3%, v/v). Growth was carried out at 30°C. Cultures were harvested during the mid-log phase of growth. Glucose-grown cells were then centrifuged (4500 g for 2 min), washed twice in deionized water and cultivated into a fresh YNB medium with lactic acid (incubation time is indicated). For GAL promoter induction conditions, the YNB medium was supplemented with a defined amino acid drop-out mixture: –uracil +40 adenine (Formedium). Cells were grown overnight [until cells reached an optical density at 640 nm (OD₆₄₀) of 1.2–1.8] in YNB medium with 2% (w/v) glucose and then, after being washed twice in deionized water, they were transferred to YNB medium with 2% (w/v) galactose at a starting OD₆₄₀ of 0.2. Alternatively, cells were grown overnight (until cells reached an OD₆₄₀ of 0.5) in YNB medium with 2% (w/v) galactose (plus 0.1%, w/v, glucose), as described by Leung et al. (2010). Glucose (2%, w/v) was added when indicated.

The protein sequences were obtained from the SGD (http://www.yeastgenome.org) and FungiDB (https://fungidb.org/fungidb/app) databases. The secondary structures were predicted by TOPCONS (Bernsel et al., 2009). Tertiary structures were predicted by HHpred (http://toolkit.tuebingen.mpg.de/hhpred) and MODELLER software (Šali et al., 1995), as described previously (Soares-Silva et al., 2007, 2011). The minimum number of residues predicted in this work for N- or C-terminus of Jen1 are listed in Table S1.

Alphafold predictions were accessed from the Alphafold protein structure database (https://alphafold.ebi.ac.uk/) (Jumper et al., 2021; Varadi et al., 2022). Molecular graphics and analysis were performed with UCSF Chimera, developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco (Pettersen et al., 2004). Raytraced images were produced with POV-Ray x (http://www.povray.org/).

All constructions were performed by in vivo gap repair (Ma et al., 1987). First, DNA fragments were amplified by PCR (Accuzyme DNA Polymerase, Bioline, or Supreme NZYProof DNA Polymerase, Nzytech) with specific oligonucleotides (listed in Table S3) using yeast genomic DNA (unless specified otherwise). The resulting PCR products were co-transformed with a linearized plasmid (digested with a specific restriction enzyme) in S. cerevisiae cells. All plasmids used and constructed are listed in Table S4. Specifically, for construction of JEN1 termini truncated versions pGPDJEN1ΔNT133, pGPDJEN1ΔCT33, pGPDJEN1ΔCT62, JEN1 gene DNA fragments were amplified using the following oligonucleotides pairs, respectively: D-NTJEN1_133 and RCTJEN1; D-CTJEN1_33 and CYC1TERM; and D-CTJEN1_62 and CYC1TERM. The resulting PCR products were co-transformed with the linearized plasmid pGPDJEN1. For pGPDJEN1ΔNT94 and pGPDJEN1ΔNT94ΔCT33 constructions, the DNA fragments were amplified from pGPDJEN1 and pGPDJEN1ΔCT33, respectively, using the oligonucleotides Fw_GPD_jen1dNT94 and Rev_GFP_jen1. The resulting PCR products were co-transformed with the linearized plasmid p416GPD. For construction of JEN1 termini truncated versions under the control of the GAL promoter, pGALJEN1ΔCT33, pGALJEN1ΔNT94 and pGPDJEN1ΔNT94ΔCT33, the GAL DNA fragment was amplified from pGALJEN1 with the oligonucleotides GPDfwd and GALrev for pGALJEN1ΔCT33 construction or with the oligonucleotides GPDfw_new and GALrev_dNT94 for constructions pGALJEN1ΔNT94 and pGPDJEN1ΔNT94ΔCT33. The resulting GAL PCR products were co-transformed with the respective linearized plasmid (pGPDJEN1ΔCT33, pGPDJEN1ΔNT94 or pGPDJEN1ΔNT94ΔCT33). For the construction of pGALUAPA/JEN1CT33, the GALUAPA DNA fragment was amplified from pDDGFP2UAPA using the primers 381 and UapA_rev_33; the JEN1CT33 DNA fragment was amplified from pDDGFP2UAPAΔCT/JEN1CT62 using the primers Jen1_fw_ct33 and 317. These DNA fragments were then co-transformed with linearized p426GPD plasmid previously digested with SacI and XhoI restriction enzymes to remove the GPD promoter. The pDDGFP2UAPAΔCT/JEN1CT62 plasmid was derived from pDDGFP2UAPA (Leung et al., 2010). For the construction of pGALUAPA/JEN1NT94, the GALJEN1NT94 DNA fragment was amplified from pGALJEN1 using the primers 381 and REV_Jen1_UapA; the UAPAGFP DNA fragment was amplified from pDDGFP2UAPA using the primers FW_UapA_Jen1 and 317. These DNA fragments were then co-transformed with linearized p426GPD plasmid previously digested with SacI and XhoI restriction enzymes to remove the GPD promoter. For the construction of pGALUAPA/JEN1NT94-CT33, GALJEN1NT94 DNA fragment was amplified from pGALJEN1 using the primers 381 and REV_Jen1_UapA; the UAPA/JEN1CT33GFP DNA fragment was amplified from pGALUAPA/JEN1CT33 using the primers FW_UapA_Jen1 and 317. These DNA fragments were then co-transformed with linearized p426GPD plasmid previously digested with SacI and XhoI restriction enzymes to remove the GPD promoter. For the construction of pGALJEN1CT33/UAPA, JEN1CT33 DNA fragment was amplified from pGPDJEN1 using the primers Fw-CT33-UapA and Rev-CT33-UapA. This DNA fragment was then co-transformed with linearized pGALUAPA/JEN1NT94 plasmid previously digested with BamHI restriction enzyme.

Plasmid isolation from S. cerevisiae and E. coli strains was performed by standard protocols. Transformations were performed by the standard lithium acetate/polyethylene glycol method (Gietz and Woods, 2002). All constructs were confirmed by DNA sequencing (GATC Biotech and MWG Eurofins).

Transport activity assays for wild-type Jen1 and truncated Jen1 transporters were performed as previously described (Soares-Silva et al., 2007) using radiolabelled D,L-[¹⁴C] lactic acid (Amersham Biosciences). Yeast cells were grown to mid-exponential phase in glucose and transferred to a fresh 0.5% (v/v) lactate medium for 4 h (Lac 4 h). For uptake measurements, yeast cells were harvested in Lac 4 h and centrifuged (4500 g, 2 min). The samples were then washed twice with ice-cold deionized water and resuspended in ice-cold deionized water to a final concentration of about 20–40 mg dry weight/ml. The reaction mixtures were prepared in 1.5 ml tubes containing 60 μl of KH₂PO₄ (0.1 M, pH 5.0), and 30 μl of the yeast cell suspension. After incubation, the reaction was started by the addition of 10 μl of 6 mM radiolabelled lactic acid, pH 5.0 [specific activity, 2000 disintegrations per minute (dpm)/nmol], rapidly mixed by vortexing, and incubated for 1 min. After 1 min, 100 μl of 100 mM non-labelled substrate, was added, quickly mixed by vortexing and the mixture was chilled on ice, to stop the reaction. The reaction solutions were centrifuged for 5 min at 6200 g. The supernatant was rejected, and the pellet was resuspended in 1 ml of deionized cold water and centrifuged for 5 min, at 6200 g. The resulting pellet was resuspended in 1 ml of scintillation liquid (Opti-Phase HiSafe II; LKB FSA Laboratory Supplies). Radioactivity was measured in a Packard TRI-CARB 4810 TR liquid scintillation spectrophotometer with dpm correction. The percentage uptake rate of wild-type Jen1 (Jen1 WT) is considered 100%. The data is represented as a scatter plot with bar (mean±s.d.) (Prism 9.0; GraphPad software, version 9.2.0) of all data points obtained in three independent experiments. For kinetic assays, the methodology used was the same as described above for transport activity assays. However, in this case, the cells were exposed for 30 s to various concentrations of radiolabelled D,L-[¹⁴C] lactic acid, ranging from 0.03 to 2 mM. The data is represented as a Michaelis–Menten plot of the net initial velocity relative to increasing lactic acid concentrations, showing the mean values of at least three independent experiments (n≥3). The error bars represent the s.d. K_m and V_max were determined using the Prism 9.0 (GraphPad software, version 9.2.0) with 95% confidence interval and P<0.0001 [K_m, Michaelis–Menten constant; s.d., standard deviation; V_max, maximum velocity]. Transport activity assays for wild-type UapA and UapA-Jen1 chimeras were performed essentially as described in Leung et al. (2010), using radiolabelled [³H]-xanthine (21.1 Ci/mmol, Moravek Biochemicals, Brea, CA).

Phenotypic growth assays on solid medium were performed according to Soares-Silva et al. (2011) and Talaia et al. (2017). A serial of 1:10 yeast cell dilutions (starting from an OD₆₄₀ of 1) were performed, and 3 μl of each yeast suspension were plated in YNB solid medium, containing the desired carbon source. Cells were incubated at 30°C or 18°C, for 4 or 7 days, respectively. Three independent experiments were performed.

Yeast cells were grown, as described above, and visualized by fluorescence microscopy. A volume of 1 ml of growing yeast cells was collected and concentrated by a factor of 10. 5 μl of each sample was then directly visualized, without fixation, on a Leica DM5000B microscope with appropriate filters. For the vacuolar staining, cells were incubated for 10 min at 30°C with 1.5 μl of CellTracker™ Blue 7-amino-4-chloromethylcoumarin (CMAC; Invitrogen). The resulting images were acquired with a Leica DFC 350FX R2 digital camera using the LAS AF software. Images were then processed in the Adobe Photoshop CC 2018 (Adobe Systems).

The quantification of the ratio between the fluorescence at the cell periphery and the total fluorescence after 4 h glucose treatment was performed using ImageJ software (version 1.53k) as described in Hovsepian et al. (2018). The data is represented as a scatter plot with median (n≥45 cells). An ordinary one-way ANOVA analysis was used followed by a Tukey's multiple comparisons test using Prism 9.0 (GraphPad software, version 9.2.0). The P values are indicated [not significant (NS), P>0.05; *P<0.05; **P<0.01; ***P<0.001; ****P<0.0001].

The pH of the culture medium was determined as previously described (Talaia et al., 2017). A volume of 1 ml of cell culture was harvested and the pH value was immediately measured by a pHmeter (Braun). The data is represented as an interleaved scatter plot (mean±s.d.) (Prism 9.0; GraphPad software, version 9.2.0) of at least three independent experiments.

Yeast cells were grown as above, and crude protein extracts were prepared as previously described (Paiva et al., 2009). Nitrocellulose membranes (GE Healthcare Life Sciences) were probed with the primary antibodies as follows: anti-GFP (monoclonal, mouse IgG1κ, clones 7.1 and 13.1, Roche, 11814460001) and anti-PGK (monoclonal, S. cerevisiae, clone 22C5D8, Invitrogen), used at 1:3000 and 1:10,000 dilutions, respectively. The anti-mouse-IgG (whole molecule)-peroxidase produced in rabbit (A9044, Sigma) was used as a secondary antibody at 1:10,000 dilution. The signal was detected by enhanced chemiluminescence using the WesternBright ECL HRP substrate (advansta). The membranes were exposed to X-ray films (Ortho CP-G Plus, Agfa) and the chemiluminescence signals were acquired using an automatic film processor (Curix 60, Afga).

The quantifications of the bands were performed using the ImageJ software (version 1.53k). The quantification levels of the Jen1–GFP signal were divided by the respective values of PGK (used as a loading control). The relative levels of protein expression of the Jen1 truncations were normalized to that of Gal at 5 h (set as 100%). The data is represented as a column bar graph (mean±s.e.m.) (Prism 9.0; GraphPad software, version 9.2.0) of at least three independent experiments.

For BiFC analyses, several plasmid constructions were made (Table S4): pGPDJEN1YFP_N (URA3), pGPDJEN1YFP_C (HIS3), pGPDYFP_NJEN1YFP_C (URA3), pGPDYFP_N (URA3) and pGPDYFP_C (URA3), using a GAP repair cloning strategy. The N-terminal half of the yellow fluorescent protein (YFP_N; 154 amino acid residues of YFP), and the C-terminal half of YFP (YFP_C; 86 amino acid residues of YFP) were amplified from plasmids PDV7 and PDV8 (Zekert et al., 2010), respectively. The JEN1 open reading frame (ORF) was amplified from pGPDJEN1 plasmid. The primers used are listed in Table S3.

To study the possible interaction of the Jen1 N-terminus with the C-terminus, jen1Δ cells expressing pGPDYFP_NJEN1YFP_C were grown overnight in glycerol (3%, v/v), supplemented with the required auxotrophies, until mid-exponential phase, to induce Jen1 expression at the PM. Then, a pulse of lactate (0.5%, v/v) was added, and fluorescent images were acquired at the indicated time points. Cells co-expressing pGPDJEN1YFP_N (URA3) and pGPDJEN1YFP_C (HIS3) or expressing pGPDYFP_N (URA3), pGPDYFP_C (URA3) or pGPDJEN1GFP (URA3) were used as controls.

We thank Olivier Vincent and Bruno André for fruitful discussions. Molecular graphics and analysis were performed with UCSF Chimera, developed by the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco, with support from NIH P41-GM103311 (Pettersen et al., 2004).

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.260059.