Partitioning to ordered membrane domains regulates the kinetics of secretory traffic

Joint Public Review:

This paper by Castello-Serrano et al. addresses the role of lipid rafts in trafficking in the secretory pathway. By performing carefully controlled experiments with synthetic membrane proteins derived from the transmembrane region of LAT, the authors describe, model and quantify the importance of transmembrane domains in the kinetics of trafficking of a protein through the cell. Their data suggest affinity for ordered domains influences the kinetics of exit from the Golgi. Additional microscopy data suggest that lipid-driven partitioning might segregate Golgi membranes into domains. However, the relationship between the partitioning of the synthetic membrane proteins into ordered domains visualised ex vivo in GPMVs, and the domains in the TGN, remain at best correlative. Additional experiments that relate to the existence and nature of domains at the TGN are necessary to provide a direct connection between the phase partitioning capability of the transmembrane regions of membrane proteins and the sorting potential of this phenomenon.

The authors have used the RUSH system to study the traffic of model secretory proteins containing single-pass transmembrane domains that confer defined affinities for liquid ordered (lo) phases in Giant Plasma Membrane derived Vesicles (GPMVs), out of the ER and Golgi. A native protein termed LAT partitioned into these lo-domains, unlike a synthetic model protein termed LAT-allL, which had a substituted transmembrane domain. The authors experiments provide support for the idea that ER exit relies on motifs in the cytosolic tails, but that accelerated Golgi exit is correlated with lo domain partitioning.

Additional experiments provided evidence for segregation of Golgi membranes into coexisting lipid-driven domains that potentially concentrate different proteins. Their inference is that lipid rafts play an important role in Golgi exit. While this is an attractive idea, the experiments described in this manuscript do not provide a convincing argument one way or the other. It does however revive the discussion about the relationship between the potential for phase partitioning and its influence on membrane traffic.

Our detailed comments are listed below:

ER exit:
The experiments conducted to identify an ER exit motif in the C-terminal domain of LAT are straightforward and convincing. This is also consistent with available literature. The authors should comment on whether the conservation of the putative COPII association motif (detailed in Fig. 2A) is significantly higher than that of other parts of the C-terminal domain. One cause of concern is that addition of a short cytoplasmic domain from LAT is sufficient to drive ER exit, and in its absence the synthetic constructs are all very slow. However, the argument presented that specific lo phase partitioning behaviour of the TMDs do not have a significant effect on exit from the ER is a little confusing. This is related to the choice of the allL-TMD as the 'non-lo domain' partitioning comparator. Previous data has shown that longer TMDs (23+) promote ER export (eg. Munro 91, Munro 95, Sharpe 2005). The mechanism for this is not, to my knowledge, known. One could postulate that it has something to do with the very subject of this manuscript- lipid phase partitioning. If this is the case, then a TMD length of 22 might be a poor choice of comparison. A TMD 17 Ls' long would be a more appropriate 'non-raft' cargo. It would be interesting to see a couple of experiments with a cargo like this.

Golgi exit:
For the LAT constructs, the kinetics of Golgi exit as shown in Fig. 3B are surprisingly slow. About half of the protein remains in the Golgi at 1 h after biotin addition. Most secretory cargo proteins would have almost completely exited the Golgi by that time, as illustrated by VSVG in Fig. S3. There is a concern that LAT may have some tendency to linger in the Golgi, presumably due to a factor independent of the transmembrane domain, and therefore cannot be viewed as a good model protein. For kinetic modeling in particular, the existence of such an additional factor would be far from ideal. A valuable control would be to examine the Golgi exit kinetics of at least one additional secretory cargo.

Comments about the trafficking model
1. In Figure 1E, the export of LAT-TMD from the ER is fitted to a single-exponential fit that the authors say is "well described". This is unclear and there is perhaps something more complex going on. It appears that there is an initial lag phase and then similar kinetics after that - perhaps the authors can comment on this?

2. The model for Golgi sorting is also complicated and controversial, and while the authors' intention to not over-interpreting their data in this regard must be respected, this data is in support of the two-phase Golgi export model (Patterson et al PMID:18555781). Furthermore contrary to the statement in lines 200-202, the kinetics of VSVG exit from the Golgi (Fig. S3) are roughly linear and so are NOT consistent with the previous report by Hirschberg et al. Moreover, the kinetics of LAT export from the Golgi (Fig. 3B) appear quite different, more closely approximating exponential decay of the signal. These points should be described accurately and discussed.

Relationship between membrane traffic and domain partitioning:
1. Phase segregation in the GPMV is dictated by thermodynamics given its composition and the measurement temperature (at low temperatures 4degC). However at physiological temperatures (32-37degC) at which membrane trafficking is taking place these GPMVs are not phase separated. Hence it is difficult to argue that a sorting mechanism based solely on the partitioning of the synthetic LAT-TMD constructs into lo domains detected at low temperatures in GPMVs provide a basis (or its lack) for the differential kinetics of traffic of out of the Golgi (or ER). The mechanism in a living cell to form any lipid based sorting platforms naturally requires further elaboration, and by definition cannot resemble the lo domains generated in GPMVs at low temperatures.

2. The lipid compositions of each of these membranes - PM, ER and Golgi are drastically different. Each is likely to phase separate at different phase transition temperatures (if at all). The transition temperature is probably even lower for Golgi and the ER membranes compared to the PM. Hence, if the reported compositions of these compartments are to be taken at face value, the propensity to form phase separated domains at a physiological temperature will be very low. Are ordered domains even formed at the Golgi at physiological temperatures?

3. The hypothesis of 'lipid rafts' is a very specific idea, related to functional segregation, and the underlying basis for domain formation has been also hotly debated. In this article the authors conflate thermodynamic phase separation mechanisms with the potential formation of functional sorting domains, further adding to the confusion in the literature. To conclude that this segregation is indeed based on lipid environments of varying degrees of lipid order, it would probably be best to look at the heterogeneity of the various membranes directly using probes designed to measure lipid packing, and then look for colocalization of domains of different cargo with these domains.

4. In the super-resolution experiments (by SIM- where the enhancement of resolution is around two fold or less compared to optical), the authors are able to discern a segregation of the two types of Golgi-resident cargo that have different preferences for the lo-domains in GPMVs. It should be noted that TMD-allL and the LATallL end up in the late endosome after exit of the Golgi. Previous work from the Bonafacino laboratory (PMID: 28978644) has shown that proteins (such as M6PR) destined to go to the late endosome bud from a different part of the Golgi in vesicular carriers, while those that are destined for the cell surface first (including TfR) bud with tubular vesicular carriers. Thus at the resolution depicted in Fig 5, the segregation seen by the authors could be due to an alternative explanation, that these molecules are present in different areas of the Golgi for reasons different from phase partitioning. The relatively high colocalization of TfR with the GPI probe in Fig 5E is consistent with this explanation. TfR and GPI prefer different domains in the GPMV assays yet they show a high degree of colocalization and also traffic to the cell surface.