Cell-Type Specific Channelopathies in the Prefrontal Cortex of the fmr1-/y Mouse Model of Fragile X Syndrome,,

Abstract Fragile X syndrome (FXS) is caused by transcriptional silencing of the fmr1 gene resulting in the loss of fragile X mental retardation protein (FMRP) expression. FXS patients display several behavioral phenotypes associated with prefrontal cortex (PFC) dysfunction. Voltage-gated ion channels, some of which are regulated by FMRP, heavily influence PFC neuron function. Although there is evidence for brain region-specific alterations to the function a single type of ion channel in FXS, it is unclear whether subtypes of principal neurons within a brain region are affected uniformly. We tested for alterations to ion channels critical in regulating neural excitability in two subtypes of prefrontal L5 pyramidal neurons. Using somatic and dendritic patch-clamp recordings, we provide evidence that the functional expression of h-channels (Ih) is down-regulated, whereas A-type K+ channel function is up-regulated in pyramidal tract-projecting (PT) neurons in the fmr1-/y mouse PFC. This is the opposite pattern of results from published findings from hippocampus where Ih is up-regulated and A-type K+ channel function is down-regulated. Additionally, we find that somatic Kv1-mediated current is down-regulated, resulting in increased excitability of fmr1-/y PT neurons. Importantly, these h- and K+ channel differences do not extend to neighboring intratelencephalic-projecting neurons. Thus, the absence of FMRP has divergent effects on the function of individual types of ion channels not only between brain regions, but also variable effects across cell types within the same brain region. Given the importance of ion channels in regulating neural circuits, these results suggest cell-type-specific phenotypes for the disease.


Introduction
Fragile X syndrome (FXS) is the most common form of inherited mental retardation and the leading identified monogenic cause of autism. Fragile X mental retardation protein (FMRP), the protein absent in FXS, regulates neuron function via multiple mechanisms, including proteinprotein interactions, translational control, and protein trafficking (Hagerman et al., 2005;Bhakar et al., 2012;Santoro et al., 2012;Brager and Johnston, 2014). FMRP is capable of binding to mRNA for several ion channels and is known to interact with ion channel-associated proteins (Brown et al., 2010;Darnell et al., 2011;Brager and Johnston, 2014). Although there is evidence for brain regionspecific alterations to channel function in FXS, (for review, see Contractor et al., 2015), it is unclear whether these effects are uniform across neuron types within a brain region. Here, we provide evidence for neuron-typespecific alterations to ion channel function within the prefrontal cortex (PFC) of fmr1-/y mouse.
Neuronal dysfunction in the fmr1-/y mouse can be attributed to differences in several types of ion channels (Meredith et al., 2007;Brager et al., 2012;Deng et al., 2013;Routh et al., 2013;Zhang et al., 2014). In some cases, reports are conflicting as to the effect of the loss of FMRP on a given type of ion channel. Some have reported an increase (Lee et al., 2011) in Kv4/A-type K ϩ channel expression, whereas others have found a decrease (Gross et al., 2011;Routh et al., 2013). Similarly, h-channels have been reported to be upregulated in hippocampus (Brager et al., 2012), but downregulated in L5B neurons of somatosensory cortex .
We tested the hypothesis that the absence of FMRP has different effects on the function of a single type of ion channel in PFC, a brain region containing a heterogeneous neuron population and that is implicated in behavioral deficits associated with FXS (Menon et al., 2004;Krueger et al., 2011;Wang et al., 2012;Dembrow and Johnston, 2014). Using current-clamp and outside-out patch-clamp recordings from fmr1-/y mice we provide evidence that I h (the current mediated by h-channels) is down-regulated in pyramidal tract (PT)-projecting neurons. This is in contrast to hippocampal CA1 pyramidal neurons, but similar to L5 neurons in somatosensory cortex. Using outside-out patch-clamp recordings, we measured three distinct types of potassium currents: rapidly inactivating A-type current, mediated by putative Kv4 channels; a slowly inactivating current, mediated by putative Kv1 channels; and a sustained current. Again, in contrast to hippocampus, we find that the maximum putative Kv4-mediated current is increased in PT neurons. Additionally, the maximum putative Kv1-mediated current is decreased at the soma of PT neurons, resulting in increased excitability. Notably, these somatic K ϩ channel differences do not extend to neighboring intra-telencephalic-projecting (IT) neurons within L5. Thus, the absence of FMRP has divergent effects on a given type of ion channel in different cell types, even within a single brain region. These findings highlight the need to understand FXS and its treatment in a neuron type/brain regionspecific manner.

Bead infusions
All procedures involving animals were approved by the Institutional Animal Care and Use Committee. Male wildtype (WT; C57BL/6) and fmr1-/y mice were anesthetized with isoflurane (1 Ϫ 4% mixed in oxygen) and prepared for stereotaxic injection of retrograde transported fluorescent-labeled microspheres (Lumafluor). Beads (100 nl) were injected into the pons (in millimeters relative to bregma: posterior 4.2; lateral 0.4; ventral 4.5), mPFC (anterior 1.0 -1.5; lateral 0.45; ventral 1-2), or striatum (posterior 0.4 -1.0; lateral 2.25; ventral 3.5), using a glass pipette (ϳ15 m diameter tip) connected to a nanoject II auto-nanoliter injector (Drummond Scientific) at a rate of 23 nl/s. For all injections, the pipette was left in place for 5 min before removing it from the brain. Mice were given analgesics (Carprofen; 5 mg/kg) and recovered for at least 2 d before their use in experiments. Mice that received different colored beads to label IT and PT neurons were anesthetized with a ketamine (100 mg/kg)/xylazine (10 mg/kg) mixture and were perfused through the heart with ice-cold saline consisting of the following (in mM): 2.5 KCl, 1.25 NaH 2 PO 4 , 25 NaHCO 3 , 0.5 CaCl 2 , 7 MgCl 2 , 7 dextrose, 205 sucrose, 1.3 ascorbate, and 3 sodium pyruvate (bubbled with 95% O 2 /5% CO 2 to maintain pH ϳ7.4) followed by 4% paraformaldehyde in 0.1 M phosphate buffer. Three hundred-micrometer-thick coronal slices were visualized with a Zeiss Axio Imager Z2 microscope running Axio Vision software (Carl Zeiss). We looked for clear instances of overlap of green and red puncta across at least four 300 m sections per animal. 34°C) recording chamber that was continually perfused (1 Ϫ 2 ml/min) with bubbled aCSF containing the following (in mM): 125 NaCl, 3.0 KCl, 1.25 NaH 2 PO 4 , 25 NaHCO 3 , 2 CaCl 2 , 1 MgCl 2 , 10 dextrose, 3 sodium pyruvate, 0.025 D-APV, 0.02 DNQX, 0.005 CGP, and 0.002 gabazine. For outside-out recordings, 0.001 mM TTX was added to the aCSF. Slices were viewed with either: (1) a Zeiss Axioskop microscope and differential interference optics, (2) a Zeiss AxioExaminer D microscope and Dodt contrast optics, or (3) a two-photon laser-scanning microscope (Leica SP5 RS) using Dodt contrast. Fluorescentlabeled neurons were visualized using a mercury lamp and a 540/605 nm or 470/502 nm excitation/emission filter set or two-photon excitation at 840 nm. Patch pipettes (4 -8 M⍀) were pulled from borosilicate glass and wrapped with Parafilm to reduce capacitance. The pipette solution for all configurations contained the following (in mM): 120 K-gluconate, 16 KCl, 10 HEPES, 8 NaCl, 7 K 2 phosphocreatine, 0.3 Na-GTP, 4 Mg-ATP, pH 7.3 with KOH. Neurobiotin (Vector Laboratories; 0.1%) was also included for histologic processing. For all experiments involving dendritic recordings and some experiments involving somatic recordings, AlexaFluor 594 (16 M; Invitrogen) was also included in the internal recording solution to determine the recording location. All drugs were prepared from concentrated stock solutions in water and were obtained from Abcam Pharmaceutical or Tocris.

Whole-cell recordings
Data were acquired using a Dagan BVC-700 (Dagan) amplifier and custom data acquisition software written using Igor Pro (Wavemetrics) or AxoGraph X (AxoGraph Scientific) data acquisition software. Data were acquired at 10 Ϫ 50 kHz, filtered at 5 Ϫ 10 kHz, and digitized by an ITC-18 (InstruTech) interface. Pipette capacitance was compensated and the bridge was balanced during each recording. Series resistance was monitored throughout each experiment and was 10 Ϫ 25 M⍀ for somatic recordings and 15 Ϫ 40 M⍀ for dendritic recordings. Voltages are not corrected for the liquid-junction potential (estimated as ϳ12 mV).
Data were analyzed using either custom analysis software written in Igor Pro or using AxoGraph X. Subthreshold membrane properties were measured at a common membrane potential (Ϫ65 mV). Input resistance (R N ) sag and rebound were calculated from the voltage response to a family of 1000 ms current injections (Ϫ150 to ϩ50 pA, 20 pA steps. Input resistance (R N ) was calculated from the linear portion of the currentϪvoltage relationship. Voltage sag was defined as the ratio of maximum to steady-state R N . Rebound slope was calculated from the slope of the rebound amplitude as a function of steady-state membrane potential. The functional membrane time constant was defined as the slow component of a doubleexponential fit of the average voltage decay in response to hyperpolarizing current injections (100 Ϫ 300 pA, 2 ms). Resonance was determined from the voltage response to a constant amplitude sinusoidal current injection that linearly increased in frequency from 1 Ϫ 15 Hz in 15 s. The impedance amplitude profile (ZAP) was constructed from the ratio of the fast Fourier transform of the voltage response to the fast Fourier transform of the current injection. The peak of the ZAP was defined as the resonant frequency. Single action potentials (APs) were elicited using just-threshold current injections of various durations. AP threshold was defined as the voltage where the first derivative first exceeded 20 mV/ms. Simulated synaptic currents were generated with exponentially rising (dendrite ϭ 0.2 ms; soma ϭ 0.3 ms) and decaying (dendrite ϭ 2 ms; soma ϭ 4 ms) waveforms. The amplitude of this waveform was adjusted such that amplitude of the voltage response to a single simulated event was ϳ3 mV. For recordings from non-labeled neurons, the presence (Ͼ2.2 Hz) or absence of resonance was used to classify neurons into projection types.

Outside-out recordings
Membrane currents were recorded using an Axopach 200B amplifier (Molecular Devices), sampled at 10 kHz, analog filtered at 2 kHz and digitized by an ITC-18 interface connected to a computer running Axograph X. For K ϩ channels, activation curves were constructed by using depolarizing voltage commands (Ϫ70 to 50 mV in 20 mV steps) to activate I K from a holding potential of Ϫ90 mV. Activation data were fit to a single Boltzmann function using a least-squares program. For voltage commands more hyperpolarized than Ϫ10 mV, we extrapolated the slow time constant of the current back to the time equal to the peak of the total current. In this way, the peak of the slow and fast inactivating currents could be separated. The accuracy of this procedure was confirmed using a small set of experiments where the slowly inactivating component was measured directly. Linear leakage and capacitive currents were digitally subtracted by scaling traces at smaller command voltages in which no voltagedependent current was activated.

Data analysis
Repeated-measures (RM) ANOVA, between-subject factors ANOVA, mixed factors ANOVA, and post hoc t tests were used to test for statistical differences between experimental conditions. Bonferroni correction was used to correct for multiple comparisons. Pearson's product moment correlation was used to test for statistically significant correlations between variables. Error bars represent SEM. Statistical analyses were performed Prism (Graphpad). Data are presented in the text as mean Ϯ SEM.
Before testing for alterations to physiology, we first asked whether the two general types of L5 projection neurons are found in the fmr1-/y mouse medial PFC (mPFC). In WT and fmr1-/y mice, we infused red retrograde tracer (Lumuflour beads) into the pontine nuclei to label PT neurons and a green tracer into either the contralateral striatum or contralateral mPFC to label IT neurons. In both genotypes, IT neurons were found throughout L2-6, whereas PT neurons were restricted to L5/6. Within the deep layers, PT and IT neurons could be observed within close proximity to each other at all depths (Ͻ20 m). Furthermore, we did not observe clearly any double-labeled neurons in either genotype (n ϭ 2 for each genotype; Fig. 1A). Together, these data suggest that in fmr1-/y mice, as in WT mice, IT and PT neurons represent two non-overlapping populations of neurons.
As an additional test, we next asked whether L5 IT and PT neurons in fmr1-/y mice could be distinguished based upon their physiologic properties. PT neurons possess conductances that endow them with membrane resonance in the 3-7 Hz range whereas IT neurons are nonresonant. Thus, the presence of resonance can be used to distinguish PT from IT neurons (Dembrow et al., 2010). For these experiments, we made whole-cell current-clamp recordings from the soma of identified L5 PT or IT neurons, as well as nearby unlabeled cells. In both WT (n ϭ 15 cells from 6 mice, 4.11 Ϯ 0.32 Hz) and fmr1-/y (n ϭ 21 cells from 7 mice, 2.79 Ϯ 0.19 Hz) mice, PT neurons displayed membrane resonance whereas IT neurons did not (WT: n ϭ 17 cells from 5 mice, 1.07 Ϯ 0.06 Hz; fmr1-/y: n ϭ 19 cells from 6 mice, 1.01 Ϯ 0.01 Hz; IT vs PT, p Ͻ 0.001 a , ANOVA; Fig. 1B,C; Table 1). Furthermore, in both genotypes, input resistance (R N ) was significantly higher in PT compared with IT neurons (WT PT ϭ 79.53 Ϯ 6.65 M⍀; WT IT ϭ 112.90 Ϯ 11.78 M⍀; fmr1-/y PT ϭ 89.04 Ϯ 5.21 M⍀, fmr1-/y IT 171.3 Ϯ 8.73 M⍀; p Ͻ 0.001 b , ANOVA; Fig. 1B,C). We also recorded from nonresonant (WT: n ϭ 3 cells from 2 mice; fmr1-/y: n ϭ 3 from 2 mice) and resonant (WT: n ϭ 13 from 4 mice; fmr1-/y: n ϭ 11 cells from 3 mice) non-labeled neurons with physiologic properties that closely resembled nearby labeled IT and PT neurons, respectively. When these putative IT and PT neurons were grouped with labeled projection neurons for analysis, we found that in both genotypes, PT neuron resonant frequency was inversely correlated with R N , whereas in IT neurons it was not ( Fig. 1C; WT PT: r 2 ϭ 0.22, p ϭ 0.01 c ; WT IT: r 2 ϭ 0.01, p ϭ 0.62; fmr1-/y PT: r 2 ϭ 0.12, p ϭ 0.05, fmr1-/y IT: r 2 ϭ 0.08, p ϭ 0.18). These data are consistent with findings in rat mPFC (Dembrow et al., 2010) and suggest that the organization of projection neurons in L5 of PFC is maintained in the fmr1-/y mouse. For our purposes, unlabeled neurons that display resonance will be referred to as PT neurons, whereas those that do not will be referred to as IT neurons.

I h is reduced in L5 neurons in the fmr1-/y mouse mPFC
The intrinsic membrane properties of pyramidal neurons are heavily influenced by the presence of voltage-gated ion channels (Magee, 2000;London and Häusser, 2005;Johnston and Narayanan, 2008). The expression of one class of channel, h-channels, which carry I h , has been reported to be up-regulated in CA1 but down-regulated in somatosensory cortex in the fmr1-/y mouse (Brager et al., 2012;Zhang et al., 2014).
To test for alterations to h-channel function in PFC we measured I h -sensitive membrane properties (membrane potential -V m , R N , time constant -tau, sag, rebound, resonance, and temporal summation; Hutcheon and Yarom, 2000;Robinson and Siegelbaum, 2003;Brager and Johnston, 2007;Narayanan and Johnston, 2007) in WT and fmr1-/y neurons. Because h-channel expression in many types of cortical pyramidal neurons increases with distance from the soma (Lörincz et al., 2002;Kole et al., 2006), we performed these measurements at the soma (WT: n ϭ 30 cells from 18 mice; fmr1-/y: n ϭ 35 cells; from 17 mice), proximal dendrite (Յ250 m; WT: n ϭ 12 cells from 9 mice; fmr1-/y: n ϭ 17 cells from 17 mice), and distal dendrite (Ͼ250 m; WT: n ϭ 34 cells from 26 mice; fmr1-/y: n ϭ 23 cells from 17 mice). We focused our initial efforts on PT neurons because they have more pronounced h-channel associated properties ( Fig. 1; Dembrow et al., 2010;Sheets et al., 2011;Kalmbach et al., 2013). R N was higher at proximal and distal dendritic recording locations (p Ͻ 0.01 d ) in fmr1-/y compared with WT neurons, but this difference was not observed at the soma (p ϭ 0.21; Fig. 2A,B). Fmr1-/y neurons were more hyperpolarized, displayed less sag and less rebound than WT neurons, at all recording locations (all p Ͻ 0.01; Fig. 2A,B e ). Furthermore, the resonant frequency of fmr1-/y neurons was lower and the functional membrane time constant longer than WT neurons at all recording locations (p Ͻ 0.01; Fig. 2C,D f ). Finally, fmr1-/y neurons displayed greater temporal summation of simulated synaptic currents at both the soma and distal dendritic recording locations (p Ͻ 0.001, ANOVA; Fig. 2E g ). Together, these differences in membrane properties between genotypes suggest that there is a decrease in I h at both the soma and dendrite of L5 PT neurons in PFC of fmr1-/y mice.
To test for differences in IT neurons, we made many of these same measurements at the soma (WT: n ϭ 20 cells from 8 mice; fmr1-/y: n ϭ 22 cells from 10 mice) and dendrite ( Ϫ72.15 Ϯ 1.48; p ϭ 0.04; ANOVA j ), suggesting that h-channels were not the main source of these differences.
To test whether the observed differences in PT neurons were due primarily to h-channels, we compared the ef-  fects of the ZD7288 on the properties of PT L5 dendrites in WT versus fmr1-/y mice. Application of 10 M ZD7288 largely eliminated the differences in membrane properties between WT (n ϭ 7 cells from 6 mice) and fmr1-/y (n ϭ 9 cells from 7 mice) neurons measured at a common membrane potential (Ϫ65 mV; Fig. 4). There were no statistical differences in R N (p ϭ 0.12), V m (p ϭ 0.94), sag (p ϭ 0.40), rebound (p ϭ 0.08; Fig. 4A Finally, we made outside-out patch-clamp recordings from the distal dendrite of PT neurons to directly compare I h in WT (n ϭ 4 cells from 3 mice; 225-400 m, 340 Ϯ 37.6 m from soma) and fmr1-/y (n ϭ 5 cells from 5 mice; 250 -350, 308 Ϯ 17.7 m from soma) mice. Voltage steps from a holding potential of Ϫ30 to Ϫ140 mV elicited an inward current in both genotypes (Fig. 5A). The activation of this current was well fit by a double-exponential. The  . The maximum dendritic I h was greater in WT compared with fmr1-/y neurons ( Fig. 5C; p ϭ 0.001, t test m ). Finally, the resonant frequency measured during whole-cell before obtaining the outside-out configuration was lower in fmr1-/y compared with WT dendrites ( Fig. 5D; p ϭ 0.03, t test n ). These data are consistent with a decrease in the functional expression of h-channels in L5 PT neurons of PFC in the fmr1-/y mouse.

Alterations to I K in L5 mPFC pyramidal neurons in fmr1-/y mice
The observed decrease in I h in PT neurons is in contrast to pyramidal neurons in CA1 of the hippocampus (Brager et al., 2012), but is similar to L5 pyramidal neurons in somatosensory cortex . Together, these observations suggest that the absence of FMRP can have divergent effects on channel expression/function depending upon brain region or neuron type. A-type K ϩ and h-channel function are affected in opposite directions by several manipulations (Harris-Warrick et al., 1995;Bernard et al., 2004;Frick et al., 2004;Fan et al., 2005;Shin et al., 2008). Thus, we next asked whether the functional expression of A-type K ϩ channels, which is decreased in CA1 pyramidal neurons in the fmr1-/y mouse (Routh et al., 2013), is increased in L5 neurons of mPFC and if so, whether the change is restricted to one class of L5 projection neuron.
We used outside-out patch recordings to measure the total K ϩ current at the soma and distal dendrite (Ն250 m) of PT L5 neurons in WT and fmr1-/y mice (Fig. 6A). The current elicited by step depolarization to 50 mV from a holding potential of Ϫ90 mV contained both a transient and a sustained component (I K-total ; Fig. 6A1). Macroscopic K ϩ currents in L5 neurons in granular cortex are composed of distinct channels with fast and slow inactivation kinetics (Bekkers, 2000;Korngreen and Sakmann, 2000). To test whether PT neurons in mPFC possess similar diversity in K ϩ currents, we used the voltagedependent properties of the putative channels underlying each of these components to isolate the individual currents. A brief (100 ms) prepulse to Ϫ20mV was used to inactivate the fast-inactivating current without significantly affecting either the sustained or slowly inactivating current (Fig. 6A2). Subtracting the resulting current from total I K revealed a rapidly inactivating K ϩ current (I KA-fast ; Fig. 6A4).
Step depolarization to 50 mV from a holding potential of Ϫ20 mV, to inactivate all transient K ϩ currents, revealed a sustained current (I K-sustained ; Fig. 6A3). Subtracting I K-sustained from the current in Fig. 6A2 revealed a slowly inactivating K ϩ current (I K-slow , Fig. 6A5). Thus, we were able to isolate three kinetically distinct outward currents: I K-sustained , I KA-fast , and I K-slow (Fig. 6A).
The maximum I K-total was greater at the soma compared with the dendrite for WT and fmr1-/y neurons (p Ͻ 0.001°, ANOVA). However, the maximum somatic I K-total in fmr1-/y neurons was larger than that in WT neurons (p ϭ 0.01; Fig.  6B p ). This difference in somatic I K-total is due, in part, to differences in both I KA-fast and I K-sustained . We observed larger I KA-fast and I K-sustained at the soma (p ϭ 0.05; fast, p ϭ 0.05 q ) but not dendrite (sustained, p ϭ 0.74; fast, p ϭ 0.52 r ), of fmr1-/y compared with WT neurons (Fig. 6C,E). Consequently, I KA-fast (p Ͻ 0.01) and I K-sustained (p Ͻ 0.05) were larger at the soma compared with the dendrite in fmr1-/y, but not WT neurons (Fig. 6C,E s ). Interestingly, we also observed genotypic differences for the I K-slow . I K-slow was smaller at the dendrite compared with the soma in both genotypes (p Ͻ 0.05 t ). In contrast to the other currents however, I K-slow was smaller at the soma of fmr1-/y compared with WT neurons (p Ͻ 0.05; Fig. 6D u ). In summary, at the soma of fmr1-/y PT neurons, there is a relative increase in the contribution of I KA-fast (57 Ϯ 3% vs 44 Ϯ 7%) and decrease in the contribution of I K-slow (50 Ϯ 6% vs 36 Ϯ 3%) to the somatic I K-total compared with WT (Fig. 6F). In contrast, we observed no differences between genotypes in relative contribution of the three K ϩ current components in the dendrites (Fig. 6G).
Relatively little is known about differences in K ϩ channel expression at the soma of different projection neurons in L5. Therefore, we also made outside-out patch record-ings from IT neurons in mPFC to test whether the change in magnitude and relative contribution of K ϩ currents observed at the soma of PT neurons is common to all pyramidal neurons within a brain region. Similar to PT neurons, the current elicited by step depolarization to 50 mV from a holding potential of Ϫ90 mV contained both transient and sustained components. In contrast to PT neurons however, we did not observe differences in any I K component between fmr1-/y and WT IT neurons at the soma or dendrite ( Fig. 7; p Ͼ 0.05, all comparisons v ). Thus, in the fmr1-/y mouse, the maximum amplitude of K ϩ currents at the soma is affected in one type of projection neuron (PT), but not a neighboring type (IT). These data suggest that channel phenotypes in the fmr1-/y mouse can vary not only between brain regions (ie, hippocampus, somatosensory cortex and mPFC), but also within a brain region in different classes of principal neurons (PT vs IT).
A-type K ϩ channels can be composed of either Kv4 or Kv1 subunits (Coetzee et al., 1999). To identify the putative channels underlying I KA-fast and I K-slow , we used two complementary approaches. First, we exploited differ- ences in the kinetic properties of these channels. Kv1 containing channels tend to inactivate and recover from inactivation an order of a magnitude slower than Kv4 containing channels (Castellino et al., 1995;Jerng et al., 2004). The time constant of inactivation ( inactivation ) of the I KA-fast was not different between WT and fmr1-/y for either PT (WT inactivation ϭ 29 ms; fmr1-/y inactivation ϭ 25 ms) or IT neurons (WT inactivation ϭ 28 ms; fmr1-/y inactivation ϭ 23 ms; p ϭ 0.88, ANOVA w ). These values were consistent with Kv4 channels. The inactivation of the I K-slow was an order of magnitude larger than the inactivation of the I KA-fast in both WT and fmr1-/y for PT and IT neurons, consistent with Kv1 channels. A closer comparison revealed that the inactivation of the I K-slow in fmr1-/y PT neurons was significantly smaller compared with WT PT neurons (WT inactivation ϭ 646 Ϯ 95 ms; fmr1-/y inactivation ϭ 244 Ϯ 44 ms; p ϭ 0.01 x ). By contrast, there was no statistically significant difference in inactivation for I K-slow in IT neurons (WT inactivation ϭ 630 Ϯ 100 ms; fmr1-/y inactivation ϭ 425 Ϯ 69 ms; p Ͼ 0.05, Fig. 8A y ). Next, we used a two-pulse protocol to measure the recovery from inactivation. We found that the time course of recovery from inactivation was well fit by two exponentials consistent with two types of transient K ϩ channels. Both the fast (p ϭ 0.2, ANOVA z ) and slow (p ϭ 0.5, ANOVA a= ) time constant of recovery from inactivation were similar in WT and fmr1-/y for L5 PT and IT neurons and were consistent with Kv4-containing and Kv1-containing channels, respectively ( Fig. 8B-D).
As an additional method of identifying the channels underlying these currents, we exploited differences in the sensitivity of Kv1 and Kv4 containing channels to K ϩ channel blockers. Specifically, we measured the sensitivity of macroscopic K ϩ currents to 50 M 4-aminopyridine (4-AP) and 150 M Ba 2ϩ , which preferentially block K V 1-containing and K V 4-containing channels, respectively (Castle et al., 1994;Russell et al., 1994;Stephens et al., 1994;Coetzee et al., 1999;Gasparini et al., 2007). Additionally, we measured the sensitivity of these currents to alpha-dendrotoxin (␣-DTX), which preferentially blocks Kv1.1-, 1.2-, and 1.6containing channels (Harvey, 2001). 4-AP (50 M) and ␣-DTX (200 nM) decreased I KA-slow without significantly affecting I KA-fast (p Ͻ 0.01, n ϭ 8 patches from 4 mice and p Ͻ 0.001, n ϭ 5 patches from 4 mice, respectively; two-way RM ANOVA; Fig. 9A,B b= ). In contrast, 150 M Ba 2ϩ decreased I KA -fast without significantly affecting I KA-slow (p Ͻ 0.005, n ϭ 5 patches from 3 mice; two-way RM ANOVA; Fig. 9C c= ). Furthermore, the amplitude of the DTX-and 4-AP-sensitive current was similar to the amplitude of I KA-slow , whereas the amplitude of the Ba 2ϩ -sensitive current was similar to the amplitude of I KA-fast (Fig. 9D). Together, these data suggest that Kv4 containing channels contribute to I KA-fast and Kv1 containing channels contribute to I K-slow . Thus, our data suggest that in the fmr1-/y mouse, there are alterations to the functional expression of two K ϩ channels in PT L5 neurons of mPFC; there is an increase in the expres- In addition to a decrease in channel expression, there is a hyperpolarizing shift in the voltage dependence of activation of Kv4 channels in CA1 pyramidal neuron dendrites of fmr1-/y mice (Routh et al., 2013). Thus, although there is a decrease in total maximum current, there are relatively more channels open at hyperpolarized potentials in fmr1-/y mice compared with WT mice. To test for similar alterations to channel function in L5 neurons of mPFC, we measured the voltage dependence of activation of fast A-type K ϩ channels in somatic and dendritic patches from WT and fmr1-/y mice (Fig. 8E). Consistent with published reports in granular cortex (Bekkers, 2000), there was a hyperpolarizing shift in the activation curve of I KA-fast at the dendrite compared with the soma in WT L5 PT neurons (p ϭ 0.03; Fig. 8F,H d= ). In contrast, there was no such shift in fmr1-/y neurons (p ϭ 0.76; Fig. 8G,H e= ). Furthermore, there was no difference in slope factor between WT and fmr1-/y neurons at either recording location (p Ͼ 0.3 f= , Figure 6. Differences in I K in PT neurons in fmr1-/y versus WT mice. A, Outside-out patch from a PT neuron visualized through fluorescence microscopy. I K-total was measured by stepping to ϩ50 mV from a holding potential of Ϫ90mV. A prepulse to Ϫ20 mV was used to inactivate fast inactivating I KA . I K-sustained was obtained by stepping from a holding potential of Ϫ20 mV to ϩ50 mV. I KA-fast was obtained by subtracting the current in A2, from the total current in A1. I K-slow was obtained by subtracting the sustained current in A3 from the current obtained in A2. B, Maximum I K-total at the soma is greater in fmr1-/y PT neurons compared with WT neurons. C, Maximum I KA-fast at the soma is greater in fmr1-/y neurons compared with WT. D, Maximum I K-slow at the soma of fmr1/y neurons is less compared with WT neurons. E, At the soma, maximum I K-sustained is greater in fmr1-/y PT neurons compared with WT. F, Summary of somatic differences in I K between WT and fmr1-/y neurons. G, There were no differences in dendritic K ϩ currents in fmr1-/y compared with WT neurons. ANOVA). Thus, the hyperpolarizing shift in the activation of I KA-fast normally observed from soma to dendrite in WT neurons is absent in fmr1-/y neurons. These data suggest that although there is no difference in maximum I KA-fast in the dendrites of fmr1-/y versus WT PT neurons, there may subtle differences in voltage dependence of activation.
Altered action potential generation in fmr1-/y L5 PT neurons of mPFC K ϩ channels modulate several aspects of neural function, including the dynamics of action potential threshold. Changes to threshold, also known as threshold accommodation, can be mediated through the recruitment of K ϩ channels, in addition to Na ϩ channel inactivation (Hodgkin and Huxley, 1952). Kv4 and Kv1 subunits have both been implicated in regulating neural excitability by setting the threshold for action potential generation (Bekkers and Delaney, 2001;Kim et al., 2005;Higgs and Spain, 2011;Carrasquillo et al., 2012). Because we observed somatic differences consistent with alterations to these two K ϩ channels, we predicted that threshold dynamics would be altered in L5 PT neurons in mPFC.
To test this hypothesis, we injected current of varied duration at the soma of L5 PT neurons in both genotypes (WT: n ϭ 20 cells from 11 mice; fmr1-/y: n ϭ 23 cells from 9 mice). The current amplitude for a given duration was adjusted to be just-threshold for the generation of an action potential (Fig. 10A). Action potential threshold varied as a function of current duration in both genotypes (p Ͻ 0.001, ANOVA g= ). However, for durations greater than 6 ms, spike threshold was significantly more hyperpolarized in fmr1-/y PT neurons compared with WT neurons (p Ͻ 0.01; post hoc comparisons h= ; Fig. 10A-C). Furthermore, the current required to generate an action potential was significantly lower in fmr1-/y compared with WT neurons at all current durations (p Ͻ 0.001, ANOVA i= ). Differences in Na ϩ channel availability can produce differences in action potential threshold (Hodgkin and Huxley, 1952). However, there was no significant difference in the maximum rate of rise of action potentials in WT versus fmr1-/y neurons, a spike parameter sensitive to Na ϩ channel availability (p ϭ 0.30; ANOVA j= ; Colbert et al., 1997). There was also no difference in action potential amplitude between genotypes (p ϭ 0.59; ANOVA). Finally, there were no differences in action potential threshold between WT and fmr1-/y IT neurons (p ϭ 0.31; ANOVA; Fig. 10D k= ). F, In WT neurons, the steady-state activation curve is shifted to hyperpolarized potentials in the dendrite relative to the soma. G, In contrast, there is no difference in the activation curves between the soma and dendrite in fmr1-/y neurons. H, the V 1/2 of activation is hyperpolarized in the dendrite compared with the soma in WT neurons. In contrast, there is no difference in the V 1/2 of activation between the soma and dendrite in fmr1-/y neurons. Figure 9. Sensitivity of K ؉ currents to K ؉ channel blockers. A, Sample sweeps recorded from the same patch comparing the ␣-DTX-sensitive current to I K-slow . ␣-DTX (200 nM) affected I K-slow , but not other K ϩ currents. B, Sample sweeps comparing the Together with our outside-out patch recordings, these data suggest that differences in K ϩ channel expression in PT, but not IT neurons, contribute to alterations in spike threshold dynamics in the fmr1-/y mouse mPFC, resulting in an increase in excitability.
As an additional test that differences in K ϩ channel function contribute to differences in excitability, we measured the sensitivity of threshold to 50 M 4-AP and 150 M Ba 2ϩ , and 200 nM ␣-DTX. Bath application of 50 M 4-AP produced a decrease in threshold for all but the shortest current durations in both genotypes (WT and fmr1-/y: n ϭ 6 cells from 4 mice; p Ͻ 0.01, ANOVA l= ). This effect, however, was larger in WT compared with fmr1-/y neurons, consistent with a lower contribution of K V 1containing channels and in agreement with our current measurements (p Ͻ 0.01; Fig. 11A m= ). Similarly, 200 nM ␣-DTX produced a larger affect on threshold in WT compared with fmr1-/y neurons (WT and fmr1-/y: n ϭ 5 cells from 4 mice; p ϭ 0.02; ANOVA; Fig. 11B n= ). Ba 2ϩ (150 M) produced a small but significant decrease in threshold for all but the shortest current durations in both genotypes (WT and fmr1-/y: n ϭ 8 from 4 and 5 mice, respectively; p Ͻ 0.01; ANOVA; Fig. 11C o= ). Interestingly, in WT mice the effect of 4-AP and ␣-DTX was larger than that of Ba 2ϩ (p Ͻ 0.001, ANOVA p= ), whereas in fmr1-/y mice there was no difference in the effect of these K ϩ channel blockers (p ϭ 0.55, ANOVA; Fig. 11D q= ). These data suggest that in fmr1-/y PT neurons, a decrease in Kv1-mediated K ϩ current contributes to an increase in excitability.

Discussion
We have provided evidence for neuron-type-specific channelopathies within a single brain region in FXS. We observed several differences in the functional expression of ion channels in L5 neurons of mPFC in the fmr1-/y mouse (Fig. 12). At the soma and dendrite, we observed differences in the subthreshold properties of fmr1-/y PT neurons consistent with a decrease in I h compared with WT neurons. These differences were largely eliminated by the application of the h-channel blocker ZD7288, suggesting that they were mediated, in part, by differences in h-channels. Consistent with this, we observed less I h at the distal dendrite of fmr1-/y PT neurons, suggesting that there is a reduction in functional h-channels. Additionally, we found alterations to several K ϩ channels at the soma of PT neurons. Compared with WT neurons, fmr1-/y neurons possessed greater maximum fast-inactivating A-type K ϩ current, but less slow-inactivating K ϩ current. The kinetics and pharmacologic sensitivity of these currents was consistent with Kv4 and Kv1 containing channels, respectively. In contrast to PT neurons, there were no genotypic differences in the maximum amplitude of these K ϩ currents at the soma of IT neurons. As a functional consequence of altered somatic K ϩ channel function in PT neurons, we observed an increase in excitability.
These findings expand on previous reports of channelopathies in various brain regions by demonstrating neuron specific phenotypes within a brain region (Brager and Johnston, 2014). In L5B neurons of somatosensory cortex a decrease in h-channel expression, together with altered BK channel function, contributes to a lowered threshold for dendritic Ca 2ϩ electrogenesis . In contrast to somatosensory cortex, there is an increase in h-channel expression in CA1 pyramidal neurons in the fmr1-/y mouse (Brager et al., 2012). Additionally, in CA1 pyramidal neurons, there is a decrease in Kv4-mediated A-type K ϩ current (Routh et al., 2013). Our data, together with these previously published findings, suggest that in FXS the function of a given ion channel can be affected in opposite ways depending upon brain region. More importantly, the absence of differences in K ϩ channel function in IT neurons suggests that neuron-specific channel phenotypes can occur even within a brain region, between different classes of principal neurons (Fig. 12). Alterations to other channel types reported in various brain regions may similarly display neuron-type specificity.

Identity of K ؉ channels altered in mPFC
Kv1 and Kv4 containing channels contribute to transient K ϩ currents (Coetzee et al., 1999). Although the details depend upon which auxiliary subunits are present, Kv4 channels inactivate and recover from inactivation an order of a magnitude faster than Kv1 channels (Castellino et al., 1995;An et al., 2000;Jerng et al., 2004;Kim et al., 2008). Thus, the I KA-fast we observed is consistent with Kv4 containing channels whereas the I K-slow is consistent with Kv1 channels. Several subunits within the Kv1 and Kv4 families can contribute to transient K ϩ currents (Coetzee et al., 1999). There is evidence that Kv4.2, 4.3, and 1.4 contribute to transient K ϩ currents at the soma of cortical pyramidal neurons (Carrasquillo et al., 2012). In addition, Kv1.2 containing channels contribute to a slowly inactivating K ϩ current at the soma (Bekkers and Delaney, 2001). All of these subunits have been reported to contribute to action potential threshold and thus alterations to any of these subunits could contribute to differences in I K and excitability we observe (Bekkers and Delaney, 2001;Kim et al., 2005;Kole et al., 2007;Higgs and Spain, 2011). Regardless of the subunit, the differences in maximum current we observed are consistent with a decrease in the number of Kv1 channels and an increase in Kv4 channels in PT neurons of mPFC in the fmr1-/y mouse.

Potential mechanisms
Brain region-specific channel phenotypes in FXS could be caused by a primary deficiency in an FMRP-mediated continued 4-AP-sensitive current to I K-slow . 4-AP (50 M) affected I K-slow , but not other K ϩ currents. C, Sample sweeps comparing the Ba 2ϩ -sensitive current to I KA-fast . Ba 2ϩ (150 M) affected I K-slow , but not other K ϩ currents. D, Summary of currents obtained via voltage separation versus K ϩ channel blockers. The amplitude of 4-AP and ␣-DTX-sensitive current is similar to the amplitude of I K-slow obtained via voltage separation, whereas the amplitude of the Ba 2ϩ -sensitive current is similar to I KA-fast . process within the cell. FMRP is an mRNA binding protein and has been shown to negatively regulate protein synthesis through ribosomal stalling (Li et al., 2001;Zalfa et al., 2003;Darnell et al., 2011). Consequently, the absence of FMRP could lead to elevated expression of ion channel proteins and/or their regulatory subunits. More recent evidence suggests that FMRP can also promote the translation of certain mRNAs (Bechara et al., 2009;Fähling et al., 2009;Gross et al., 2011). Thus, the absence of FMRP might then reduce the expression of ion channel proteins. In addition to its role in translation, FMRP participates in the activity-dependent transport of mRNA granules from the soma to axonal and dendritic locations (Antar et al., 2004; Kanai et al., 2004; Dictenberg et al., Figure 10. Fmr1-/y PT neurons are more excitable compared with WT neurons. A, The threshold for action potential generation in WT and fmr1-/y neurons was measured in response to just threshold current injections of increasing duration. Sample action potentials for each genotype are shown below. B, Phase plane plot of action potentials in fmr1-/y and WT PT neurons in response to 1.5 and 100 ms current injection. C, fmr1-/y neurons displayed a lower threshold for action potential generation for all current injections longer than 6 ms. D, There were no differences in threshold between fmr1-/y and WT IT neurons. ‫ء‬p Ͻ 0.01 comparison between genotypes. 2008). The absence of FMRP could therefore result in the trapping of ion channel mRNAs at the soma or the mislocalization of mRNAs and the encoded proteins. Both h-channel and A-type K ϩ channel mRNAs are known targets of FMRP (Gross et al., 2011;Lee et al., 2011;Darnell et al., 2011). Therefore, the channel phenotypes observed in FXS could be because of either of these mRNA regulatory processes. Lastly, FMRP can directly bind to target proteins, including ion channel subunits. Consequently, the absence of FMRP could alter the biophysical properties or surface expression of ion channels by promoting inappropriate subunit assemblies (Brown et al., 2010;Deng et al., 2013). It is noteworthy that in order to sufficiently explain the divergent channel phenotype we observe, these cellular processes would have to promote or suppress the function of a given ion channel in a neuron-specific manner.
Alternatively, the differences in channel function we observe could be the result of secondary changes in response to altered processes extrinsic to the cell. Both I h and I KA undergo synaptic-driven changes (Frick et al., 2004;Fan et al., 2005;Brager and Johnston, 2007;Losonczy et al., 2008). Thus, a potential cell extrinsic mechanism could involve changes in synaptic input to mPFC neurons that result in changes in ion channel function opposite to those that occur in CA1. Indeed there is evidence that synaptic input to neocortical and hippocampal pyramidal neurons is altered in the fmr1-/y mouse (Bureau et al., 2008;Gibson et al., 2008;Krueger et al., 2011;Testa-Silva et al., 2012;Patel et al., 2013;Tyzio et al., 2014). In addition, the function of both I h and I KA are affected by neuromodulation Carr et al., 2007;Dembrow et al., 2010). Thus, the divergent channel phenotypes observed in CA1 and neocortex could be caused by differences in neuromodulation. Determining whether neuron-type-specific alterations to ion channel function are caused by a cell autonomous or cell extrinsic mech- Figure 11. Fmr1-/y neurons are less sensitive to Kv1 channel blockers compared with WT neurons. A, Phase plane plots for action potentials generated in response to 1.5 and 50 ms current injections before and after 50 M 4-AP application. 4-AP decreased threshold in both genotypes, but more so in WT neurons. B, Phase plane plots for action potentials generated before and after 200 nM ␣-DTX application. ␣-DTX decreased threshold in both genotypes, but more so in WT neurons. C, Phase plane plots for action potentials generated before and after 150 M Ba 2ϩ application. Ba 2ϩ decreased action potential threshold in both genotypes. D, 4-AP and ␣-DTX application had a larger effect on action potential threshold compared with Ba 2ϩ in WT neurons. In contrast, 4-AP and ␣-DTX had the same effect on action potential threshold compared with Ba 2ϩ in fmr1-/y neurons. For current steps longer than 6 ms, there was significant difference in the effect of 4-AP and/or ␣-DTX on threshold between WT and fmr1-/y mice. anism (eg, differences in network activity) is a difficult challenge, yet fundamental to our understanding of the neurobiology of FXS.

Functional implications
The combined effects of the channel phenotypes we observe could have a substantial impact on mPFC function. PT neurons receive synaptic input in L1 from various sources and the presence of I h greatly influences the transfer of these inputs to the soma (Dembrow et al., 2015). Specifically, I h contributes to the filtering properties of the membrane such that inputs arriving in the theta frequency range are selectively transferred to the soma (Ulrich, 2002;Cook et al., 2007). Additionally, I h compensates for distance dependent temporal delays imposed by the dendrite, ensuring that distal inputs arrive at the soma with a similar latency to proximal ones (Vaidya and Johnston, 2013). This has the effect of narrowing the temporal window under which synaptic input arriving at different dendritic locations can summate at the soma to drive an action potential (Dembrow et al., 2015). Thus, the decrease in I h we observe in the fmr1-/y mouse could: (1) affect the frequency preference of PT neurons to synaptic input, and (2) increase integration at the soma of asynchronous dendritic synaptic inputs. Consequently, fmr1-/y PT neurons may function as temporal integrators of synaptic input, in contrast with WT PT neurons, which function as coincidence detectors (Dembrow et al., 2015).
These changes in I h could work synergistically with K ϩ channel alterations to make fmr1-/y PT neurons generally more responsive to synaptic input. K ϩ channels affect the threshold for spike generation in part by influencing the voltage at the axon initiation site (Bekkers and Delaney, 2001;Kole et al., 2007). Like h-channels, K ϩ channels contribute to the band pass filtering properties of the membrane. Thus, the changes in K ϩ channels we observe Figure 12. Summary of neuron type-specific channel phenotypes observed in the fmr1-/y mouse. Dendritic and somatic I h , I KA -fast, and I K -slow levels in the fmr1-/y mouse relative to WT in CA1 pyramidal neurons (Brager et al., 2012;Routh et al., 2013), L5 PT neurons of PFC and IT neurons of PFC. There is also a decrease in I h in the dendrites of L5 neurons of somatosensory cortex (data not shown; Zhang et al, 2014). Upward arrow indicates increase relative to WT. Downward arrow indicates decrease relative to WT. Horizontal arrow indicates no difference relative to WT. could also decrease the preference of PT neurons for high-frequency input as has been observed upon blockade of Kv1 channels in L2/3 neurons (Higgs and Spain, 2011). Together with changes to synaptic input, these changes may contribute to altered up-states observed in cortical pyramidal neurons in the fmr1-/y mouse Gonçalves et al., 2013).
FXS patients display behavioral deficits associated with PFC dysfunction, including hyperactivity, attention deficits, impulsivity, and behavioral inflexibility. The channel phenotypes we describe here may thus contribute to these behavioral phenotypes. Finally, our findings of celltype-specific phenotypes highlights the challenges associated with developing effective treatment strategies in FXS. A critical implication of our findings is that existing therapies that target a given channel phenotype, although effective in treating one feature of FXS could simultaneously exacerbate others. Understanding the mechanisms of these neuron/brain region-specific effects in FXS may prove critical in developing treatment strategies.