DPP6 Domains Responsible for Its Localization and Function*

Background: DPP6, a transmembrane protein with a large extracellular domain, is an auxiliary subunit of Kv4.2 potassium channels. Results: The extracellular domain is required for DPP6 export from the ER while intracellular domains impart the functional impact on Kv4.2. Conclusion: Different DPP6 domains are responsible for its localization and function. Significance: Understanding DPP6 function may provide insight into its role in neuronal development and disease. Dipeptidyl peptidase-like protein 6 (DPP6) is an auxiliary subunit of the Kv4 family of voltage-gated K+ channels known to enhance channel surface expression and potently accelerate their kinetics. DPP6 is a single transmembrane protein, which is structurally remarkable for its large extracellular domain. Included in this domain is a cysteine-rich motif, the function of which is unknown. Here we show that this cysteine-rich domain of DPP6 is required for its export from the ER and expression on the cell surface. Disulfide bridges formed at C349/C356 and C465/C468 of the cysteine-rich domain are necessary for the enhancement of Kv4.2 channel surface expression but not its interaction with Kv4.2 subunits. The short intracellular N-terminal and transmembrane domains of DPP6 associates with and accelerates the recovery from inactivation of Kv4.2, but the entire extracellular domain is necessary to enhance Kv4.2 surface expression and stabilization. Our findings show that the cysteine-rich domain of DPP6 plays an important role in protein folding of DPP6 that is required for transport of DPP6/Kv4.2 complexes out of the ER.

Dipeptidyl aminopeptidase-like protein 6, or DPP6 2 (also referred to as DPPX, BSPL, or potassium-channel accelerating factor KAF) (1), is a member of the prolyl oligopeptidase family of serine proteases. This family also includes proteins such as DPP8, DPP9, and most notably DPPIV (also known as CD26) (1,2). Most of these proteases work to remove dipeptides from other regulatory proteins and peptides. CD26 in particular is important for the regulation of immune, metabolic, CNS, and inflammatory functions, mediated by its exopeptidase activity (3)(4)(5). However, DPP6 has a point mutation at the enzymatic serine (to aspartate), and is among those in the family that have lost their peptidase function. This group is often referred to as the dipeptidyl aminopeptidase-like proteins (DPPLs) (1,6). The DPP6 gene encodes at least three different splice isoforms that diverge in the N-terminal domain. In situ hybridization of rat brain tissue showed that DPP6-S is the most prominent isoform in the CA1 region (7).
Structurally, DPP6 is a type II transmembrane protein that consists of a single transmembrane domain and a short intracellular N terminus. Strop et al. (2004) published the x-ray crystal structure of DPP6, illuminating many specific features of the protein. They found that two monomers of DPP6 associate to form a homodimer. Each monomer possesses an eight-bladed ␤-propeller and an ␣/␤ hydrolase domain, which both participate in dimer formation outside of the cell. By far the largest portion of the protein is the extracellular C terminus (803 aa).
The DPP6 protein structure is highly similar to the antigenic exopeptidase family member DPPIV/CD26 (32% sequence identity and 50% sequence similarity) (8). They share large extracellular C termini that are divisible into three domains. In DPP6, the first extracellular domain (231aa) is a glycosylation domain containing seven n-glycosylation sites. These sites occur primarily in the ␤ propeller, with only one site in the ␣/␤ hydrolase domain. The second extracellular domain (193 aa) is a cysteine-rich domain along the middle blades of the C terminus. Each monomer contains four disulfide bridges in this domain. CD26 also contains these disulfide bonds, along with a fifth bridge in blade 6 of its ␤-propeller. The third, "aminopeptidase" domain (134 aa) is at the end of the C terminus and contains the DPP6 serine mutation. Strop et al. (2004) speculate that this change may have actually resulted in the gain of function as a potassium channel accelerating factor. They also note that DPP6 contains several ␤-propeller motifs that "commonly act as scaffolds for protein-protein interactions," and may have additional unidentified binding partners. DPP6 similarity to CD26 goes beyond structure. CD26, like DPP6, also binds the extracellular matrix. In vitro binding assays have shown that CD26 binds to collagens I and II at its cysteine-rich domain, and also binds to fibronectin (9 -11).
DPP6 is mostly studied as an auxiliary subunit of voltagegated K ϩ channels of the Kv4 family. Kv4 channels, along with channels from the Kv1 family, primarily underlie the transient subthreshold-activating A-type current (I A ) in neurons. I A plays a major role in neuronal excitability by opposing depolar-ization, dampening action potential (AP) initiation and frequency (12). Kv4.2 is the main ␣ subunit for channels of this type in the somatodendritic regions of CA1 hippocampal neurons (13). As Kv4.2 is highly expressed in distal CA1 dendrites, these channels control backpropagation of APs into dendrites, significantly impacting synaptic plasticity (13)(14)(15).
A-type K ϩ channels underlying the dendritic A-current in CA1 neurons appear to exist as a complex consisting of three main proteins: the pore-forming Kv4 subunit along with two auxiliary proteins, a Kv channel-interacting protein (KChIP) and a dipeptidyl aminopeptidase-like protein (DPPL) (16,17). The expression of all three subunits is required to reproduce native-like I A currents in heterologous systems. In CA1 pyramidal hippocampal neurons, these channels are primarily Kv4.2, KChIP2, KChIP4, and DPP6 (18,19).
In heterologous co-expression and knockdown studies in neurons, DPP6 increases Kv4.2 surface expression and accelerates channel activation, inactivation, and recovery from inactivation (20,21). In addition, we have recently found that DPP6 is important for localizing Kv4.2 to the distal dendrites in CA1 neurons, impacting dendritic excitability synaptic and plasticity (22). DPP6 knock-out mice exhibited hyperexcitable dendrites with enhanced dendritic AP back-propagation, calcium electrogenesis, and induction of synaptic longterm potentiation.
DPP6 effects on Kv4.2 make it an important protein to study to better understand such crucial neuronal processes as synaptic integration and plasticity. However, recent reports suggest that DPP6 may have additional roles independent of that as a Kv4 auxiliary subunit (11,23,24). Recently, we reported a novel, Kv4-independent role for DPP6 in neuronal development. Electrophysiological, molecular, and imaging experiments showed that DPP6 binds to the extracellular matrix and has a significant impact on filopodia formation and stability, consequently affecting dendritic arborization, spine density, and synaptic function in CA1 neurons (11). Further characterization of DPP6's structure and function may contribute to our understanding of these processes and uncover possible roles in neuropsychiatric pathologies where the DPP6 gene has been implicated, including autism spectrum disorders (25,26), mental retardation (27), and amyotrophic lateral sclerosis (28,29).
Little is known about the intracellular trafficking of DPP6. However, a recent study showed that DPP6 expression and stability are not dependent on Kv4.2 (23). DPP6 expression levels are completely unaffected by the absence of Kv4.2 in mouse cortex, and DPP6 protein levels appear to be stable without Kv4.2. These results indicate that DPP6 has the capacity to traffic independently of Kv4.2, but this trafficking has not been examined in the literature. We show here, using live imaging and biochemistry that the cysteine-rich domain is important for DPP6 exit from the ER and localization to the cell surface. We also show that while the extracellular domain of DPP6 is important for enhancing the surface expression of Kv4.2, the short intracellular N-terminal plus transmembrane domains of DPP6 are necessary for it to associate with and accelerate the properties of Kv4.2.
Live Imaging and Measurements-HEK cells, which have an elevated morphology, were used in all live imaging experiments. HEK cells transfected with DPP6-WT or mutations and either ER-DsRed or membrane mCherry markers, after 24 h, live image with a Zeiss LSM 510 confocal microscope. Each coverslip was kept at 37°C until use, and was only used for imaging over a period of 15 min after being placed in the microscope with Invitrogen Leibovitz's (1ϫ) L-15 Medium. Measure the intensity plot along the line drawn.
Electrophysiology-HEK 293 cells were co-transfected with Kv4.2 (0.5 g) and DPP6 or one of its mutations (0.5 g) in 35-mm dishes using the Xtreme Gene 9 system (Roche Applied Science, Indianapolis, IN). For control recordings, Kv4.2 (0.5 g) was co-transfected with empty vector (0.5 g). Transfected cells were re-plated onto glass coverslips and recordings were made 16 to 20 h after transfection. Coverslips containing transfected HEK 293 cells were submerged in the recording chamber and exposed to a continuous flow of ACSF consisting of the following (mM): 145 NaCl, 3 KCl, 2 CaCl 2 , 2 MgCl 2 , 8 glucose, 10 HEPES. An infrared differential interference contrast (IR-DIC) videomicroscopy system (Zeiss Instruments, Diagnostic Instruments) was used to visualize cells. Patch pipettes were pulled from thick-walled borosilicate glass (Warner Instruments) to achieve a tip resistance of 2-5 M⍀. Pipettes were filled with an internal solution containing (mM): 146.5 K-gluconate, 7.5 KCl, 9 NaCl, 10 HEPES, 0.2 EGTA, and 1 MgCl 2 . Whole-cell voltage clamp recordings were made using a Multiclamp 700B amplifier (Molecular Devices) and Clampex 10.1 software (Molecular Devices). Signals were digitized at 10 kHz with Digidata 1440A (Molecular Devices) and filtered at 2 kHz. All recordings were performed at room temperature, and recordings with cell capacitances less than 7 pF were excluded.
During the current density experiments, cells were initially held at Ϫ60 mv followed by a 400 ms pulse at Ϫ120 mv and then a 500 ms pulse at ϩ60 mv to elicit the A-type current. This protocol was repeated 9 times for each cell. This protocol was also used to measure the A-type current of HEK 293T cells co-transfected with Kv4.2, KChIP2a, and in some cases either DPP6 or one of its mutations. The nine sweeps were averaged to form a trace from which the peak amplitude of I A was calculated. This amplitude was then divided by the cell capacitance to obtain the current density (pA/pF).
For the recovery experiments, HEK 293 cells were co-transfected with Kv4.2 (0.5 g) and DPP6-WT or DPP6-D-extra (0.5 g). Cells were prepared for recording in the same manner as above. Cells were initially held at Ϫ60 mv followed by a 400 ms pulse to Ϫ100 mv and then a 400 ms pulse to ϩ60 mv. After this induction of I A ,a second induction protocol was given 5,10, 15, 20, 25, 100, 200, or 500 ms later. At the end of each sweep the cell was given a 400 ms pulse to Ϫ100 mv, and then returned to Ϫ60 mv for 200 ms. This was repeated three times for each cell. Percent recovery was calculated by dividing the peak amplitude of each secondary current the initial peak I A .
All recordings were analyzed using Clampfit 10.1 (Molecular Devices), Microsoft Excel, and Igor Pro 6.02A (WaveMetrics). All values are expressed as means Ϯ S.E. Statistical significance was evaluated using Student's t test (unpaired, two tails) or ANOVA with Dunnett's post hoc multiple comparison test when comparing values against a basal control.

RESULTS
The Cysteine-rich Domain of DPP6 Is Required for ER-to-Membrane Transport-To investigate the functional roles of each extracellular DPP6 domain, we generated three truncation mutants: 1) "DPP6-D-Pep" with the aminopeptidase domain is deleted, 2) "DPP6-D-PepϩCys" with both the cysteine-rich and aminopeptidase domains are deleted, and 3) "DPP6-D-Extra" with the entire extracellular domain is deleted, including the glycosylation, cysteine-rich, and aminopeptidase domains (Fig.  1A).
We began by examining the effect of the extracellular domain deletions on membrane localization. HEK293 cells were cotransfected with DPP6-WT or deletion mutants and a membrane marker (Mem-mCherry), which encodes a fusion protein consisting of the N-terminal 20 amino acids of neuromodulin tagged with mCherry red fluorescent protein. Live-cell images were captured 24 h after transfection. We found DPP6-WT to be expressed in the plasma membrane, showing colocalization with the membrane marker ( Fig. 1B). DPP6-D-Pep showed similar expression as DPP6-WT (Fig. 1D). However, deletion of the cysteine-rich domain (DPP6-D-PepϩCys) showed an intracellular expression pattern (Fig. 1F). Interestingly, DPP6-D-Extra, which is missing the entire extracellular domain, appears to have retained membrane expression, as demonstrated by colocalization with the membrane marker (Fig. 1H).
We quantified these mutation effects on DPP6 surface membrane expression with a biotinylation assay. COS7 cells were transfected with DPP6-WT or the C-terminal deletions (Fig. 1, J and K). DPP6-WT and deletions DPP6-D-Pep and DPP6-D-Extra showed similar levels of membrane expression as suggested in the live imaging experiments. Mutation DPP6-D-PepϩCys, however, showed only 40% of WT surface expression (Fig. 1K, p Ͻ 0.05). Together, these results suggest that the cysteine-rich domain of DPP6 may be important for proper folding of the DPP6 extracellular domain, which is critical to its forward trafficking to the plasma membrane.
In the next experiment we examined the intracellular expression of the extracellular DPP6 mutations by co-expressing them in HEK293 cells with an ER marker labeled with Ds-Red to see if the poor surface expression of DPP6-D-PepϩCys was due to ER retention (Fig. 2). Accordingly, we did find that this mutation colocalized with the ER marker (Fig. 2E), while DPP6-WT and the other mutations showed less ER expression (Fig. 2, A, C, and G).
C349/C356 and C465/C468 Disulfide Bridges Are Key to DPP6 Forward Trafficking-Given our results showing that the cysteine-rich domain is necessary for DPP6 transport out of the ER, and since disulfide bonds play an important role in the folding and stability of proteins, we hypothesized that the two disulfide bridges formed at C349/C356 and C465/C468 have key roles in the proper folding of DPP6. We made cysteine point mutations at DPP6-C349A/C356A and C465A/C468A (Fig.  3A), and performed biotinylation and live imaging assays. The live imaging data (Fig. 3B) and intensity plots (Fig. 3C) shown  that mutation DPP6-C349A/C356A colocalizes with the ER marker, but not with the membrane marker (Fig. 3, D and E). Similar results were found for DPP6-C465AϩC468A (Fig. 4, B and E) and the double mutation of DPP6-C349A/C356A and C465A/C468A (images not shown). These findings were supported in a biotinylation assay showing that both single mutations and the double mutation DPP6-C349A/C356A ϩ C465A/ C468A have decreased surface expression compared with DPP6-WT (Fig. 3, F and G; Fig. 4, F and G). Together, these results indicate that both mutations individually play important roles in the transport of DPP6 from ER to the cell membrane but that these effects are not additive, implying that intact disulfide bridges are necessary for the proper folding of the protein.
C349/C356 and C465/C468 Disulfide Bridges of DPP6 Are Not Required for Interaction with Kv4.2, but Are Functionally Important-After finding that disulfide bridges in the cysteine rich domain are necessary for DPP6 transport from the ER to cell membrane we examined whether the disulfide bridges are important for DPP6 interaction with the primary subunit Kv4.2. We performed Co-IP experiment in COS7 cells, which were transfected with Kv4.2 and either DPP6-WT, DPP6-C349A/C356A, or DPP6-C349A/C356AϩC465A/C468A. After pull-down with a Kv4.2 antibody, Western blot analysis showed that both single and double bridge mutants are still able to bind Kv4.2 (Fig. 5A). However, since the mutants get trapped in the ER they do not act to enhance Kv4.2 surface expression as DPP6-WT does, as evidenced in both biotinylation assays (Fig.  5, B and C) in COS7 cells and electrophysiological recordings from HEK293 cells (Fig. 5D). Peak current density of Kv4.2mediated currents significantly increased after co-expression with DPP6 compared with the control (Fig. 5D; p Ͻ 0.05). In contrast, peak current density of Kv4.2 co-expressed with DPP6-C349A/C356A or DPP6-C349A/C356AϩC465A/C468A showed no significant difference ( Fig. 5D; p Ͼ 0.05). Together these data from the biochemical assay and the electrophysiological recording confirm that C349/C356 and C465/C468 disulfide bridges of DPP6 play a key role in the enhancement of Kv4.2 surface expression by DPP6 in these heterologous expression systems.   6, A and B, p Ͻ 0.05). Although DPP6-WT further increased membrane expression (Fig. 6, A and B; p Ͻ 0.05), the mutations of DPP6-C349A/C356A or DPP6-C349A/C356AϩC465A/ C468A abolished both the DPP6-and KChIP2-mediated enhancements (Fig. 6, A and B; p Ͻ 0.05). Similar results were found using KChIP3 (data not shown). These results were confirmed by measuring current density of Kv4.2 in HEK293 cells in electrophysiological recordings (Fig. 6C). That KChIP2 cannot rescue the trafficking effects of mutant DPP6 proteins seems to indicate that Kv4.2 complexing with DPP6 precedes that with KChIPs. This possibility is supported by our results showing that DPP6 mutants are trapped in the ER while a previous report found that KChIP proteins interact with Kv4.2 in post-ER vesicles (32). Taken together, these data show that there is a pool of Kv4.2 channels which reliably traffic to the surface without chaperone proteins but that both DPP6 and KChIP proteins, when properly folded, enhance the surface expression/stability of Kv4.2 channels.
Previously we found that deleting the entire extracellular domain of DPP6 (DPP6-D-Extra) showed some membrane expression (Figs. 1 and 2). As we found that the extracellular domain of DPP6 is not important for its interaction with Kv4.2, we investigated whether the extracellular domain of DPP6 is required for its regulation of Kv4.2's electrophysiological properties. Electrophysiological recordings of Kv4.2's recovery from inactivation were performed in HEK293 cells (Fig. 7, D and E). Recovery from inactivation was, as expected, significantly faster for cells co-expressing Kv4.2 and DPP6-WT (n ϭ 9) compared with the Kv4.2 with a GFP control (n ϭ 7, p Ͻ 0.05). Interest-ingly, recovery from inactivation was also significantly faster when Kv4.2 was co-expressed with DPP6-D-Extra (n ϭ 6, p Ͻ 0.05) and nearly identical to that found in DPP6-WT recordings. These findings indicate that the N-terminal intracellular domain of DPP6, but not the entire extracellular domain, is necessary for interaction and acceleration of Kv4.2 channel properties.
Finally, we examined which extracellular domains of DPP6 affect its function to enhance the surface expression of Kv4.2. We performed biotinylation assays in COS7 cells transfected with Kv4.2 and either a control plasmid, DPP6-WT or the other  NOVEMBER 14, 2014 • VOLUME 289 • NUMBER 46 C-terminal deletions (Fig. 8A). The surface expression of Kv4 In E, recovery percentage was calculated by dividing the second peak by the initial peak and plotted by delay between sweeps. Time to full I A recovery was much faster in cells co-transfected with Kv4.2 and DPP6-WT (n ϭ 9) than the control (n ϭ 7). Kv4.2 and DPP6-D-Extra (n ϭ 6) also showed a significant acceleration in recovery from inactivation and was not significantly different from Kv4.2 and DPP6-WT. Error bars represent S.E.

DISCUSSION
DPP6 Domains Responsible for Its Transport-In this study we examined the functional role of distinct domains within the structure of the transmembrane protein DPP6. We found that the extracellular, cysteine-rich domain of DPP6 is required for its trafficking out of the ER (Figs. 1 and 2). Either of two disul-  NOVEMBER 14, 2014 • VOLUME 289 • NUMBER 46 fide bridges (C349/C356 and C465/C468) in this region resulted in ER retention (Figs. 3 and 4), implying that proper folding of the protein is required for exit. Interestingly, a more severe deletion of the entire extracellular domain allowed membrane expression, perhaps indicating an ER-retention domain in the interceding glycosylation domain. ER retention was not rescued by co-expression with another prominent Kv4.2 accessory subunit KChIP2 (Fig. 6). DPP6 Domains Responsible for Its Interaction and Function in Kv4 Complexes-As shown previously (20,21), we found that DPP6-WT associates with Kv4.2 increasing its surface expression and accelerating its recovery from inactivation in HEK293 cells (Figs. 5, 7, and 8). Here we extend these results to show which domains are required for each function of DPP6. DPP6 association with Kv4.2 requires DPP6 N-terminal plus transmembrane domains (54 aa, Fig. 7). Interestingly, this domain fully replicated DPP6 functional effect of accelerating Kv4.2 recovery from inactivation in HEK293 cells (Fig. 7). Our results support a previous report (33) which used chimeric proteins to identify a region of the DPP family member DPPY (DPP10) from its N terminus to the end of its transmembrane domain to interact with another Kv4 family member, Kv4.3.

DPP6 Domains Responsible for Localization and Function
Although we found Kv4.2 to interact with the N-terminal and transmembrane domains of DPP6, we did not find their co-expression to result in enhanced Kv4.2 surface expression. To increase surface Kv4.2 beyond basal levels required the intact extracellular, cysteine-rich domain but not the extracellular aminopeptidase domain (Fig. 8, A and B). Data from biochemical and electrophysiological recordings confirmed that intact disulfide bridges of DPP6 (C349/C356 and C465/C46) are necessary to enhance Kv4.2 surface expression by DPP6 in these heterologous expression systems (Fig. 5). In contrast, total Kv4.2 protein was not affected by loss the cysteine-rich domain but did require the glycoslyation domain (Fig. 8, C and D) perhaps indicating that the unglycosylated protein degrades quickly.
Together, these results suggest that there exist multiple pools of Kv4.2 channels. First, a basal pool of Kv4.2 is able to traffic out of the ER without binding any accessory subunits. Although caution is warranted in extrapolating results obtained in heterologous expression systems to the native environment, this finding was suggested previously in a study investigating the effect of knocking out DPP6 on A-channel expression in hippocampal CA1 pyramidal neuron dendrites (22). Our results presented here suggest a second pool of Kv4.2 channels, which bind with DPP6 in the ER, resulting in increased surface expression of Kv4.2. This pool may correspond to the enhanced A-current found in distal CA1 dendrites of wild type but not DPP6 knock out mice (22).
Once out of the ER, Kv4-DPP6 channels can be further modulated by interaction with KChIP auxiliary subunits in a post ER vesicular pathway (32). Knockdown of DPP6 in cerebellar granule neurons and hippocampal pyramidal neurons results in a loss of both Kv4.2 and KChIP3 (24) suggesting that DPP6 may be required for Kv4 interaction with KChIPs.
Potential DPP6 Domains Responsible for Its Role in Neurodevelopment-We have recently published evidence indicating a role for DPP6 in neurodevelopment, which appears to be independent of its role as a Kv4 auxiliary subunit (11). In that study we found dendritic filopodia formation and stability to be impacted in DPP6-KO mice. Other groups have also found indications that DPP6, like its related, structurally similar family member DPPIV, is a multifunctional protein. Foeger et al. (2012) found that heterologously expressed DPP6 localizes to the cell surface in the absence of Kv4.2 in mouse cortex (23) while Nadin and Pfaffinger (34) reported that DPP6 interacts with the K2P channel TASK-3 to regulate resting membrane potential and input resistance in cerebellar granule cells (34). Moreover numerous genome-wide association screens for a variety of maladies implicate a potential role for DPP6 (26,27,35,36). Further studies will be required to determine which DPP6 domains are responsible for its various functions but it seems likely that its role in regulating filopodia involves an interaction between its ␤-propeller structure located in the cysteine-rich domain and the extracellular matrix protein fibronectin (11) as found for DPPIV (10) (37). Future studies may also address the role of the non-functional aminopeptidase domain of DPP6, which did not greatly impact protein trafficking in the present study.