KLHL1 Controls CaV3.2 Expression in DRG Neurons and Mechanical Sensitivity to Pain

Dorsal root ganglion (DRG) neurons process pain signaling through specialized nociceptors located in their peripheral endings. It has long been established low voltage-activated (LVA) CaV3.2 calcium channels control neuronal excitability during sensory perception in these neurons. Silencing CaV3.2 activity with antisense RNA or genetic ablation results in anti-nociceptive, anti-hyperalgesic and anti-allodynic effects. CaV3.2 channels are regulated by many proteins (Weiss and Zamponi, 2017), including KLHL1, a neuronal actin-binding protein that stabilizes channel activity by recycling it back to the plasma membrane through the recycling endosome. We explored whether manipulation of KLHL1 levels and thereby function as a CaV3.2 modifier can modulate DRG excitability and mechanical pain transmission or sensitivity to pain. We first assessed the mechanical sensitivity threshold and DRG properties in the KLHL1 KO mouse model. KO DRG neurons exhibited smaller T-type current density compared to WT without significant changes in voltage dependence, as expected in the absence of its modulator. Western blot analysis confirmed CaV3.2 but not CaV3.1, CaV3.3, CaV2.1, or CaV2.2 protein levels were significantly decreased; and reduced neuron excitability and decreased pain sensitivity were also found in the KLHL1 KO model. Analogously, transient down-regulation of KLHL1 levels in WT mice with viral delivery of anti-KLHL1 shRNA also resulted in decreased pain sensitivity. These two experimental approaches confirm KLHL1 as a physiological modulator of excitability and pain sensitivity, providing a novel target to control peripheral pain.


INTRODUCTION
Nociceptive pathways are generally activated in response to noxious stimuli as protection from injury, yet chronic pain induces allodynia and hyperalgesia due to primary dysfunction, usually caused by nerve injury. Long-term changes triggered by nerve injury include altered gene expression in Dorsal root ganglion (DRG) and the spinal cord (Choi et al., 2007;Basbaum et al., 2009;Bourinet et al., 2016). The role of low voltage-activated (LVA) calcium Ca V 3.2 channels in pain sensation is well established; they contribute to nociception by lowering the threshold for action potential (AP) in DRG neurons (White et al., 1989;Cain and Snutch, 2010;Todorovic and Jevtovic-Todorovic, 2011). LVA channels (also called T-type) are comprised of Ca V 3.1, Ca V 3.2, and Ca V 3.3 channels; their biophysical properties such as the relatively small depolarization required for their activation, and window currents confer them the capability to act as burst firing modulators Perez-Reyes et al., 1998;Bourinet et al., 2016).
T-type currents are up-regulated in various models of chronic pain, such as chronic constriction injury, spinal nerve ligation, STZ-diabetes, the carregin pain model, and drug-induced diabetic neuropathy (Jagodic et al., 2007(Jagodic et al., , 2008Melrose et al., 2007;Takahashi et al., 2010;Marger et al., 2011;Watanabe et al., 2015;Li et al., 2017;Bellampalli et al., 2019). Ca V 3.2 can be upregulated by increased expression of USP5, which interacts with and de-ubiquitinates these channels thereby decreasing their degradation (Garcia-Caballero et al., 2014;Stemkowski et al., 2016). Moreover, treatment with T-type channel blockers results in reduced mechanical hyperalgesia in the spinal nerve ligation model (Dogrul et al., 2003;Chen et al., 2015). Similarly, manipulation of the expression levels of Ca V 3.2, as in the KO mouse model (Choi et al., 2007) or by selective knockdown in DRG neurons using antisense (Bourinet et al., 2005) results in attenuated pain responses, confirming the critical role of Ca V 3.2 in pain transmission.
Ca V 3.2 are therefore viable pharmacological targets to control pain (Dogrul et al., 2003;Flatters and Bennett, 2004;Okubo et al., 2011;Chen et al., 2015). An alternative strategy to modulate channel function is to target auxiliary or modulatory subunits to indirectly affect channel activity or trafficking to the plasma membrane (Weiss and Zamponi, 2019). This approach has been highly successful in the modulation of high-voltage-activated (HVA) channels by targeting of the α 2 δ subunit. This auxiliary subunit enhances Ca 2+ currents in part by modulating Ca V trafficking, altering their density and kinetics (Davies et al., 2007;Hendrich et al., 2008) and is one of the molecular targets of the antiepileptic and analgesic drug Gabapentin (GBP; Gee et al., 1996;Suárez et al., 2005;Martins et al., 2015). Consequently, treatment with GBP results in a significant decrease of N-type Ca V 2.2 currents due to a reduction of functional channels at the plasma membrane (Vega-Hernández and Felix, 2002;Hendrich et al., 2008).
Here, we targeted a protein that affects Ca V 3.2 trafficking to modulate its function and excitability in DRG neurons. Our target is Kelch-like 1 (KLHL1), a structural protein that binds to Ca V 3.2 and actin and alters Ca V 3.2 function. KLHL1's primary effect on Ca V 3.2 recycle it back to the plasma membrane via direct association with the channel and actin filaments, thus preventing its degradation; this process is mediated through increased recycling endosome-mediated channel insertion in the plasma membrane and results in an increased number of functional channels and ultimately increased Ca V 3.2-mediated T-type current density. KLHL1 also remains bound to Cav3.2 and F-actin at the plasma membrane, altering the channel kinetics of deactivation (Aromolaran et al., 2009(Aromolaran et al., , 2010(Aromolaran et al., , 2012. Here, we show that the expression levels of the structural protein KLHL1 can be altered to manipulate DRG neuron excitability and mechanical sensitivity in mice.

Cell Culture
DRG cultures were obtained as described (Gandini et al., 2014). In brief, DRG were dissected from C57BL/6 mice (P6-P10) in Advanced DMEM Medium (Gibco) supplemented with 20% of Fetal Bovine Serum (Gibco), washed, and digested for 40 min at 37 • C with a mixture of trypsin type XI (1.25 mg/ml, Sigma) and collagenase IV (1.25 mg/ml, Sigma), followed by mechanical dissociation. Cells were spun down at 1,000 g for 5 min at 10 • C and re-suspended in Advanced DMEM medium supplemented with 10% FBS. Cells were plated onto L-lysine-covered coverslips (12 mm, Carolina Biological Supply, Burlington, NC, USA) and kept in a 5% CO 2 humidified atmosphere at 37 • C. The Patchclamp recordings were made 24 h after dissociation (1 day in vitro, 1 DIV).
Immunoprecipitation. Crude membrane preparations were obtained using standard protocols (Aromolaran et al., 2010); a fraction of the sample was reserved prior to immunoprecipitation (input, 30× less concentrated than the IP samples) and the remaining volume was divided up for all experiments. Samples were processed by addition of primary antibodies (Ca V 3.2, 1:40 and KLHL1, 1:40, Santa Cruz Biotechnology, or IgG, 1:40, Alpha Diagnostic Intl. Inc., San Antonio, TX, USA) and incubated for 1-3 h at 4 • C followed by overnight incubation with protein A/G agarose beads (Biovision, Mountain View, CA, USA) on a shaking plate at 4 • C. Samples were washed and processed for western blot analysis as described.
Total currents were elicited using depolarizing steps (test potentials, V T ) from −60 to +60 mV (∆V = 10 mV) from a holding potential (V H ) of −90 mV. HVA currents were obtained from V H = −50 mV to V T = −60 to +60 mV (∆V = 10 mV). HVA currents traces were subtracted from the total current traces at each V H to obtain the LVA current component.

AAV Injections
All animal protocols used in this study were reviewed and approved by an independent Institutional Animal Care and Use Committee (IACUC). Hind paws of 13-weeks old WT male mice were injected with a control virus (EGFP-AAV, shCtrl,) or shKLHL1-AAV under blind conditions. The summary of the experimental conditions used is depicted in Figure 6A. The initial trial (n = 7) received 4.2 × 10 10 shKLHL1-AAV or 5.5 × 10 10 EGP-AAV vector genomes. The second trial (n = 11) received a high titer, 9.0 × 10 10 shKLHL1-AAV or EGFP-AAV vector genomes over 2 days. Viruses were diluted such that each individual injection volume was 5 µl total. Mice were given pain medication (Buprenorphine, 0.05 mg/kg, s.c.) for the first 2 days following the last injection and were allowed to recover in observation for 4-5 days while checked for any limp or lameness; all mice were confirmed healthy after injections.
Behavioral tests were performed twice a week (and averaged) for a total of 3 weeks after injections. Baseline withdrawal threshold responses were determined for 1 week before injections.
% Paw-withdrawal threshold was reported as the % of mice in the total population displaying withdrawal thresholds at all forces tested.

Von Frey Filament Tests
Hind paw withdrawal experiments were carried out in male mice ∼16 weeks old; animals had access to food and water ad libitum, all experiments followed IACUC-approved standard procedures. Only males were used in the first study because our preliminary data show male KLHL1 KO mice display a clear phenotype in contrast with females, where differences are more difficult to establish if present. Mice were placed on a wire mesh-bottom testing apparatus and allowed to acclimate for 15 min before assessing mechanical allodynia. Measurements were recorded by applying von Frey filaments (North Coast, Morgan Hill, CA, USA) ranging from 1.4 to 10 g to the plantar surface of the mouse hind paw; each filament was assessed for a total of five consecutive times. Hind paw withdrawal response times of less than 2 s were considered positive. The withdrawal threshold was calculated as the filament force at which each mouse had a positive response more than three times out of five (Chaplan et al., 1994;Bonin et al., 2014).

DRG Slices
Mice were anesthetized with isoflurane before euthanasia; DRG were excised and fixed with 4% paraformaldehyde for 4 h, cryoprotected overnight (30% sucrose in PBS), embedded in OCT (Tissue Tek, Fisher Scientific, Hampton, NH, USA), and frozen with dry ice. 30 µm-thick sections were cut using a cryostat and mounted onto Superfrost Plus slides (Fisher Scientific, Hampton, NH, USA). Slides were washed three times with PBS-glycine, dried and protected with coverslips. Fluorescence images were captured using IX80 Olympus inverted epifluorescence microscope using a 10× objective and analyzed using deconvolution.

Statistical Analysis
Statistical analysis was performed with SigmaPlot 11 Software. Statistical significance was determined as P < 0.05, using student's t-test or Kolmogorov-Smirnov non-parametric analysis (Kolmogorov-Smirnov test). Results are presented as mean ± SEM.

KLHL1 Is Expressed and Interacts With Ca V 3.2 in DRG Neurons
Co-immunoprecipitation of Ca V 3.2 and KLHL1 is detected in overexpression experiments in HEK-293 cells and in whole brain samples, demonstrating direct interaction between these FIGURE 1 | Ca V 3.2 and KLHL1 interact in mice dorsal root ganglion (DRG) ganglia. Co-immunoprecipitation of Ca V 3.2 and KLHL1 using antibodies against Ca V 3.2 (top) and KLHL1 (bottom). two proteins (Aromolaran et al., 2009(Aromolaran et al., , 2010(Aromolaran et al., , 2012. Figure 1 shows an example of pull-down experiments from DRG protein extracts using Ca V 3.2 (top) or KLHL1 antibodies (bottom, IgG was used as negative control). These data confirm the presence of KLHL1 in DRG neurons and its interaction with Ca V 3.2 T-type channels.

KLHL1 KO DRG Neurons Exhibit Low Ca V 3.2 Channel Expression and Reduced Excitability
We next assessed the effect of KLHL1 deletion on Ca V expression in DRG neurons. HVA channel expression was statistically similar in both KLHL1 KO and WT mice, as seen in Figure 2, which shows the KO/WT protein ratio for Ca V 2.1 (1.3 ± 0.4, n = 4) and Ca V 2.2 (1.0 ± 0.1, n = 3). In contrast, Ca V 3.2 expression was statistically lower among LVA channels (0.3 ± 0.09, n = 4) in the KLHL1 KO tissue (p = 0.04) whereas Ca V 3.1 and Ca V 3.3 expression remained constant (1.0 ± 0.2, n = 3; and 0.9 ± 0.06, n = 4 respectively). Thus, the absence of KLHL1 results in decreased Ca V 3.2 protein expression, which remains uncompensated for in the adult KLHL1 KO mice DRG.
To assess the physiological impact of lower Ca V 3.2 expression in the absence of KLHL1 we analyzed Ca 2+ current densities in DRG neurons. We found two neuronal populations according to the Ca 2+ currents they expressed ( Figure 3A): 56% of all WT neurons elicited only HVA currents (capacitance = 23.0 ± 3.1 pF, n = 14; gray symbols); the remaining cells (44%) expressed both HVA and LVA currents. The latter group had an average capacitance of 18 pF ± 2.4 pF (black symbols, n = 11); this value was not statistically different from HVA-only neurons (p = 0.1).
Overall the HVA-only population was identical between WT and KLHL1 KO neurons (56% and 53% of the total population, respectively), with current densities of 45.1 ± 5.9 and 50.1 ± 7.9 pA/pF (n = 10 and 13 (WT, KO); p = 0.1). Figure 3B depicts an example of a recording from a neuron displaying HVA currents-only. Figure 3C shows the I-V curves for WT (black circles) and KO HVA-only neurons (white circles; n = 10, 13; p = 0.2).
Neurons expressing LVA+HVA currents represented 44% of the total population in WT vs. 47% in KO neurons (p = 0.1). Figure 3D depicts an example of HVA+LVA currents; the rapidly inactivating LVA current component is noticeable at lower voltages. Figure 3E shows an example of current traces recorded at V T = −30 mV from V H = −90 and from −50 mV, respectively, which when subtracted yield the LVA current component (red trace). Figure 3F shows the LVA-only current I-V curves from WT and KO DRGs. The peak LVA current was −37.0 ± 3.6 pA/pF (n = 11) in WT compared to −23.0 ± 4.6 pA/pF (n = 10) in the KO (at −30 mV; p = 0.004). We studied small neurons with capacitances ranging from 18 to 28 pF to ensure only nociceptor neurons were analyzed (Andrade et al., 2010), given that D-Hair cells also display a high density of T-type currents (Dubreuil et al., 2004;Coste et al., 2007;Bernal Sierra et al., 2017), but their capacitance ranges from ∼39 to 65 pA (Coste et al., 2007). The impact of decreased Ca V 3.2 channel expression on DRG neuron excitability was assessed using current clamp experiments. Figure 4A shows representative traces of the APs elicited by a depolarizing current ramp delivered at 60 pA/s. The KLHL1 KO neuron rheobase was significantly larger (39.0 ± 4.1 pA) than WT (22.1 ± 2.9 pA; n = 9, 11, p = 0.03; Figure 4B), in line with a reduction in LVA calcium channel expression. This increase was accompanied by a concomitant reduction in action potential number (AP; 9.2 ± 1.1, n = 9) compared to WT (13.1 ± 1.4, n = 11, p = 0.03; Figure 4C). This rheobase difference was abolished by application of a low dose (100 nM) of NCC 55-0396 (NCC) to partially block T-type channels in WT neurons (27.2 ± 3.7 pA, n = 8, p = 0.003).

KLHL1 Mice Display Increased Mechanical Sensitivity Threshold
Our data shows that Ca V 3.2 expression is down-regulated in the absence of KLHL1; KO DRG neurons display significantly lower neuronal excitability, which may, in turn, alter pain sensation. We assessed the responses to mechanical stimulation by measuring paw withdrawal thresholds in WT and KLHL1 KO mice using von Frey filaments. KLHL1 KO mice displayed significantly higher withdrawal threshold (6.1 ± 0.2 g, n = 20) compared to WT mice (4.7 ± 0.3 g, n = 20; p = 0.009; Figure 5A). Non-parametric analysis of paw withdrawal threshold responses within the mice population demonstrates statistical differences at 4 and 6 g of force (p < 0.05; Figure 5B). Thus, decreased Ca V 3.2 expression in KLHL1 KO DRG neurons results in decreased excitability and altered pain sensitivity, confirming KLHL1 is a physiological modulator of Ca V 3.2 in sensory neurons.

Modulation of KLHL1 Expression Levels Alters Mechanical Sensitivity in WT Mice
Induction of excitability changes by the manipulation of KLHL1 levels could represent a novel method in the regulation of Cav3.2 expression. Therefore, we tested whether knockdown of KLHL1 expression alters mechanical sensitivity in WT mice by injecting adeno-associated viral particles (AAV) containing shRNA designed against KLHL1 into the mice hind paws (US Patent 10,047,377).
Preliminary data from neuronal cultures indicated that titers ∼5.0 × 10 10 shRNA-containing viral particles appeared to be less efficient in vitro. Therefore, we carried out two blind trials assessing the effect of two titer viral loads (4.2 × 10 10 vs. 9.0 × 10 10 viral particles; Figure 6A). All trials were performed following the timeline showed in Figure 6B and described in ''Materials and Methods'' section. Figure 6C shows representative images of DRG slices obtained from WT mice injected with EGFP-AAV or shKLHL1-AAV, confirming successful delivery and uptake of the AAV. Ca V 3.2 and KLHL1 levels from protein samples pooled from  three L4 DRGs ipsilateral to the shKLHL1-AAV-injected paw were analyzed by western blot (sh). L4 DRG ipsilateral to the EGFP-AAV injection were also collected as a negative control (Ctrl; Liu et al., 2019).
Further analysis is shown in Figure 8 where trial 2 data is shown as the percentage of the mice population displaying paw withdrawal threshold (PWT %) at a given Von Frey filament force value. There are no significant differences in  baseline values (A) or after the first week after injection (B) between the experimental and control-treated mice populations (Kolmogorov-Smirnov test). However, 69% of EGFP-AAV injected mice responded to the 4 g von Frey filament stimulus 2 weeks after injection (C) in comparison with only 34% in the shKLHL1-AAV injected population (p < 0.05). This difference was more pronounced after 3 weeks of injection (D), with changes between the two populations at 4 g (35% shKLHL1-AAV vs. 64% EGFP-AAV) and 2 g (0% shKLHL1-AAV vs. 24% EGFP-AAV). Note that all mice were more sensitive after injections as a result of the AAV injections (compare to untreated).

DISCUSSION
T-type Ca V 3.2 channel up-regulation is associated with diabetic neuropathy (Jagodic et al., 2007), neuropathic pain (Choi et al., 2016) and irritable bowel syndrome (Marger et al., 2011). Similarly, increased Ca V 3.2 function is found in chemotherapyinduced toxic neuropathies, and their inhibition with T-type calcium channels blockers decreases pain sensitivity (Flatters and Bennett, 2004;Okubo et al., 2011;Li et al., 2017). Also, paracetamol fails to induce analgesic effects in Ca V 3.2 knockout mice, suggesting these channels are necessary for analgesic actions (Kerckhove et al., 2014). KLHL1 protein is a constitutive modulator of Ca V 3.2 channels, and here we show that in its absence, KLHL1 KO mice elicit increased mechanical sensitivity threshold (decreased sensitivity to pain). Thus, KLHL1 could have significant potential as a molecular target to modulate neuropathic pain, akin to the effect of the auxiliary subunit α 2 δ's role on HVA channels (Field et al., 2006;Nguyen et al., 2009). KLHL1 functions in an analogous manner as the α 2 δ subunit, which is targeted by GBP and similar drugs resulting in a significant decrease of Ca V 2.2 currents in part by a reduction of functional channels at the plasma membrane (Vega-Hernández and Felix, 2002;Field et al., 2006;Hendrich et al., 2008;Aromolaran et al., 2009Aromolaran et al., , 2010Aromolaran et al., , 2012Martínez-Hernández et al., 2011).
KLHL1 KO neurons displayed decreased T-type calcium current density due to the down-regulation of Ca V 3.2, they also displayed decreased DRG neuron excitability, in line with the absence of KLHL1. Partial blockade of T-type channels with 100 nM NCC 55-0396 in WT neurons reduced their excitability to a comparable level to that of the KO DRGs, suggesting T-type channel down-regulation is solely responsible for the decrease in excitability.
Similar to other studies (Shin et al., 2003;Wang et al., 2015), we found Ca V 3.1 expression is detectable in DRG neurons from mice (in contrast to studies performed in rats, which report Ca V 3.1 is absent in DRG neurons from that species (Talley et al., 1999;Wen et al., 2006), however, it is well established that Ca V 3.1 does not have a functional role in DRGs, and accordingly, RNAseq data, Allen Atlas data and our own ICC data (not shown) demonstrate Ca V 3.1 expression in DRG is much lower in mice (if present) compared to Ca V 3.2. The fact that Ca V 3.1 is not physiologically relevant in DRG neurons possibly explains our observation that this channel type was not upregulated in KLHL1 KO DRG neurons, in contrast with our observations in hippocampal neurons from KLHL1 KO (Perissinotti et al., 2014). Similarly, KLHL1 also interacts with Ca V 2.1, and their levels were also unaffected in this system (in contrast to KLHL1 KO hippocampal neurons). The cause of this differential regulation is not known yet but may again be because Ca V 2.1 is not physiologically relevant in DRG neurons, thus suggesting tissue-specific modulatory mechanisms are tuned to control the expression of functionally relevant channels and their isoforms differently in central nervous system (CNS) vs. DRGneurons (Zamponi and Snutch, 2013).
Administration of T-type calcium channels blockers such as mibefradil or ethosuximide via intraperitoneal or paw injection, or chronic intrathecal infusion show reversal of neuropathic pain in rats (Dogrul et al., 2003;Chen et al., 2015); similarly, knockdown of Ca V 3.2 resulted in decreased analgesic effect (Bourinet et al., 2005(Bourinet et al., , 2016. Here, we were able to alter mechanical sensitivity in WT mice by knocking down KLHL1 levels in DRG neurons using anti-KLHL1 shRNA AAV injected into the hind paws of WT mice. We were able to sample the decreased expression of KLHL1 and Ca V 3.2 T-type channels by Western Blot analysis in a sample of three pooled L4 DRGs injected with shKLHL1 AAV compared to an EGFP AAV-injected DRGs sample. von Frey filament tests confirmed that mice whose hind paws were injected with shKLHL1 were less sensitive to pain than those injected with control EGFP-AAV (5.6 ± 0.5 g threshold compared to 3.5 ± 0.4 g, respectively at week 3 of treatment). These values are in line with the majority of data in the literature (Watanabe et al., 2015;Garcia-Caballero et al., 2016;M'Dahoma et al., 2016;Stemkowski et al., 2016;Ogawa et al., 2018); however, two groups have reported withdrawal threshold values around 1 g (Costigan et al., 2009;Chiu et al., 2013;Vicuna et al., 2015;Choi et al., 2016). It is no clear the reason for these differences, given that most studies were done using C57B/6 mice (6-14 weeks old).
In vitro data from neuronal cultures indicated that viral titers of less than 5.0 × 10 10 particles would be less efficient, here we found we found that in vivo delivery of 5.1 × 10 10 shKLHL1-AAV vector genomes was the minimal titer that exerted an effect, although it was only attained after 3 weeks post-treatment. In contrast, delivery of 9.0 × 10 10 vector genomes of virus-containing shKLHL1 sequences was efficacious at increasing the mice's withdrawal threshold, demonstrating reduced sensitivity to pain. Alternative delivery routes such as subcutaneous or intramuscular injections are also known as viable options, and can be assessed in the future (Towne et al., 2009).
In summary, our study shows that KLHL1 is a physiological modulator Ca V 3.2 expression and function in DRG neurons and that KLHL1 may be a viable molecular target to reduce pain transmission by lowering Ca V 3.2 expression. Modulation of neuronal excitability by alteration of KLHL1 levels and/or function may represent a novel method of treatment for neuropathic disorders and may help facilitate the development of novel therapeutic alternatives.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation, to any qualified researcher.

ETHICS STATEMENT
All animal studies presented in this study were reviewed and approved by an independent Institutional Animal Care and Use Committee (IACUC) at Loyola University Chicago Stritch School of Medicine.