INPP5E controls ciliary localization of phospholipids and the odor response in olfactory sensory neurons

Abstract

The lipid composition of the primary cilia membrane is emerging as a critical regulator of cilia formation, maintenance and function. Here, we show that conditional deletion of the phosphoinositide 5'-phosphatase gene Inpp5e, mutation of which is causative of Joubert syndrome, in terminally developed mouse olfactory sensory neurons (OSNs), leads to a dramatic remodeling of ciliary phospholipids that is accompanied by marked elongation of cilia. Phosphatidylinositol (4,5)-bisphosphate [PI(4,5)P2], which is normally restricted to the proximal segment redistributed to the entire length of cilia in Inpp5e knockout mice with a reduction in phosphatidylinositol (3,4)-bisphosphate [PI(3,4)P2] and elevation of phosphatidylinositol (3,4,5)-trisphosphate [PI(3,4,5)P3] in the dendritic knob. The redistribution of phosphoinositides impaired odor adaptation, resulting in less efficient recovery and altered inactivation kinetics of the odor-evoked electrical response and the odor-induced elevation of cytoplasmic Ca2+. Gene replacement of Inpp5e through adenoviral expression restored the ciliary localization of PI(4,5)P2 and odor response kinetics in OSNs. Our findings support the role of phosphoinositides as a modulator of the odor response and in ciliary biology of native multi-ciliated OSNs.

Keywords: INPP5E; Mouse; Odor response; Olfactory cilia; Phospholipids.

Fig. 1.

Loss of INPP5E causes redistribution of PIP₂ and elongation of cilia in mouse OSNs. (A) PLCPH–GFP, a probe for PIP₂, is mostly restricted to the knob of WT OSNs. In a small percentage of OSNs, a ciliary segment of varying length up to the full length (black arrows) is also enriched in PIP₂. The inert membrane-bound lipid probe MP–mCherry was used as a counterstain to label the full length of axoneme, and does not have the highly restricted localization of PLCPH-GFP resulting in overlapping colors (middle panel, white). PIP₂ was evenly distributed in the plasma membrane of the knob as shown in z-stack view (right panel). Yellow lines denote z-stack projection shown at the bottom and right side of the image. (B) In contrast to what is seen in the WT, in Inpp5e^osnKO (INPP5E-KO) PLCPH–GFP decorated the entire length of every cilium. A color-shifted image is shown to accentuate the equal distribution of PLCPH–GFP and MP–mCherry labeling (B, middle panel). The PIP₂ redistribution is evident also in the z-stack side view showing substantial enrichment at the base of cilia and along the proximal segment (PS, red arrows) whereby PIP₂ level in the knob periciliary plasma membrane was not changed (right panel). (C) More than 50% of WT OSNs showed no PIP₂ in their cilia. 18% of OSNs had only a single PIP₂-positive cilium whereas three other groups of neurons equally represented the remaining 30%. Conversely, PIP₂ was detected in 100% of OSNs in Inpp5e^osnKO (KO, green bar). A total of 318 cells in 4 mice were analyzed in the WT group and 36 cells in 3 mice were analyzed in the KO group. (D) Length distribution within the same sets of cells of PLCPH–GFP-positive aspects of cilia (PIP₂ domain) in WT was substantially shifted to shorter values compared to the full cilia length measured with MP-mCherry, yielding 29.5±0.5 µm (n=753, 4 mice). (E) Distribution of both PLCPH–GFP and MP–mCherry length values showed a complete overlap in Inpp5e^osnKO OSNs. The average full ciliary length in the KO OSNs, 35.3±0.6 µm (n=495, 3 mice) was significantly longer than in the WT (unpaired t-test, t=7.363, d.f.=1246, P<0.0001). Data shown as mean±s.e.m.

Fig. 2.

Virally induced ectopic expression of full-length WT human INPP5E tagged with GFP completely reversed mislocalization of PIP₂ in Inpp5e^osnKO mouse cilia. (A,B) Inpp5e^osnKO mice were infected at P8–P14 with a triple dose of Ad-GFP-INPP5E-WT and tested 8–10 days later. GFP–INPP5E-WT is enriched in OSN knobs and also localizes to cilia. The KO mice were co-infected with PLCPH–mCherry to measure rescue of the PIP₂ localization. Several knobs of co-infected OSNs are indicated with arrowheads. (C) Magnified dual-color view of the area marked with a square in B shows several knobs of OSNs co-infected with both viruses (arrowheads) resulting in a complete loss of ciliary PIP₂ (magenta). (D) Rescue was quantified by measuring length of PIP₂ positive ciliary aspect in the WT littermates and KO mice. The KO OSNs were identified within the same preparation by a strong ciliary distribution of PLCPH–mCherry, and also lacking any detectable GFP–INPP5E-WT fluorescence. Rescue completely reversed Inpp5e^osnKO deficiency [PIP2 domain length 4.9±0.27 µm (n=110, 16 cells, 3 mice), WT; 28.5±1.37 µm (n=54, 5 cells, 3 mice), KO; 4.2±0.3 µm (n=122, 17 cells, 3 mice), Rescue, one-way ANOVA, F(DFn, DFd) 86.73 (2283), ****P<0.0001]. ns, not significant. (E,G) Inpp5e^osnKO KO mice in a different group were infected with Ad-PLCPH-mCherry and Ad-GFP-INPP5E-D477N encoding for catalytically inactive phosphatase. The GFP–NPP5E-D477N mutant was localized to the full cilia length (E). Knobs of co-infected OSNs showing no change in PLCPH ciliary localization are marked with solid arrows. Some knobs had less PLCPH probe localized to cilia (open arrows) reminiscent of the KO phenotype. (G,H) Expression of GFP–INPP5E-D477N resulted in a significantly smaller number of OSNs having a complement of PIP2-decorated cilia. This reduction was quantified in H, 17.6±0.09% (D477N, n=61, 3 mice), compared to GFP–INPP5E-WT, 61.2±0.05% (INPP5E-WT, n=83 cells, 3 mice), unpaired t-test, t=4.536, d.f.=24, ***P=0.001). Data shown as mean±s.e.m.

Fig. 3.

Other phosphoinositides than PIP₂ in mouse OSNs are almost exclusively restricted to the knobs and changed their level in an INPP5E-dependent manner. (A,D) The location of the PI(4)P probe mCherry–P4M-SidM was not significantly affected by loss of INPP5E showing only insignificant trending decrease in the knobs (179±26 relative units, WT, n=94, 3 mice; 143±17 relative units, KO, n=54, 3 mice; unpaired t-test, t=0.9777, d.f.=146, P=0.3298). (B,E) A tandem PH domain, Tapp1 tagged with GFP, was used to specifically label membrane PI(3,4)P₂, which was found to only be enriched in the knobs and in cilia in a small fraction of OSNs. Importantly, the overall pattern of PI(3,4)P₂ distribution did not change in Inpp5e^osnKO. Fluorescence intensity, however, measured in OSN knobs showed a significant decrease in the KO compared to WT mice (280±11 relative units, n=830, 3 mice, WT; 174±7 relative units, n=858, 3 mice, KO; unpaired t-test, t=8.453, d.f.=1686, P<0.0001). (C,F) PIP₃ detected with a Btk-PH domain tagged with GFP, was restricted mostly to the knobs with a relatively low presence in cilia of the WT and KO. Quantitative analysis of fluorescence showed increase of the intensity in the knobs of the KO (668±64 relative units, n=60, 3 mice, WT; 1495±185 relative units, n=91, 3 mice, KO; unpaired t-test, t=3.536, d.f.=149, ***P=0.0005). (G) Fluorescence intensity of MP–mCherry, used as a negative control, was not significantly different in the OSN knobs of WT and KO mice (340±31 relative units, n=46, 3 mice, WT; 378±23 relative units, n=70, 3 mice, KO; unpaired t-test, t=1.001, d.f.=114, P=0.3188). Data shown as mean±s.e.m.

Fig. 4.

The distribution of integral membrane lipids was not changed in the OSNs and cilia in Inpp5e^osnKO KO mice. (A) D4H–mCherry, a cholesterol-binding probe was enriched in the proximal segment of olfactory cilia equally in the WT and KO OSNs (arrowheads). Cholesterol was also detected, albeit at a lower level, in the full length of the ciliary axoneme. (B) Phosphatidylserine, probed with C2 motif of lactadherin, was uniformly distributed along the cilia and was also enriched in the dendritic knobs of OSNs. (C) A sphingomyelin-specific probe, Eqt2-SMP–GFP, was mostly enriched in the OSN knobs and detected at a low level in cilia. (D) Glycosylated phosphatidylinositol was probed in OSNs with a human folate 1 receptor, GPI–GFP which failed to detect any presence in cilia and it was mostly restricted to the knobs in both the WT and Inpp5e^osnKO mouse.

Fig. 5.

Soluble and polytopic proteins with affinity to PIP₂ mislocalize in olfactory cilia of Inpp5e^osnKO. (A) Tubby-like proteins tagged with GFP (TULP1–GFP and TULP3–GFP) were preferentially restricted to the knobs in the WT (upper left panel). Build-up of PIP₂ in cilia of the KO resulted in complete redistribution of TULP1 (bottom panels). Note that loss of INPP5E activity led to a depletion of TULP1 within knobs, revealing the proximal segment of cilia decorated with TULP1–GFP (arrowheads, right bottom panel). Quantification of the percentage of the OSN knobs having TULP1-positive cilia per analyzed image showed significant increase in the KO (25.49±0.06%, n=4, 3 mice, WT; 100%, n=6, 3 mice, KO; Mann–Whitney t-test, **P=0.0048). (B) TULP3–GFP, like TULP1, also showed dramatic redistribution between the knob and cilia resulting in a significant increase of percentage of knobs with TULP3-positive cilia (30.87±0.12%, n=6, 3 mice, WT; 100%, n=4, 3 mice, KO; Mann–Whitney t-test, **P=0.0095). (C) The K⁺ inward rectifier ion channel Kir2.1-mCherry, a polytopic protein with two membrane-spanning loops and a known affinity to PIP₂, also changed its ciliary distribution in Inpp5e^osnKO OSNs. Kir2.1–mCherry moved into the ciliary membrane in a significantly larger fraction of OSNs in the KO (right upper panel, 3.02±0.02%, n=12, 3 mice, WT; 24.34±5.89%, n=10, 3 mice, KO; Mann–Whitney t-test, **P=0.0023). (D,E) As a negative control, we used a different ion channel, PC2 (PKD2 or TRPP1) tagged with mCherry (mCherry–PC2) and a microtubule-binding protein Efhc1 (GFP-Efhc1), both of which did not change their distribution in the knobs of Inpp5e^osnKO (upper panels, WT; bottom panels, KO). Data shown as mean±s.e.m.

Fig. 6.

INPP5E is responsible for shaping the odor-evoked intracellular Ca²⁺ transient in the knob of OSNs. (A,B) Ectopically expressed GCaMP6F was visualized in the en face preparation of mouse OE by wide-field fluorescence microscopy. (A) Bright spots represent numerous OSN knobs. (B) Stimulation micropipette filled with a mixture of 132 different odorants diluted to 1:10,000 in ACSF was positioned as indicated. A single 100-ms pulse at 10 psi pressure generated a plume of fluorescein covering an area over the epithelial surface demarcated by a dotted line. (C) Repetitive application of a single odor pulse (arrowheads) evoked nearly identical responses. GCaMP6F fluorescence corrected for background was calculated as (F–Fo)/Fo. (D) Individual traces measured in more than 100 OSNs across several areas and 3 mice per each genotype were averaged to create the graph. Traces were normalized to the peak value before averaging. Arrowheads show the time of stimulation. (E,F) The odor-evoked GCaMP6F response had a faster decay in the KO OSNs than the response in the WT control group (WT, 6.49±0.37s, n=167, 3 mice; KO, 3.59±0.18s, n=110, 3 mice, unpaired t-test, t=6.077, d.f.=275, ****P<0.0001). The response in the KO also had a faster rising phase (WT: 1.12±0.07s, n=46, 3 mice; KO: 0.80±0.08s, n=30, 3 mice, unpaired t-test, t=2.936, d.f.=74, **P=0.0044). To calculate termination phase time constant (decay tau) each individual trace was fit to an exponential function. Rise time 10–90% was defined as time to reach from 10% to 90% of the response peak level. Data shown as mean±s.e.m.

Fig. 7.

A faster single-cell odor response translates into a more transient EOG in Inpp5e^osnKO. (A) Representative EOG traces recorded in response to 100-ms pulse of amyl acetate vapor, driven from the 90-ml head space of bottles containing increasing concentration ranging from 10⁻⁶ M to a maximum of 1 M (indicated at the individual traces). Odor application is denoted by a black arrowhead. (B) Dose–response relationship showing that there is no significant difference between the WT and KO (WT, n=7, 3 mice; KO, n=11, 4 mice; two-way ANOVA, F(5, 102)=0.1858, P=0.9674). (C,D) Rise time of the EOG evoked by a single 100-ms pulse of 10⁻² M amyl acetate (rise time 10–90%) was decreased in the KO compared to the WT (WT, 174.5±7.7 ms, n=37, 5 mice; KO, 157.9±10.9 ms, n=40, 7 mice; Mann–Whitney test, *P=0.0221), similar to the time constant (decay tau) of the termination phase (WT, 4.57±0.15s, n=81, 5 mice; KO, 3.40±0.16s, n=28, 4 mice; unpaired t-test, t=4.386, d.f.=107, ****P<0.0001). (E) EOG evoked by a longer 5-s pulse of 10⁻³ M amyl acetate applied at the time indicated by a square step (aac, 10⁻³M) also appeared more transient in the Inpp5e^osnKO KO (F). Ectopic expression of the full-length WT INPP5E partially rescued the EOG shape (G). (H–J). The ratio between peak amplitude of second and first EOG response, plateau-to-peak ratio and time constant of termination phase (decay tau) were significantly affected by the loss of INPP5E activity and restored by ectopic expression in OSNs of the WT INPP5E. Second/first peak ratio (WT, 0.733±0.026, n=18; KO, 0.514±0.022, n=9; Rescue, 0.582±0.021, n=12; Mann–Whitney t-test, WT versus KO, ****P<0.0001; KO versus rescue, P=0.0409). Peak/plateau ratio (WT, 0.462±0.028, n=11; KO: 0.230±0.017, n=13; Rescue: 0.336±0.024, n=16; Mann–Whitney t-test, WT versus KO, ****P<0.0001; KO versus Rescue, P=0.0021). Time constant of termination phase (WT, 1.707±0.124s, n=19; KO, 1.311±0.080s, n=20; Rescue: 1.991±0.134, n=16; Mann–Whitney t-test, WT versus KO, P=0.0083; KO versus rescue, ****P<0.0001). Data shown in H–J are based on the experiments performed on 9 WT, 6 KO and 6 rescued mice and are presented as mean±s.e.m.