Direct relationship between increased expression and mistrafficking of the Charcot-Marie-Tooth-Associated protein PMP22

: Charcot-Marie-Tooth disease (CMT) is a neuropathy of the peripheral nervous system that afflicts ~1:2500 people. The most common form of this disease (CMT1A, 1:4000) is associated with duplication of chromosome fragment 17p11.2-12, which results in a third wild-type PMP22 allele. In rodent models overexpressing the peripheral myelin protein 22 (PMP22) protein and in dermal fibroblasts from patients with CMT1A, PMP22 aggregates have been observed. This suggests that overexpression of PMP22 under CMT1A conditions overwhelms the endoplasmic reticulum (ER) quality control, leading to formation of cytotoxic aggregates. In this work, we used a single-cell flow-cytometry trafficking assay to quantitatively examine the relationship between PMP22 expression and trafficking efficiency in individual cells. We observed that as expression of wild-type or disease variants of PMP22 is increased, the amount of intracellular PMP22 increases to a greater extent than the amount of surface-trafficked protein. This was true for both transiently transfected cells as well as PMP22 stable expressing cells. Our results support the notion that overexpression of PMP22 in CMT1A leads to a disproportionate increase in misfolding and mis-trafficking of PMP22, which is likely a contributor to disease pathology and progression.

mutations in the PMP22 gene, including the most common form of the disease (CMT1A) (6). CMT1A is linked to a heterozygous duplication of chromosome fragment 17p11. 2-12, resulting in the production of a third copy of the PMP22 allele (trisomy). While 9 other proteins are also overexpressed in this process, it has been shown that CMT1A is caused by the additional copy of PMP22 (1, 2, 7). Whether the disease arises from an absolute increase in protein expression or resulting fluctuations in levels of PMP22 is up for debate; however, it is clear that increasing the copy number of PMP22 in rodents results in pathological phenotypes similar to those observed in patients with CMT1A (7)(8)(9). Deletion of a PMP22 allele (WT/null conditions) and genetically dominant point mutations in PMP22 also result in forms of CMT(1, 6).
PMP22 encodes a tetraspan integral membrane protein, PMP22, comprising 2-5% of the protein content in compact myelin of the PNS (10,11). The specific biological functions of PMP22 are still under investigation; however, substantial evidence suggests that it plays a structural role in the maintenance and development of compact myelin(1). Indeed, when PMP22 is reconstituted into lipid bilayers, it is sufficient to induce wrapping of the membranes to produce myelin-like-assemblies (12). However, PMP22 has also been implicated in a number of other processes within myelinproducing Schwann cells (1, 2, [13][14][15] (16). In vitro studies revealed that PMP22 is only modestly stable in detergent micelles, with the native conformation favored over the denatured ensemble by only 1.5  0.1 kcal mol -1 (17,18). That the folded form is marginally stable appears to be directly related to why only ~20% of synthesized wild type (WT) protein is able to fully mature and traffic to the plasma membrane (PM) (17,19,20).
Why a third copy of PMP22 results in neuropathy is not clearly established. Patients afflicted with CMT1A present slowed nerve conduction velocity and axonal loss, accompanied by a shortening of internodal length(1). Beyond these morphological changes in compact myelin, overexpression of PMP22 has also been shown to cause Schwann cell apoptosis (21,22). Mouse models with multiple copies of WT PMP22 or expressing disease mutant forms of PMP22, as well as CMT1A patientderived dermal fibroblasts, display the presence of cytosolic PMP22 aggregates coinciding with a decrease in proteasomal activity (1,8,13,23,24). Under normal conditions, PMP22 that is not able to mature beyond the endoplasmic reticulum (ER) is removed from the membrane by ER quality control and degraded in the cytosol via proteosomal and/or lysosomal pathways (23). One hypothesis for CMT1A pathology is that under normal conditions expression of PMP22 occurs at levels that approach saturation of the ER protein folding quality control system, such that introduction of a third copy of PMP22 overwhelms the system leading to ER stress and accumulation of cytotoxic aggregates. This is supported by the fact that stimulation of autophagy leads to increased degradation of such aggregates and improved myelination in cultures and in mice (25,26). In this work, we quantitatively examine the question of whether increased expression of PMP22 in model cell lines results in increased formation of intracellularly-trapped protein and a decrease in PM trafficking efficiency.

Measurement of PMP22 Trafficking Efficiency
For this study we used a previously described single-cell flow cytometry-based assay that quantitates the levels of both PM and intracellular PMP22 (17). Because the fusion of fluorescent reporter proteins to PMP22 has been shown to cause intracellular retention and aggregation, we inserted a c-myc epitope tag into the 2 nd extracellular loop of the protein (Fig. 1A). This modification has been shown not to affect protein function or turnover (17,27). In our experimental workflow (Fig. S1), PMP22-expressing cells are harvested and fixed followed by incubation with a myc antibody conjugated to phycoerythrin (PE), which labels surface-expressed PMP22 (properly trafficked, mature PMP22). Following a wash step, cells were permeabilized and internal PMP22 (misfolded PMP22 plus a likely small population of actively folding/trafficking protein) was labeled with the same anti-myc antibody but this time conjugated to AlexaFluor647 (AF647). Surface and internal concentrations of PMP22 in single cells were then quantified using flow cytometry with fluorimetric detection. Fluorescent intensities of the two fluorophores were normalized by fixing and permeabilizing a population of PMP22 expressing cells, splitting the population in two, and labeling each half with only one antibody.
This experimental setup was used to measure WT PMP22 trafficking efficiency in transiently transfected HEK293 cells. We define trafficking efficiency as the fraction of PMP22 localized to the PM compared to the total PMP22 in the cell (= PM Intensity/(PM Intensity + Internal Intensity)). Efficiency can vary from 0.0 (all PMP22 is intracellular) to 1.0 (all PMP22 is found at the PM). In three biological replicates, interrogating 2500 individual HEK293 cells per experiment, we found that WT PMP22 traffics to the PM with a mean trafficking efficiency of 0.27  0.01 (Mean  95% Confidence Interval; Fig. 1B).
We pooled the 7500 cells and binned them according to their trafficking efficiencies in bin sizes of 0.04 efficiency units (Fig. 1B). A population analysis of the cells revealed a left-skewed gaussian distribution centered around 0.2 corresponding with previously measured PMP22 trafficking efficiencies. We also reanalyzed previously reported data collected using the same experimental setup in Madin-Darby Canine Kidney (MDCK) cells (17). As shown in Fig. 1C, we observed a similar population distribution of trafficking efficiencies, but the curve is left-shifted relative to HEK293 and the mean trafficking efficiency (0.17 ± 0.03) is lower.

WT PMP22 Trafficking Efficiency as a Function of Total Expression
Normalized, background-corrected, total relative fluorescent intensity (RFU) of PMP22, corresponding to total cellular PMP22, ranged from roughly 50 to 125,000 RFU in individual HEK293 cells. In Fig. 2A we grouped measurements into 10 equal sized bins (750 cells per bin) based on total single-cell expression of PMP22 and plotted mean trafficking efficiency in each bin (for mean PMP22 expression in each bin see Fig. S2A). We observed a concentrationdependence wherein increased expression leads to decreased trafficking efficiency. This data is interesting in light of the common CMT1A phenotype in which overexpression of PMP22 causes disease. To ensure that this observation was not cell line dependent, we performed the same analysis on data from MDCK cells (17). MDCK cells also displayed decreased PMP22 trafficking efficiency at high total expression levels( Fig.  2B).
We next deconvoluted our data to examine the concentration-dependence of PM and intracellular populations of PMP22. For both cell lines, as the amount of total PMP22 increases, both internal and PM PMP22 increases (Fig. 2C, 2D). Fig. 2E-F quantitate the bin-to-bin changes in internal and surface PMP22. For both HEK293 and MDCK cells there is an initial (bin 1 to bin 2) jump in internal protein, followed by a plateau of steady increase where the bin to bin growth rate (measured as the fraction of mean protein expression in the two bins) is the same for both internal and surface PMP22 population. The growth of the internal PMP22 population increases at higher total expression levels, with the growth of the PM population remaining constant (MDCK cells) or increasing only between bins 9 and 10 (HEK293 cells). In total, Fig. 2 shows that PMP22 trafficking efficiency decreases as total expression increases. The burden of trapped PMP22 in the cell increases disproportionately relative to the growth of the PM population of the protein.

Trafficking Efficiency for PMP22 Disease Mutants as a Function of Total Expression
Trafficking efficiency versus total expression was examined for three diseasecausing PMP22 variants: L16P, A67T, and G93R. The L16P mutation, which results in the "Trembler-J" mouse phenotype, causes severe demyelination and consequent disease.
G93R causes moderate demyelination and moderate CMT, while A67T causes a mild form of demyelination and a phenotype known as hereditary neuropathy with liability to pressure palsies (HNPP)(1, 6). The trafficking results for these three mutants are shown in  Fig. 3D shows that the average per-cell total expression levels for each mutant and for WT are comparable.
For L16P PMP22 the most precipitous decrease in trafficking efficiency was seen from bin 1 to bin 2 ( Fig. 3A and Fig. S3A-B). This suggests that the pathway for productive folding and trafficking of L16P PMP22 is relatively efficient at very low total expression levels. After bin 1, the surface trafficking efficiency is reduced to roughly 3%, where it remains over a wide range of total expression levels. The data for A67T and G93R ( Fig. 3B-C and Fig. S3C-F) is similar to that of WT PMP22 in that the highest trafficking efficiency for these mutants occurs at the lowest levels of total expression and trafficking efficiency gradually decreases as total expression increases. As for WT, the fractional growth of internal PMP22 population is usually higher than for the PM population at the initial bin 1to-2 transition and also for the later bins (Fig.  S3B, S3D, S3F).

PMP22 Trafficking Efficiency Under Conditions of Stable Expression
In the above experiments, transient transfection was used to express PMP22 in HEK293 and MDCK cells. This process results in high levels of PMP22 expression.
In contrast, stable expressor cell lines, which involve stable integration of plasmid DNA into the host cell genome, generally result in lower protein expression per cell (28). To examine PMP22 trafficking under these conditions we generated HEK293 cells that stably expressed myc-tagged WT PMP22 (stable cells; SI Fig. 4).
In Fig. 4A, we measured WT PMP22 PM and internal expression in single cells and plotted the population distribution of PMP22 trafficking efficiencies. The mean trafficking efficiency was higher (0.57  0.01) and the population distribution was significantly different (p<0.05) than in transiently transfected cells (distribution compared using a pairwise Kolmogorov-Smirnov test (29)). Fig. 4B shows that the average per cell expression of PMP22 in our stable cell line was significantly lower than in transiently transfected cells. These results further support the idea that lower PMP22 expression leads to higher trafficking efficiency.
We binned WT PMP22 stable cell trafficking data as described above, splitting the data into 10 equal sized bins based on the total PMP22 expression in each cell (~2200 cells per bin; for average PMP22 expression per bin see Fig. S5A). Fig. 4C shows that as total expression increases, there is an initial reduction in trafficking efficiency from about 80% at very low total PMP22 to about 55%, where it remains fairly constant over the remainder of the bins. Fig.  S5B-C show that from bin 1 to bin 5 there is a major jump in the proportion of internal PMP22 relative to a much smaller increase in the PM population, while in later bins both PM and internal populations grow at roughly the same pace.

Discussion:
WT PMP22 is overexpressed in the most common form of CMT disease, CMT1A. This has led to interest in the possibility that misfolded, non-degraded, PMP22 is a source of cytotoxicity contributing to disease etiology (13,23,24,30,31). Here, we examined expression of PMP22 in both transiently-transfected model cell lines and in stable cells. It should be considered that PMP22 expression in myelinating Schwann cells may be better modeled by the results from the higher expressing transiently-transfected cells (Fig.  4B). In myelinating Schwann cells, PMP22 is expressed at very high levels. PMP22 is under the control of a powerful transcriptional system that utilizes two promotors, P1 and P2, and a super-enhancer upstream of P1 (7). Most cell types only express transcripts using P2 and consequently express only low levels of PMP22, whereas Schwann cells utilize both promotors to drive higher PMP22 expression. Transcription factor binding and open chromatin markers were found to be much more abundant at the PMP22 super-enhancer in myelinating Schwann cells compared to oligodendrocytes (32). Furthermore, deletion of this super-enhancer in a Schwann cell line dramatically reduced PMP22 transcripts (33). Under conditions of CMT1A trisomy, in which the super-enhancer and promotors are also duplicated, the level of expressed PMP22 may be even higher. We therefore argue that the results in transiently-transfected cells are more revealing than the results from stable cells.
Large populations of transiently transfected HEK293 and MDCK cells exhibited average WT PMP22 trafficking efficiencies in the vicinity of 0.20, which is what was previously documented for Schwann cells (19,20). It is notable, that the average trafficking efficiency in MDCK cells (0.17) is lower than in HEK293 cells (0.27). We suggest that this is because the average expression level of PMP22 in MDCK cells is higher than in HEK293 cells, leading to increased misfolding (Fig. 2C-D). This is supported by the fact that the trafficking efficiency in stable cells, where average total expression level is significantly lower (Fig.  4B), jumps to 0.57. This same trend of lower trafficking efficiency at high expression levels is seen within each population of transientlytransfected cells ( Fig. 2A-B) and to a modest degree in the stable cells (Fig. 4C). For both transiently-transfected cell lines, a jump is seen in the growth of the internal PMP22 population that is disproportionately higher than for the PM population both at low and high total expression levels ( Fig. 2E-F). Between the two extremes, the growth of the internal fraction is roughly the same as the PM fraction. This suggests complexity in terms of the cellular PMP22 folding, misfolding, and trafficking processes, complexity that is not hard to imagine given the intricacies of the ER folding quality control system (34). Fig. 2C-D show that high level expression of PMP22 results in a considerable burden of internal protein in cells, consistent with the hypothesis that misfolding of PMP22 and subsequent failure of the misfolded protein to be degraded contributes to CMT1A. Nevertheless, it is also seen that the population of PM PMP22 continues to grow as total expression increases, such that our data are consistent with the possibility that the added expression of WT PMP22 may also result in an aberrant gain of function effect.
Results for mutant forms of PMP22, A67T (mild CMT), G93R (moderate CMT), and L16P (severe CMT) show that the efficiency profile of A67T is similar to that of WT, the profile for G93R is similar to WT (except that the decline in efficiency with increased total expression is more steep), and L16P exhibits dramatically reduced efficiency even at the lowest expression levels, with a further reduction to a low level "basin" occurring as total expression increases (Fig. 3, Fig. S3). It is interesting that the total amount of internal protein seen for all three of the disease mutants is similar at the highest total expression levels (compare bin 10 in Fig. S3A, C, E), while the amount of PM protein increases dramatically for L16P<G93R<A67T. This does not imply that the loss of PM protein alone explains the differences in disease severity. CMT disease caused by heterozygous L16P/WT PMP22 expression (Dejerine-Sottas Syndrome, DSS) is much more severe than for WT/null patients who suffer from HNPP. The most likely explanation for this is that L16P undergoes misfolding and entrapment early in the secretory pathway of patients, and drags some of the WT protein down with it via WT/L16P heterodimerization (30,35). Consequently, the total loss in both L16P and WT forms of PMP22 is likely greater than occurs in WT/null conditions. Thus, possible toxicity associated with the misfolded/mistrafficked protein may be compounded by the reduction in the surfacetrafficked protein, leading to a more severe disease phenotype.

Conclusions:
Previous studies have shown that PMP22 trafficking efficiency is related to the stability of its folded structure. Even WT PMP22 appears to have only marginal conformational stability (18,36), which helps to explain its modest folding efficiency. Disease mutant forms of PMP22 are even less stable with mistrafficking and disease severity correlating with the degree of destabilization (17). Aggregates and other misfolded forms of proteins in the ER can be recognized and removed from the ER for degradation via either proteasomal or lysosomal pathways (23,34). However, if these systems are saturated by an overabundance of misfolded protein, then that protein will accumulate-potentially becoming a source of cytotoxicity. Saturation of the membrane protein quality control machinery appears to occur when PMP22 is overexpressed (37). Here, we explored the relationship between expression levels and surface trafficking efficiency of PMP22. It was found that there was a direct, negative, relationship between expression and trafficking efficiency. Moreover, reduced trafficking efficiency was due more to an increase in internal (likely misfolded) PMP22 at higher total expression levels than to reduced PM trafficking. While it has yet to be explored in tissue from CMT1A patients, the results of this work support the plausibility that the etiology of CMT1A is related to the accumulation of misfolded PMP22. Moreover, the long-term accumulation of misfolded PMP22 in Schwann cells and/or the weakened efficiency of protein degradation pathways with aging would help to explain why CMT1A is a progressive disorder.

Cloning
Human cDNA for PMP22 was subcloned into a pIRES2-EGFP mammalian expression vector. QuikChange mutagenesis was used to insert a myc epitope into the second extracellular loop of PMP22 within the pIRES2-EGFP vector and to make the various point mutations used in this study. Plasmids were purified using a GenElute HP Plasmid MidiPrep Kit (Sigma-Aldrich).

Cell Culture and Transfections
HEK293 cells were cultured in Dulbecco's modified Eagle medium (DMEM) containing 10% fetal bovine serum (FBS) and 1% pen/strep at 37 o C and 5% CO 2 . ~24 hours prior to transfection, cells were plated to be 50-60% confluent at the time of transfection. Cells were transfected using the calcium phosphate methodology, 6 cm 2 plates were transfected with 1.5 µg DNA. Transfection media was removed from cells ~12-15 hours post-transfection and cells were washed with phosphate buffered saline (PBS) and fresh culture media was added to each plate. Transfection efficiencies were routinely >50%.

Single Cell Flow Cytometry Trafficking Assay
The experimental workflow is depicted in Fig. S2 and a detailed description of the assay protocol is found in (17).

Stable Cell Line Generation
HEK293 cells were transfected with WT PMP22 as described above. After removal of the transfection media cells were kept under selective pressure via the addition of 750 µg/mL G418 (Invitrogen, Grand Island, NY) to the culture media. The medium was changed every other day for two weeks until cells were harvested via trypsinization and resuspended in 300 µL PBS+5% FBS and placed on ice. Using a FACS Aria III (BD Biosciences, San Jose, CA), single cells were selected based on their scattering area and width profiles and plated in individual wells in a 96-well plate containing growth media without G418. Cells were grown until confluency and progressively expanded to 6-well plates. Cells that stably expressed myc-tagged WT PMP22 were selected based on analysis via western blotting.

Statistics
Student's t-test was used to compare mean values and the Kolmogorov-Smirnov test was used to compare population distributions. All statistical analysis was performed in GraphPad Prism.

Data Availability
All data are either presented in the manuscript, in the on-line supporting information, or are available from the corresponding author (Charles Sanders, Vanderbilt University, chuck.sanders@vanderbilt.edu) upon request.

Acknowledgments:
We thank Arina Hadziselimovic for assistance with cloning, and Jonathan Schlebach for use of his original MDCK cell data and critical revisions of the manuscript.

Flow
Cytometry experiments were performed in the VMC Flow Cytometry Shared Resource.

Author contributions:
JTM conducted all experiments and statistical analysis for this work. All authors participated in data analysis and wrote the paper. CRS conceived of the work and directed the approaches used.