The ciliary lumen accommodates passive diffusion and vesicle-assisted trafficking in cytoplasm–ciliary transport

Abstract

Transport of membrane and cytosolic proteins into the primary cilium is essential for its role in cellular signaling. Using virtual three-dimensional superresolution light microscopy, the movements of membrane and soluble proteins from the cytoplasm to the primary cilium were mapped. In addition to the well-characterized intraflagellar transport (IFT) route, we found two new pathways within the lumen of the primary cilium: passive diffusion and vesicle-assisted transport routes that are adopted by proteins for cytoplasm–cilium transport in live cells. Through these pathways, approximately half of IFT motors (KIF3A) and cargo (α-tubulin) take the passive diffusion route, and more than half of membrane-embedded G protein–coupled receptors (SSTR3 and HTR6) use RAB8A-regulated vesicles to transport into and inside primary cilia. Ciliary lumen transport is the preferred route for membrane proteins in the early stages of ciliogenesis, and inhibition of SSTR3 vesicle transport completely blocks ciliogenesis. Furthermore, clathrin-mediated, signal-dependent internalization of SSTR3 also occurs through the ciliary lumen. These transport routes were also observed in Chlamydomonas reinhardtii flagella, suggesting their conserved roles in trafficking of ciliary proteins.

INTRODUCTION

The primary cilium is an antennalike projection present on nearly all mammalian cells, which is involved in a variety of signal transduction pathways, including planar cell polarity, Hedgehog signaling, neuronal signaling, nutrient sensing, mechanosensation, olfaction, phototransduction, and cellular growth (Nauli et al., 2003; Corbit et al., 2005; Gerdes and Katsanis, 2005; Ross et al., 2005; Marshall and Nonaka, 2006; Scholey and Anderson, 2006; Rohatgi et al., 2007; Breunig et al., 2008; Jones et al., 2008; Craft et al., 2015; Wang and Dynlacht, 2018; Wheway et al., 2018; Anvarian et al., 2019; Nachury and Mick, 2019). Central to these processes is the unique composition of structural, soluble, and transmembrane (TM) proteins in primary cilia (Ostrowski et al., 2002; Pazour et al., 2005; Ishikawa et al., 2012; Breslow et al., 2018), each of which serves a specific purpose. For example, structural proteins, such as tubulins and their associated proteins, form the backbone of primary cilia and facilitate mechanosensation (Singla and Reiter, 2006; Shida et al., 2010; Wloga et al., 2017), TM proteins typically function as signal receptors (Lerea et al., 1986; Anholt et al., 1987; Pazour et al., 2002; Yoder et al., 2002; Barzi et al., 2011; He et al., 2014; Singh et al., 2015; Hilgendorf et al., 2016; Bangs and Anderson, 2017), and soluble proteins function as intracellular messengers (Nair et al., 2005; Rosenzweig et al., 2007; Jensen and Leroux, 2017; Brooks et al., 2018).

Primary cilia are dynamic signaling organelles; therefore the efficient cytoplasm–ciliary transport of protein is essential for proper function. Previous work has shown that a diffusion barrier at the base of the primary cilium restricts the entry and exit of soluble and TM proteins (Calvert et al., 2010; Kee et al., 2012; Breslow et al., 2013; Lin et al., 2013). Further, both active transport via microtubule-based motors and free diffusion drive the entry of structural/soluble proteins into Chlamydomonas reinhardtii flagella (Hao et al., 2011; Craft et al., 2015). Regarding TM proteins, two models have been proposed to explain cytoplasm–cilium transport: 1) vesicle fusion outside the primary cilium and subsequent transport of TM proteins into primary cilia after loading onto an intraflagellar transport (IFT) train, and 2) vesicle fusion at an unknown location inside primary cilia (Figure 1A; Hunnicutt et al., 1990; Vieira et al., 2006; Pazour and Bloodgood, 2008; Nachury et al., 2010; Chuang et al., 2015; Jensen and Leroux, 2017). While the precise site of extraciliary fusion is unclear and may vary between cell types, the first model is largely derived from early findings in frog photoreceptors that show that rhodopsin localizes to vesicles that appear to be fusing at the periciliary ridge complex, a unique membrane domain near the base of the photoreceptors’ connecting cilia (Papermaster et al., 1985). The second model, which appears to accomplish the transport and enrichment of TM proteins in primary cilia in a single step, arises from the occasional appearance of vesicle-like structures inside primary cilia and photoreceptors (Reese, 1965; Poole et al., 1985; Jensen et al., 2004; Chuang et al., 2015; Jana et al., 2018). The role of IFT in the import of TM proteins is less clear in this model.

FIGURE 1:

Transport routes of IFT, structural, and soluble proteins in live and permeabilized cells. (A) Schematic highlighting the structure of TZ and two current models for the transport of structural and membrane proteins into primary cilia. (B) Schematic of SPEED microscopy–based imaging of GFP-labeled proteins moving through the TZ labeled with NPHP4-mCherry. The spatial distributions of transiting molecules are described by both Cartesian (x,y,z) and cylindrical (x, θ, r) coordinate systems. (C–E) Imaging of IFT-B component. (C) Epifluorescence microscopy image of IFT20-GFP (green) and NPHP4-mCherry (red) overlaid with single-molecule IFT20-GFP locations (white spots). Scale bar: 1 μm. (D) Single-molecule IFT20-GFP locations were converted to three-dimensional locations and plotted as a histogram along the R dimension in the cylindrical system. Numbers indicate radial distance as mean ± SE. (E) Spatial representation of the histogram in D as cross-sectional view of IFT20-GFP’s transport route (green cloud) in the TZ overlaid on the ultrastructure of the TZ. (F–H) Imaging of IFT-A component. (F) Epifluorescence microscopy image of GFP-IFT43 (green), NPHP4-mCherry (red), and Arl13b-mCherry (red) overlaid with single-molecule GFP-IFT43 locations (white). Dashed white circle indicates region of photobleaching to remove Arl13b-mCherry fluorescence. Scale bar: 1 μm. Small images on the left throughout this figure show the result of photobleaching Arl13b-mCherry to reveal the precise location of NPHP4-mCherry. Scale bar: 1 μm. (G) Single-molecule GFP-IFT43 locations along the R dimension. (H) Spatial representation of the histogram in G. (I–K) Imaging of IFT motor. (I) Epifluorescence microscopy image of KIF3A-GFP (green), NPHP4-mCherry (red), and Arl13b-mCherry (red) overlaid with single-molecule KIF3A-GFP locations (white). Scale bar: 1 μm. (J) Single-molecule KIF3A-GFP locations along the R dimension. (K) Spatial representation of the histogram in J. (L–N) Imaging of IFT cargo. (L) Epifluorescence microscopy image of α-tubulin-GFP (green), NPHP4-mCherry (red), and Arl13b-mCherry (red) overlaid with single-molecule α-tubulin-GFP locations (black). Scale bar: 1 μm. (M) Single-molecule α-tubulin-GFP locations along the R dimension. (N) Spatial representation of the histogram in (M). (O–Q) Imaging of GFP. (O) Epifluorescence microscopy image of Arl13b-mCherry (red) and NPHP4-mCherry (red) overlaid with single-molecule GFP locations (black). Scale bar: 1 μm. (P) Single-molecule GFP locations along the R dimension. (Q) Spatial representation of the histogram in P. (R–T) Imaging of IFT motor in permeabilized cells. (R) Epifluorescence microscopy image of Arl13b-mCherry (red) and NPHP4-mCherry (red) overlaid with single-molecule KIF3A-GFP locations (black) in permeabilized cells. Scale bar: 1 μm. (S) Single-molecule KIF3A-GFP locations along the R dimension. (T) Spatial representation of the histogram in S. (U–W) Imaging of IFT cargo in permeabilized cells. (U) Epifluorescence microscopy image of Arl13b-mCherry (red) and NPHP4-mCherry (red) overlaid with single-molecule α-tubulin-GFP locations (black) in permeabilized cells. Scale bar: 1 μm. (V) Single-molecule α-tubulin-GFP locations along the R dimension. (W) Spatial representation of the histogram in V.

From an ultrastructural perspective, primary cilia are composed of three main subregions. First, the basal body, which contains the nine triplet microtubules of the mother centriole and their associated distal and subdistal appendages. Second, the transition zone (TZ, 300–1000 nm in length and 160–250 nm in diameter), which lies distal to the basal body. The TZ is where the nine microtubule triplets transition to nine microtubule doublets (termed 9+0) and is the proposed location for the components of the selectivity barrier (Craige et al., 2010; Czarnecki and Shah, 2012; Kee et al., 2012; Najafi et al., 2012; Tony Yang et al., 2015). Further, the TZ contains Y-shaped structures in electron micrographs that appear to tether the microtubules to the encompassing ciliary membrane. Third, the main ciliary shaft, which is composed of the 9+0 microtubule axoneme, is 200–250 nm in diameter and extends into the extracellular environment (Czarnecki and Shah, 2012), where it receives and transmits signals via TM receptors (Hilgendorf et al., 2016; Ye et al., 2018) and ectosomal release (Wang et al., 2014; Nager et al., 2017; Phua et al., 2017), respectively. The TZ and the ciliary shaft both are encompassed by a membrane with a unique lipid and protein composition (Kaneshiro, 1987; Nauli et al., 2003; Praetorius and Spring, 2003; Molla-Herman et al., 2010; Praetorius, 2015; Tony Yang et al., 2015). While the primary cilium axoneme is generally modeled as a rigid 9+0 structure, various asymmetries may occur, such as a microtubule doublet turning into a singlet and/or collapsing toward the inner lumen. This is especially likely toward the distal end (Rogowski et al., 2013; O’Hagan et al., 2017; Sun et al., 2019).

To interrogate the significant question regarding the localization of TM proteins at the three main landmark regions of the primary cilium (basal body, transition zone, and ciliary shaft), single-point edge-excitation subdiffraction (SPEED) microscopy, a virtual three-dimensional superresolution imaging technique, was employed (Ma and Yang, 2010; Ruba et al., 2018b; Li et al., 2021). SPEED microscopy relies on high–spatiotemporal resolution (2 ms, 10–20 nm) capture of 2D single-molecule locations of labeled proteins in live cells and the subsequent back-projection algorithm to obtain the 3D probability spatial density distribution of targeted proteins inside the primary cilium (Supplemental Figure S6; Ma and Yang, 2010; Ma et al., 2016; Ruba et al., 2018a,b, 2019). This approach has been utilized to determine the contributions of IFT and free diffusion to the movement of soluble proteins within the lumen of the cilium shaft (Luo et al., 2017). Here, the transit of soluble and TM proteins at the TZ was evaluated and found to demonstrate that the lumen mediates vesicular trafficking of TM proteins as well as passive diffusion of structural and soluble proteins. Further, the axonemal lumen was identified as the preferred route for TM proteins during two functionally critical ciliary stages: ciliogenesis and clathrin-mediated, signal-dependent internalization. In addition, this luminal transport route of TM proteins is distinct from passively diffusing ciliary structural and soluble proteins. Last, it was demonstrated that the luminal route is also present in flagella of the model system, Chlamydomonas reinhardtii, suggesting a broader evolutionary origin.

RESULTS

To validate SPEED microscopy’s ability to track protein localization within the TZ of live primary cilia, the transport routes for proteins whose localizations are well characterized were evaluated using SPEED and compared with the published data. Specifically, IFT components (IFT20 and IFT43), freely diffusing soluble molecules (free GFP), IFT motors and cargos (KIF3A and α-tubulin), and an externally labeled TM protein (SSTR3) were evaluated. The NIH-3T3 cell line was selected, as serum starvation causes cell cycle arrest and ciliogenesis in the majority of the cells (Ott and Lippincott‐Schwartz, 2012).

These experiments necessitate a precisely localized transition zone centroid as well as determination of the axis along which transport is utilized by the cell to move material from the cell body into the primary cilium (Materials and Methods). First, to obtain the centroid of the transition zone, the NPHP4-mCherry fluorescence, a TZ marker that localizes to the Y-shaped linkers (Awata et al., 2014), is fitted with a 2D Gaussian function. As has been detailed elsewhere (Zhang et al., 2006, 2007; Stallinga and Rieger, 2010; von Diezmann et al., 2017; Samuylov et al., 2019), the point spread functions of fluorescently emitting fluorophores closely approximate a Gaussian distribution. Therefore, the centroid of the emitting fluorophores can be determined by fitting a two-dimensional Gaussian function to the isolated fluorophore, provided there is sufficient temporal or physical space between the detected fluorophore and similarly emitting fluorescent proteins (Li et al., 2021; Tingey et al., 2023). Second, to determine the orientation and location of the transport axis, the center line of the ciliary marker fluorescence is extrapolated back into the transition zone. After this procedure, it was found that the single-molecule distribution overlapped with the transition zone centroid well within error, thereby engendering confidence in the subsequent determination of the central transport axis.

In greater detail, this is accomplished experimentally by first localizing the position of the TZ in live primary cilia by detecting the fluorescence of NPHP4-mCherry. The fluorophore-labeled protein of interest was then prephotobleached via SPEED microscopy to reduce the labeled concentration of that protein locally within the TZ (Figure 1B; Supplemental Figure S1A). Then a high-speed CCD camera set at 500 frames per second (2 ms/frame) was employed to record the two-dimensional localizations of single labeled proteins of interest in the TZ after they entered the photobleached area from neighboring regions (Video 1). Typically, 30,000–60,000 frames were captured within 1–2 min. In this time frame, it was found that the spatial shift of the primary cilium is <10 nm. Furthermore, thousands of single-molecule events with localization precisions of 10–20 nm were selected from these frames. These single-molecule localizations were then superimposed on the centroid of the NPHP4-mCherry fluorescence spot used to mark the TZ to generate a two-dimensional high-resolution spatial distribution of the protein of interest (Supplemental Figure S1, B and C). Finally, a two- to three-dimensional transformation algorithm was utilized to obtain the three-dimensional spatial density histogram of the protein of interest along the radius of the primary cilium (Figure S1, D–F; Ma and Yang, 2010; Ma et al., 2016; Ruba et al., 2018a, b, 2019; Li et al., 2021; Tingey et al., 2023).

Movie S1.

Download video file ^{(10.1KB, mov)}

Typical events of SSTR3-GFP single molecules captured at the transition zone of primary cilium. A high-speed CCD camera set at 500 frames per second (2 ms/frame) was used to record the 2D localizations of SSTR3-GFP single molecules (white) in the NPHP4-mCherry-labeled transition zone (red).

As has been detailed elsewhere, for the three-dimensional reconstructed back projection to validly represent the actual three-dimensional distribution of single molecules, the single-molecule localizations are ideally distributed with rotational symmetry (Ruba et al., 2018a, 2019; Li et al., 2021). However, the algorithm has been demonstrated to be sufficiently robust so that, should deformation or incomplete symmetry be present within the structure, the algorithm is still capable of producing a highly reproducible three-dimensional reconstruction of the distribution (Ruba et al., 2019). Electron microscopy (EM) studies of the primary cilium validate the rotational symmetry of the ciliary ultrastructure, which provides a scaffolding for protein transport routes to occur (Czarnecki and Shah, 2012). Through Monte Carlo simulation, given typical experimental parameters (single-molecule localization precision, number of single-molecule locations, imperfect degree of labeling symmetry and rotational symmetry), we are able to obtain <10 nm standard error on the mean peak position of routes for targeted proteins in primary cilia (Supplemental Figures S7–S11; Table 1). The code for the simulations, two- to three-dimensional transformation, and sample and experimental data is available at https://github.com/YangLab-TempleTemple.

TABLE 1:

Transport route localization precision and precision/radius ratio for all three-dimensional transport routes.

Protein	Cilia location	Condition	# of cilia	Route mean (nm)	Route S.D. (nm)	# of single molecules	Average precision (nm)	Precision/radius ratio	σTR (nm)
α-tubulin	TZ	Normal	5	0	49	579	16	≫0.63	<0.1
				111	22	517	16	0.14	3.2
SSTR3	TZ	Normal	6	55	31	1471	18	0.32	2.0
				129	17	1152	18	0.14	3.2
IFT20	TZ	Normal	5	105	23	1162	16	0.15	1.8
IFT43	TZ	Normal	5	108	26	1078	16	0.15	1.8
KIF3A	TZ	Normal	5	0	27	613	16	≫0.63	<0.1
				92	36	633	16	0.17	2.7
GFP	TZ	Normal	5	0	33	1135	16	≫0.63	<0.1
α-tubulin	TZ	Perm.	6	0	39	808	16	≫0.63	<0.1
KIF3A	TZ	Perm.	5	0	37	918	16	≫0.63	<0.1
SSTR3	TZ	Perm.	5	44	33	1137	23	0.53	4.2
				126	17	509	23	0.18	4.9
SSTR3	TZ	Membrane	4	131	12	1142	13	0.10	1.7
SSTR3	BB	Normal	4	55	27	660	23	0.42	3.9
				120	20	159	23	0.19	7.6
SSTR3	Shaft	Normal	5	55	25	3117	15	0.27	1.0
				131	22	2078	15	0.11	1.4
HTR6	BB	Normal	4	45	26	765	27	0.59	10.3
				127	20	187	27	0.21	9.1
HTR6	TZ	Normal	5	39	34	1364	24	0.62	9.2
				130	24	337	24	0.19	6.1
HTR6	Shaft	Normal	5	46	27	1659	22	0.48	2.4
				122	20	457	22	0.18	4.6
RAB8A	TZ	Normal	4	50	38	2326	13	0.26	0.9
				133	20	402	13	0.10	2.9
RAB8A	Shaft	Normal	4	53	34	2648	14	0.26	1.3
				134	16	669	14	0.10	2.8
SSTR3	TZ	1 µM GCA	3	53	31	957	20	0.37	2.4
				130	22	372	20	0.15	4.4
SSTR3	TZ	5 µM GCA	4	60	35	824	22	0.37	4.1
				136	17	424	22	0.16	4.6
SSTR3	TZ	1 h growth	3	42	26	1269	21	0.50	3.1
SSTR3	TZ	3 h growth	4	49	37	1596	14	0.29	1.6
				135	43	239	14	0.10	3.9
SSTR3	TZ	6 h growth	3	50	28	1122	21	0.42	3.2
				133	27	230	21	0.16	5.1
SSTR3	TZ	12 h growth	3	40	28	742	35	0.88	14.3
				134	25	121	35	0.26	17.4
HTR6	TZ	1 h growth	3	41	36	1132	19	0.45	2.5
HTR6	TZ	3 h growth	3	50	30	729	14	0.28	2.0
				127	21	182	14	0.11	4.1
HTR6	TZ	6 h growth	4	35	32	876	19	0.56	4.7
				125	24	341	19	0.15	4.7
HTR6	TZ	12 h growth	3	45	28	1134	23	0.50	4.0
				124	20	534	23	0.19	4.2
SSTR3	TZ	10 µM sst	3	47	40	942	19	0.40	3.0
				137	19	143	19	0.14	7.9
SSTR3	TZ	100 µM sst	1	22	26	1476	22	1.0	8.8
SSTR3	TZ	100 µM sst + 30 µM PS2	2	48	28	393	23	0.48	5.4
SSTR3	TZ	100 µM sst + 30 µM PS2	2	131	24	138	23	0.18	7.8
β-tubulin	Flagella	Normal	6	11	27	1960	21	1.91	10.6
				75	37	1058	21	0.28	2.6
PKD2	Flagella	Normal	2	63	25	185	54	0.86	25.7
				134	17	125	54	0.4	27.8
IFT54	Flagella	Normal	2	79	36	1330	17	0.22	1.8
KAP	Flagella	Normal	2	29	12	164	25	0.86	10.9
				77	28	147	25	0.32	7.2

Using SPEED microscopy, the three-dimensional transport routes of components of the IFT A and B subcomplexes, IFT43 and IFT20, respectively, were determined (Follit et al., 2006; Arts et al., 2011; Hirano et al., 2017). EM work has demonstrated that IFT occurs between the axonemal microtubules and ciliary membrane in Chlamydomonas flagella (Kozminski et al., 1995; Rosenbaum and Witman, 2002). Therefore, it was expected that the three-dimensional density histograms of IFT43 and IFT20 would lie in the space between the axonemal microtubules and the ciliary membrane. Indeed, the transport routes for IFT20-GFP and GFP-IFT43 were found at 105 ± 2 and 108 ± 1 nm along the cilium radius with route widths (defined as ±1 SD about the peak position in Gaussian function) of 46 ± 1 nm and 52 ± 2 nm (Figure 1, C–H). Using the published EM structural data as a reference, these localizations indicate that the IFT components localize between the microtubules and the ciliary membrane.

Next, the localizations of IFT motors, IFT cargos, and freely diffusing small molecules were validated in SPEED microscopy. For the IFT motor, the transport route for KIF3A-GFP, a component of the heterotrimeric kinesin-2 (KIF3A/KIF3B/KAP) complex that drives anterograde IFT (Kozminski et al., 1995; Pazour and Rosenbaum, 2002; Engelke et al., 2019), was determined. In agreement with our previous work in the cilium shaft (Luo et al., 2017), KIF3A was found to have two transport routes at the TZ: an outer route at 92 ± 7 nm along the cilium radius with a route width of 71 ± 17 nm, and an inner route at 0 ± 3 nm along the cilium radius with a 52 ± 2 nm route width (Figure 1, I–K). The IFT cargo α-tubulin-GFP (10, 39) also occupies two distinct transport routes in the TZ: one between the microtubules and ciliary membrane with a radial peak position at 111 ± 1 nm and a route width of 43 ± 3 nm, and the other inside the axonemal lumen with a radial peak position at 0 ± 1 nm and a route width of 98 ± 3 nm (Figure 1, L–N). It was expected that the outer transport routes of both KIF3A and α-tubulin localize between the microtubules and ciliary membrane, as was the microtubule-proximal localization of the motor in relation to IFT proteins and cargo. We hypothesize that the inner transport route of both proteins corresponds to a passive diffusion route that has been observed previously in the cilium shaft (Luo et al., 2017).

To map the passive diffusion route in the TZ, the freely diffusing molecule GFP was localized using Arl13b-mCherry as a ciliary marker and NPHP4-mCherry as a marker for the TZ (Higginbotham et al., 2012). To identify the location of NPHP4-mCherry in the presence of Arl13b-mCherry as the ciliary marker, we first photobleached the ciliary shaft. As Arl13b-mCherry is mobile, while NPHP4-mCherry is immobile, we can selectively photobleach only Arl13b-mCherry and use the intact fluorescence of NPHP4-mCherry for localizing the TZ. It is demonstrated here that GFP localizes at the center of the TZ with a peak position at 0 ± 1 nm along the cilium radius and a route width of 70 ± 1 nm (Figure 1, O–Q), which colocalizes well with the inner transport routes of the IFT motor and cargo. To demonstrate that transit of IFT motors and cargos in the lumen corresponds to passive diffusion, the ATP dependence of these targets was evaluated for impact on transport. To this end, NIH-3T3 cells were permeabilized with digitonin, which has been shown to perforate the cellular membrane selectively, causing ATP to flow out of the cell (Breslow et al., 2013). This reduces ATP levels in the primary cilium by ∼90%, inhibiting IFT (Figure 2I; Ye et al., 2013). It is shown here that within permeabilized cells, the IFT motor KIF3A and the IFT cargo α-tubulin no longer occupied the outer transport route but continued to undergo transport through the luminal route (Figure 1, R–W). Thus, the luminal transport routes for α-tubulin and KIF3A are able to persist in an energy-independent environment via passive diffusion.

FIGURE 2:

A previously uncharacterized transport route exists for membrane proteins inside the lumen of the primary cilium. (A–D) Imaging of externally-labeled AP-SSTR3-GFP. (A) Schematic of the external labeling procedure. The extracellular SSTR3 N-terminus (green) is tagged with the AP (yellow) and biotinylated (arrow) for binding to AlexaFluor647 (red)-labeled streptavidin (orange). (B) Image of primary cilium in live cells coexpressing AP-SSTR3-GFP (green) and NPHP4-mCherry (red) overlaid with two-dimensional single-molecule Alexa-Fluor-647 externally labeled SSTR3 locations (white). Scale bar: 1 μm. (C) Single-molecule AP-SSTR3-GFP locations in the TZ plotted along the R dimension in the cylindrical system. (D) Spatial representation of the histogram in C. (E–H) Imaging of SSTR3-GFP. (E) Epifluorescence microscopy image of live cell coexpressing SSTR3-GFP (green) and NPHP4-mCherry (red; scale bar: 5 µm). The dashed white line represents the cell border. (F) Enlarged image of the primary cilium overlaid with single molecule SSTR3-GFP locations (black dots). Scale bar: 1 µm. Single-molecule data and epifluorescence images were rotated together clockwise, so that TZ is parallel to the x dimension to maintain consistency with subsequent data analyses. (G) Single-molecule SSTR3-GFP locations in the TZ along the R dimension in the cylindrical system. (H) Spatial representation of the histogram in G. (I–L) Imaging of SSTR3-GFP in permeabilized cells. (I) Schematic of the permeabilization procedure, resulting in ATP diffusion from the cell. (J) Image of primary cilium in live cells coexpressing SSTR3-GFP (green) and NPHP4-mCherry (red) overlaid with two-dimensional single-molecule SSTR3-GFP locations (white) in permeabilized cells. Scale bar: 1 μm. (K) Single-molecule SSTR3-GFP locations in the TZ plotted along the R dimension in the cylindrical system. (L) Spatial representation of the histogram in K. (M, N) Same as F and (G) for AP-SSTR3-GFP in the ciliary shaft. (O, P) Same as J and K for SSTR3-GFP in the ciliary shaft in permeabilized cells.

Tracking of transmembrane proteins through the transition zone shows their localization at both the ciliary membrane and the ciliary lumen

To evaluate TM membrane transport routes at the single-molecule level, the SPEED technique previously validated in IFT transport was applied to TM proteins. This was accomplished using an externally labeled version of somatostatin receptor 3 (SSTR3), a G-protein coupled receptor that localizes to primary cilia and has critical functions in the hippocampus (Händel et al., 1999; Berbari et al., 2007; Einstein et al., 2010). The labeled SSTR3 construct, AP-SSTR3-GFP, can be externally labeled in live NIH-3T3 cells and marks the ciliary membrane (Howarth and Ting, 2008; Ye et al., 2013). This particular construct consists of an AP domain on the N terminus and a GFP on the C terminus. The AP domain is an acceptor peptide that, when expressed with the biotin ligase BirA, becomes biotinylated shortly after translation at the endoplasmic reticulum. The biotinylated AP-SSTR3-GFP is trafficked to the ciliary membrane, where it can be externally labeled using Alexa Fluor 647(AF647)-labeled streptavidin (Figure 2A; Ye et al., 2013; Ruba et al., 2018b). As expected, the AF647-labeled AP-SSTR3-GFP localized near the TZ ciliary membrane with a peak position at 131 ± 3 nm along the cilium radius and a route width of 24 ± 1 nm (Figure 2, B–D), after the estimated length of the external label is taken into consideration (Materials and Methods).

The localization of SSTR3-GFP at the TZ in live cells was also evaluated (Figure 2, E and F). Surprisingly, as is demonstrated here, SSTR3-GFP occupies two transport paths through the TZ: one near the ciliary membrane with a peak position at 129 ± 3 nm and a route width of 34 ± 5 nm, and the other inside the TZ lumen with a peak position at 55 ± 2 nm and a route width of 62 ± 6 nm (Figure 2, G and H). Based on Monte Carlo simulations, which estimate the localization error of the mean for the 3D density histograms (Supplemental Figures S7 and S8 and Table 1), the inner and outer routes are spatially resolvable with localization precisions of 2.0 and 3.2 nm, respectively. This is made possible by the collection of a sufficient number of high-resolution single-molecule localizations. A comparison of the peak fitting areas determined that transiting molecules had a 73% frequency in the inner route and 27% in the outer route. The outer transport route for SSTR3-GFP likely corresponds to the ciliary membrane, as it colocalizes with the externally labeled AP-SSTR3-GFP (Figure 2, C and D). However, the inner transport route has not been visualized in previous measurements. It seems unlikely that the inner route reflects the GFP-labeled C-terminus of SSTR3 extending into the axonemal lumen as the 102 aa C-terminal domain of SSTR3 could extend a maximum of ∼44 nm (Ainavarapu et al., 2007), given an average length of ∼ 0.4 nm/aa (4 Å) and the (2–4)-nm size of GFP, a distance that is not sufficient to bridge the ∼70–100 nm between the outer and inner transport routes. It also seems unlikely that free GFP is cleaved from SSTR3-GFP and contributes to the inner route localization, as the inner route of SSTR3-GFP and the passive diffusion route of free GFP are well separated by a distance of ∼ 50 nm. Together, these high-resolution 3D data in live cells may support a model for TM protein transport that utilizes the ciliary lumen to a greater extent than previously recognized.

Given that transit in the ciliary lumen occurs by passive diffusion for free GFP, IFT motors, and IFT cargos (Figure 1), we hypothesized that energy-independent transport may account for SSTR3 mobility in the luminal route. To test this, cells were permeabilized with digitonin and then SPEED microscopy was performed. Although permeabilization did reduce the total frequency of single-molecule events (from 32.4 ± 14.9 events/s to 10.1 ± 1.6 events/s, Supplemental Figure S2C), the location of transiting SSTR3-GFP molecules was largely unaffected (Figure 2, I–L). The majority of SSTR3-GFP molecules continued to utilize the inner route (85% compared with 73% in unpermeabilized cells). The location of the SSTR3 inner transport route showed a slight shift toward the cilium center in permeabilized cells, with a peak position at 44 ± 6 nm along the cilium radius and a route width of 66 ± 9 nm (Figure 2, K and L). For the outer transport route, permeabilization and loss of ATP increased the immobile portion and decreased the directionally moving SSTR3 molecules (determined via mean squared displacement analysis of single molecule trajectories), an observation that has been reported previously in the literature (Ye et al., 2013; Supplemental Figure S2, A and B).

To determine whether these SSTR3 transport routes occur in different regions of the primary cilium, a three-dimensional density map for SSTR3 at the basal body and the cilium shaft was generated (Supplemental Figure S3, A–G). At the basal body, using γ-tubulin-mCherry as a marker (Muresan et al., 1993), SSTR3 again exhibited two distinct transport routes: one near the outer basal body with a peak position at 120 ± 4 nm and a route width of 40 ± 6 nm and the other inside the lumen of the basal body with a peak position at 55 ± 2 nm and a route width of 54 ± 4 nm (Supplemental Figure S3, H and K). These localizations suggest that the majority of SSTR3 likely enters and/or exits the transition zone through the lumen of the basal body rather than through the gaps in the 9+0 microtubules or through diffusion along the membrane. Along the cilium shaft, two transport routes were also detected for SSTR3-GFP (Supplemental Figure S3, J and M), while only the outer route was detected for externally labeled AP-SSTR3-GFP (Figure 2, M and N).

To determine whether this paradigm is similar for other TM proteins, three-dimensional transport routes at the basal body, the TZ, and the cilium shaft were evaluated for HTR6, a GPCR that binds serotonin and localizes to primary cilia, tagged with GFP on its C-terminus (Berbari et al., 2008). It was demonstrated that HTR6 also occupies two distinct transport routes, one in the lumen and one at the ciliary membrane, with localizations similar to those of SSTR3. When compared with SSTR3, HTR6 exhibited an even larger proportion localized to the luminal route (Supplemental Figure S3, N–P). One possible explanation for this differential localization may be that HTR6 undergoes higher turnover in its targeting to the primary cilium and thus spends more time in intraciliary and intracellular transport rather than embedded in the ciliary membrane.

RAB8A, a regulator of ciliary vesicle targeting, colocalizes and cotransports with SSTR3

The luminal localization of SSTR3 and HTR6 is unexpected and suggests that current models for ciliary protein trafficking may be incomplete. In fact, this data implicates a previously uncharacterized import mechanism for TM proteins. One possible explanation for the unanticipated import localizations of SSTR3 and HTR6 is that vesicular trafficking may contribute to the luminal transport of TM proteins. As has been reported previously in the literature, vesicular trafficking components such as Rabin8 and RAB8A promote docking of ciliary vesicles to the microtubule triplets that comprise the basal body (Nachury et al., 2007; Yoshimura et al., 2007). In light of these observations, it was hypothesized that RAB8A may mediate the transport of SSTR3. To interrogate this possibility, the transport routes of GFP-RAB8A through the TZ and cilium shaft in live primary cilia were imaged and mapped via SPEED microscopy. Interestingly, it was found that RAB8A utilized two transport routes in both the TZ and the cilium shaft (Figure 3, A–F). The majority of transit events (87%) at the TZ localized to the luminal transport route with a mean position of 50 ± 6 nm along the cilium radius and a route width of 75 ± 11 nm, while the remainder of events localized at an outer transport route with a mean position of 133 ± 3 nm along the cilium radius and a route width of 40 ± 7 nm (Figure 3, A–C).

FIGURE 3:

The transport routes for RAB8A are similar to those of SSTR3. (A–C) Imaging of GFP-RAB8A in the TZ. (A) Epifluorescence microscopy image of GFP-RAB8A (green) and NPHP4-mCherry (red) overlaid with single-molecule GFP-RAB8A locations (black). Scale bar: 1 μm. (B) Single-molecule GFP-RAB8A locations plotted along the R dimension. (C) Spatial representation of B. (D–F) Imaging of GFP-RAB8A in the cilium shaft. (D) Epifluorescence microscopy image of GFP-RAB8A (green) and Arl13b-mCherry (red) overlaid with single-molecule GFP-RAB8A locations (black). Scale bar: 1 μm. (E) Single-molecule GFP-RAB8A locations plotted along the R dimension. (F) Spatial representation of E.

The 3D transport routes for RAB8A, SSTR3, and HTR6 mapped by SPEED microscopy localize at the same position along the cilium radius that is outward from the passively diffusing soluble molecules. Intriguingly, it has been noted in the literature that RAB8A largely associates with membranes and has little turnover during the vesicle transport process in live cells (Grigoriev et al., 2011). However, the data demonstrate an as yet uncharacterized transport route for RAB8A that involves the luminal transport route. One possible explanation for this seeming discrepancy is that RAB8A may be tightly associated with vesicles bearing ciliary TM proteins and cause them to hug the microtubule walls of the ciliary lumen. If this is indeed the case, the single-molecule trajectories of SSTR3 and RAB8A within primary cilia should be correlated. To evaluate whether or not SSTR3 and RAB8A follow a similar tracking route, simultaneous dual-channel tracking of SSTR3-mCherry and GFP-RAB8A in transfected NIH-3T3 cells was utilized. Further, for each individual trace, the angle difference between every step of each trajectory was calculated to evaluate the degree of comovement between SSTR3-mCherry and GFP-RAB8A. In principle, an average angle difference close to 0° indicates a high degree of correlation between the trajectories of the proteins, whereas a wide distribution of angle differences indicates no correlation between the trajectory movements. Measurements of the level of comovement between positive and negative controls were used to generate a standard curve (Supplemental Figure S4M). The positive control experiments indicate that green and red dyes attached to 100-nm beads (Supplemental Figure S4, C–F, and Video 2) have a probability of 95% for comovement, where comovement is defined as the total angle differences of their trajectories that fell within ∼37° centered on the 0° bin (Supplemental Figure S4, D–F). The negative control indicates that free fluorescein and JF561 dyes (Supplemental Figure S4, G–J) have a probability of comovement of ∼13% (Supplemental Figure S4, I and J). By referencing the standard curve (Supplemental Figure S4M), it was determined that a probability of 24% comovement is equivalent to a 15% comovement for SSTR3 and RAB8A trajectories in cilia above the negative control. Both the colocalization of the 3D transport and the degree of comovement suggest that RAB8A is correlated with the transport of SSTR3 and may potentially play a role in the luminal transport of TM proteins (Supplemental Figure S4, A, B, K, and L).

Movie S2.

Download video file ^{(4.3MB, mov)}

Dual-channel co-tracking of green and red dyes attached 100-nm beads by SPEED microcopy shown in green and red channels respectively.

Golgicide A suppresses transmembrane protein frequency equally in both routes and prevents ciliogenesis

To further evaluate the role of vesicular trafficking in SSTR3 transport within the ciliary lumen, cells were treated with varying concentrations of Golgicide A (GCA), an inhibitor of Golgi export of TM proteins that works by inhibiting GBF1 from interacting both with Arf4 and with the ciliary targeting sequence, a necessary step in sorting and export from the Golgi, thereby severely attenuating the vesicular transport of TM proteins (Figure 4A; Wang et al., 2012, 2014). NIH-3T3 cells were serum-starved in the absence or presence of GCA for 24 h, resulting in inhibition of the length and ultimately the presence of primary cilia in a dose-dependent manner (Figure 4, B–D). This indicates that membrane flux to primary cilia is an important component of ciliogenesis, particularly in relation to vesicle transport (Nachury et al., 2007). GCA treatment also reduced the frequency of single-molecule SSTR3-GFP events at the TZ in a dose-dependent manner (Figure 4E). For the remaining events, the relative density and locations of the 3D transport routes were statistically unaffected (Figure 4, F–I), indicating that a similar mechanism of entry is utilized by both the inner and outer routes. It may be that the choice of which route to take is stochastic and based on the cross-sectional area of ciliary entry, given that the space between the microtubules and the ciliary membrane accounts for ∼30% of the cross-sectional area, while the ciliary lumen accounts for ∼70%. Together, these data suggest that both the inner and outer SSTR3 transport routes are linked to vesicular transport mechanisms, with the inner route accommodating the bulk of SSTR3 transport under these experimental conditions.

FIGURE 4:

Golgicide A reduces SSTR3 frequency but does not alter transport routes. (A) Schematic outlining Golgicide A’s mechanism of action. (B, C) Epifluorescence images of SSTR3-GFP in NIH-3T3 cells treated with B, no Golgicide A or C, 5 µM Golgicide A for 24 h. Scale bars: 1 µm. (D) Bar graph showing cilia length vs. Golgicide A concentration. (0 µM n = 29, 1 µM n = 8, 5 µM n = 12, bar graph represents mean ± SE) (E) Bar graph showing SSTR3 single-molecule frequency vs. Golgicide A concentration (0 µM n = 5, 1 µM n = 3, 5 µM n = 4, bar graph represents mean ± SE) (F–H) Three-dimensional transport routes for SSTR3 in 0, 1, and 5 µM Golgicide A, respectively, plotted along the R dimension. (I) Summary of percentages of SSTR3 transport in the inner transport route for 0, 1, 5, and 10 µM Golgicide A (statistics summarized in Table 1).

The ciliary lumen is the preferred transport route for transmembrane proteins during clathrin-mediated, signal-dependent internalization

The presence of both inner and outer routes for SSTR3 transit at the TZ invites the question of whether or not the relative utilization of these routes could be affected under conditions of active GPCR signaling. Previous work has reported that SSTR3 is actively removed from primary cilia after binding somatostatin (Green et al., 2016; Ye et al., 2018). Here, it is demonstrated that after stimulation with 10 µM somatostatin for 1 h, the majority of SSTR3 single-molecule trajectories at the base of primary cilia displayed characteristics of both random diffusion and unclear movement patterns (Figure 5, D–I), while a minority displayed long, directional movement (Figure 5, A–C and Video 3). To further characterize these trajectories, each step of each trajectory was measured and the average directionality of all the trajectories for each concentration of somatostatin in a range from 0 to 100 µM was calculated. As the concentration of somatostatin increased, the average directionality of trajectories shifted from 6% into the primary cilium to 16% out of the primary cilium (Figure 5J), reflecting an overall removal of SSTR3 from the primary cilium. This observation is consistent with previously published work (Ye et al., 2018). In concert with the shift in the average directionality of the recorded trajectories, the proportion of transiting molecules in the luminal transport route also increased until 100% of molecules utilized the inner route at 100 µM somatostatin (Figure 5, K–M, O).

FIGURE 5:

Somatostatin stimulation causes an increase in the transport of SSTR3 in the inner transport route. (A) Epifluorescence image of SSTR3-GFP overlaid with single-molecule locations (white). Scale bar: 1 µm. (B) Two-dimensional scatterplot of the Gaussian fitting results of the single-molecule trajectory in A. (C) MSD vs. time plot of trajectory in A and B. (D–F, G–I) Same as (A–C) showing trajectories with different types of movement patterns. (J) Bar chart showing the average percentage directionality of every trajectory step at different concentrations of somatostatin and Pitstop 2. (0 µM sst n = 489, 1 µM sst n = 908, 5 µM sst n = 889, 10 µM sst n = 1080, 50 µM sst n = 275, 100 µM sst n = 159, 100 µM sst/30 µM Pitstop 2 n = 427, bar graphs represent mean ± SE) (K–N) Three-dimensional transport routes in the TZ for SSTR3 in 0, 10, and 100 µM somatostatin and 100 µM somatostatin with 30 µM Pitstop 2. (O) Percentage of SSTR3-GFP molecules utilizing the inner transport route from K–N (statistics summarized in Table 1).

Movie S3.

Download video file ^{(25.9KB, mov)}

The movie shows that SSTR3-GFP single molecules (white) can move directionally in the primary cilium (green) under stimulation with 10 μM somatostatin. The tip and the base indicate the orientation of the primary cilium.

The mechanism for SSTR3 removal from primary cilia upon somatostatin stimulation has been reported to involve recruitment of β-arrestin to primary cilia, which subsequently binds to SSTR3 and promotes clathrin-coated endocytosis (Oakley et al., 1999; Green et al., 2016). To investigate the role of clathrin-coated endocytosis in SSTR3 removal from primary cilia, the directionality and three-dimensional transport route of SSTR3 were measured at 100 µM somatostatin following a 1-h preincubation with 30 µM Pitstop 2, a potent clathrin-coated endocytosis inhibitor (Von Kleist et al., 2011). Pitstop 2 treatment completely blocks the dose-dependent reversal of SSTR3 directionality during somatostatin stimulation (Figure 5J) and prevents the previously observed shift of transiting SSTR3 to the lumenal transport route (Figure 5, N and O), suggesting that clathrin-coated endocytosis plays a role in facilitating movement through the inner transport route.

The ciliary lumen is the preferred transport route for transmembrane proteins during ciliogenesis

As a second test of whether the relative utilization of the inner and outer transport routes could be affected by cellular conditions, the transit of SSTR3 and HTR6 molecules was examined during different stages of ciliary growth. Due to the nonlinear growth kinetics of primary cilia (Figure 6A), time points of 1, 3, 6, 12, and 24 h following serum starvation were selected. For both SSTR3 and HTR6, the majority of transit events at the TZ occurred via the luminal transport route early in ciliary growth and shifted to include more outer transit events as the primary cilia matured (Figure 6, B–L). We used the Pearson correlation coefficient to correlate these changes, with values close to +1 or –1 indicating high degrees of positive and negative correlation, respectively, and values close to 0 indicating no correlation. Changes in percentage transit in the inner route and ciliary length had a strong negative correlation at -0.90 and -0.83 for SSTR3 and HTR6, respectively (Figure 6M). This suggests that these phenomena may arise from the same biological process that regulates ciliogenesis. The percentage transit in the inner route showed no correlation with the frequency of single-molecule events (Pearson coefficients of 0.13 and 0.15 for SSTR3 and HTR6, respectively; Supplemental Figure S5). These results suggest that, unlike structural and IFT proteins in flagellum elongation (Dirksen and Staprans, 1975; Wood and Rosenbaum, 2014), TM proteins may not be transported to the growing cilium at a higher rate earlier in ciliogenesis. Thus, the changes in percentage transit in the central transport route may be due to other factors, such as structural and selective competency of the transition zone thath forms early in ciliogenesis (Williams et al., 2011).

FIGURE 6:

SSTR3 and HTR6 use the inner transport route at a higher frequency at earlier times of ciliogenesis. (A) Growth curve of primary cilia following serum starvation. Insets: green = SSTR3-GFP, red = NPHP4-mCherry. (B–F) 3D transport routes for SSTR3 following 1, 3, 6, 12, and 24 h of serum starvation. (G–K) Same as B–F except with HTR6. (L) Summary of B–K. (M) Graph of percentage transit in inner route vs. average cilia length with Pearson’s correlation coefficient for both SSTR3 and HTR6.

The ciliary lumen accommodates similar transport routes in Chlamydomonas reinhardtii flagella

The utilization of an inner transport route by a TM protein in the TZ of murine primary cilia was unexpected and raised the question of whether such events occur in the cilia of other species. To interrogate this question, single-molecule tracking was performed in Chlamydomonas reinhardtii flagella (Video 4), enabling evaluation of the three-dimensional transport routes of several classes of protein: specifically, the IFT cargo β-tubulin (Craft et al., 2015), the TM protein PKD2 (Huang et al., 2007), the IFT protein IFT54 (Wingfield et al., 2017), and the kinesin-2 motor subunit KAP (Mueller et al., 2005). Intriguingly, the single-molecule imaging data are strikingly similar to those for murine primary cilia in terms of transport routes observed for each discrete protein. β-Tubulin, PKD2, and KAP all demonstrate inner and outer transport routes within the shaft of the Chlamydomonas flagellum (Figure 7, A–J) that roughly correlate to the same transport routes for α-tubulin, SSTR3, and KIF3A within murine primary cilia (Figure 1). While the results are largely similar, two significant differences between the systems were observed: first, radial shifts in the inner routes for β-tubulin and KAP by 11 and 29 nm, respectively. These shifts are perhaps due to the presence of the central pair of microtubules in the motile Chlamydomonas flagellum (Czarnecki and Shah, 2012). Second, a shift of the IFT54 transport route to 79 nm along the radii of flagella was observed, which is ∼25 nm more central than IFT20 and IFT43 in murine primary cilia. This discrepancy may be due to slightly different locations of the axoneme in flagella and cilia and/or different placement of IFT54 within the IFT particle than of IFT20 and IFT43. Taken together, the general location of the three-dimensional transport routes appears to be similar in the motile flagellum of Chlamydomonas and the primary cilium of mammalian cells.

FIGURE 7:

Transport routes in Chlamydomonas reinhardtii flagella closely reflect those in primary cilia. (A) Brightfield image of C. reinhardtii stably expressing β-tubulin-GFP. Red dashed box shows the region of the flagellum selected for single-molecule imaging. Scale bar: 1 µm. (B) Brightfield image from boxed region in A with dashed white lines indicating location of flagellum and superimposed with single-molecule localizations (black dots). Scale bar: 1 µm. (C–F) Three-dimensional transport routes for β-tubulin, PKD2, IFT54, and KAP, respectively, in the shaft of the flagellum. (G–J) Spatial representations of C–F, respectively. (K) Summary of various transport routes observed in a cross-sectional view of a primary cilium. (L) Table indicating which proteins utilize the transport routes depicted in K. Underlined proteins indicate transport routes determined in C. reinhardtii.

Movie S4.

Download video file ^{(468.5KB, mov)}

β-tubulin-GFP single molecules were tracked in Chlamydomonas reinhardtii flagella at 2 ms/frame.

DISCUSSION

Vesicles as transmembrane protein carriers in the ciliary lumen

Single-molecule data indicate that the lumina of primary cilia are a prominent pathway for the transport of TM proteins in two model systems. Further, the relative usage of each transport route can be regulated by a variety of ciliary states. In addition, the ciliary lumen accommodates the transport of passively diffusing soluble molecules including α-tubulin, kinesin-2, and free GFP. It is important to note that the axonemal lumen transport route is offset by ∼50 nm from the passive diffusion route on the central axis (Figure 7K). Here, results are presented that suggest that vesicle transport is involved in the inner route. Specifically, there are three key pieces of data that support this possibility: 1) colocalization with Rab8A, 2) reduction in transport in response to GCA treatment, and 3) sensitivity to Pitstop 2 treatment. Together, these data indicate that TM protein transport both correlates with a well-characterized vesicular protein and undergoes a transport dynamic shift in response to vesicle transport attenuation and inhibition. A summary of the transport routes and the proteins that traverse these routes is presented here (Figure 7, K and L).

These results characterize the ciliary lumen as a viable transport route for vesicles, a finding that has been entertained in the field based on EM studies of chondrocytes and rod photoreceptors (Jensen et al., 2004; Chuang et al., 2015). Recent work has also visualized vesicles in the amphid cilia of C. elegans and vesicles accumulated in genetically modified worms’ null for various IFT transport proteins (Doroquez et al., 2014; Li et al., 2019). However, these studies suffered from the criticism that EM work has the propensity to form vesicle-like bubbles as a result of the fixation process associated with EM imaging. Further, EM is also limited in its ability to resolve low-contrast structures such as vesicles unless vesicular proteins are specifically and densely labeled via immunogold labeling (Nachury et al., 2010). In fact, early characterization studies of SSTR3 localization in the hippocampus demonstrated dense immunogold labeling along the entire width of primary cilia (Händel et al., 1999) and electron-dense particulates filling the empty space of primary cilia (Rogowski et al., 2013). EM would be ideal for observing vesicle transport of TM proteins in primary cilia if not for these limitations. To bypass the limitations of EM imaging, a live-cell, superresolution, and fluorescence microscopy–based approach was utilized to address this question.

Further, as SPEED microscopy and its subsequent two- to three-dimensional transformation algorithm are a relatively new and cutting-edge technique, this technique was validated for its capacity to accurately reconstruct known details of protein transport in primary cilia using IFT components and the AP-SSTR3-GFP construct. These results confirmed and expanded upon the transport routes and mechanisms for TM and other proteins in primary cilia. We also examined the effects of single-molecule localization precision (Supplemental Figure S8), number of single-molecule localizations (Supplemental Figure S9), labeling efficiency (Supplemental Figure S10), and distortion (Supplemental Figure S11) on the precision of the final three-dimensional transport routes via Monte Carlo simulation (Supplemental Figure S7) for every transport route presented in this paper. Each transport route conforms to the precision standards of the biological claims presented here, according to the simulations. The results are summarized in Table 1. Furthermore, the reasonable consistency of the ciliary ultrastructure and the experimental reproducibility of each transport route are emphasized by showing the three-dimensional transport route histograms for SSTR3 in six different cilia and quantifying the error (Supplemental Figure S12).

Despite the lumen of the TZ being a relatively open and clear channel compared with the surrounding structural regions and likely free from the cartwheel structure that characterizes nonciliary procentrioles (Alvey, 1986; Hoyer-Fender, 2013), an important consideration is how vesicles may pass through the basal body, TZ, and cilium shaft. First, some vesicle distortion may occur to allow vesicle passage into and along the primary cilia. Indeed, the recycling endosome, a waypoint for TM proteins sent to the primary cilia, has a convoluted membrane structure and its vesicles are highly nonuniform (Goldenring, 2015). Second, there appears to be a requirement for proteins to mediate the transport of vesicles past the diffusion barrier into primary cilia. This is interesting, as RAB8A is responsible for mediating entry into the connecting cilia of frog photoreceptors (Moritz et al., 2001) and is shown here to occupy the same transport routes as SSTR3 in live murine primary cilia. This presents the possibility that RAB8A may play some role in the vesicular transport of SSTR3. Third, small vesicles carrying TM proteins may be able to pass through the diffusion barrier, albeit at a low frequency, due to being above the widely accepted barrier limit (Calvert et al., 2010; Kee et al., 2012; Breslow et al., 2013; Lin et al., 2013).

Previous models have suggested that the fusion of TM protein-containing vesicles likely occurs outside the primary cilia, at the periciliary base, or at the TZ ciliary membrane (Nachury et al., 2010). The results presented here do not refute these findings. Indeed, RAB8A possessed transport routes within the lumen as well as near the ciliary membrane. In addition, inhibition of TM protein vesicle export from Golgi with GCA demonstrated reduction in both the inner and outer transport routes. This suggests that vesicle transport is a component in transiting to both the inner and outer TZ locations.

The ciliary lumen as a specialized signaling transport route

Fluorescence microscopy has been utilized as a tool by several research groups to demonstrate that markers for clathrin-coated pits localize to the ciliary pocket (Molla-Herman et al., 2010). Further, other research groups have reported that β-arrestin is recruited to primary cilia upon SSTR3 stimulation and has a causal role in its removal (Green et al., 2016). Based on this evidence, it appears that β-arrestin is recruited to primary cilia following SSTR3 stimulation, attenuates SSTR3’s signaling, and facilitates transport by IFT out of primary cilia, where it promotes interaction with endocytic machinery. Our data expand upon this model by suggesting that the ciliary membrane possesses endocytic capabilities and that the internalized receptors can be transported through the axonemal lumen.

In addition, these data support a model where soluble protein entry into the ciliary lumen may occur to provide a staging area for ciliogenesis, the construction of the axoneme, and/or the construction of new IFT particles. During this aggregation within the lumen, the proteins in question would likely remain nonfunctional until they were transported to their final destination, similarly to how proteins trapped within the axonemal lumen of C. reinhardtii are nonfunctional until incorporated into their final destination (Lechtreck et al., 2013). Such a model is similar to that observed in α-tubulin during ciliogenesis, where diffusive concentration of α-tubulin is upregulated to permit more free tubulin to be incorporated into the growing end of the cilium (Craft et al., 2015; Craft Van De Weghe et al., 2020). A similar phenomenon has been observed in the assembly of IFT trains. Current thinking is that IFT trains assemble at the base of the cilium and then transport into the cilium. Therefore, a variety of components diffuse into the region until they are incorporated into a new IFT train (Wingfield et al., 2017). In light of these observations as well as the data presented within this study, such a model is probable. However, further investigation is required to determine the physiological and pathological relevance of this observation as only single cells were evaluated within this study.

The primary cilium is a specialized signaling organelle whose evolutionary progenitor is the motile flagellum (Mitchell, 2007). Over time and with selection pressure against motility in eukaryotic cells, it appears that protoprimary cilia lost the central pair of microtubules and associated motility proteins. This loss of motility function may have paved the way for an increase in signaling capabilities by reducing the geometric constraints of the flagellum. Here, we show that TM protein transport can occur at various stages of the primary cilium lifecycle—specifically, growth, steady-state, and signaling states—at least in part through the axonemal lumen. In future work, probing the differences between flagella and primary cilia and incorporating other well-known ciliary protein transport components, such as the BBSome, will be essential in developing a molecular view of the various ciliopathies.

MATERIALS AND METHODS

Request a protocol through Bio-protocol.

Tissue culture and transfection

NIH-3T3 cells or NIH-3T3 cells stably expressing NPHP4-mCherry/IFT20-GFP were grown in DMEM, high glucose, GlutaMAX Supplement (Life Technologies), 10% fetal bovine serum (Fisher Scientific), and 1% penicillin–streptomycin (Thermo Fisher) and split every 2 d to 40–50% confluency. At 24 h before imaging, the cells were transferred to glass-bottomed dishes (MatTek) and grown in OPTIMEM (Life Technologies) to induce growth of primary cilia. Transfection was performed concurrently with induction of primary cilia growth using Transit-LT1 (Mirrus) according to the manufacturer’s protocol. Before imaging, media were replaced with transport buffer (20 mM HEPES, 110 mM KOAc, 5 mM NaOAc, 2 mM MgOAc, and 1 mM EGTA, pH 7.3). For permeabilization, cells were permeabilized in glass-bottomed dishes for 2 min with 30 μg/ml digitonin in transport buffer and washed again with transport buffer. Transport buffer was supplemented with 1.5% polyvinylpyrrolidone. For external membrane labeling, biotin ligase (BirA) and AP-SSTR3-GFP were cotransfected using Transit-LT1 according to the manufacturer’s protocol except for the media, which was supplemented with 1 μM biotin. Before imaging, cells were incubated with 1 μM AlexaFluor647–streptavidin (Life Technologies) for 30 min to cause efficient binding to transiently expressed, biotinylated AP-SSTR3-GFP on the cell surface. Cells were then washed five times with PBS to efficiently wash away unbound AlexaFluor647–streptavidin before placing cells in transport buffer. For Golgicide A inhibition, cells were serum-starved as described above and the media were supplemented with the given concentration of Golgicide A. For somatostatin stimulation, cells were incubated with 10 µM somatostatin for 1 h before imaging. For Pitstop 2 inhibition of somatostatin signaling, cells were incubated with the given concentration of Pitstop 2 for 1 h before addition of somatostatin. For the growth status experiments, cells were serum-starved as described above and imaging was performed at the given time interval post–serum starvation.

Plasmids and stable cell lines

The SSTR3-GFP plasmid was a gift from Kirk Mykytyn (College of Medicine, Ohio State University). The GFP-IFT43 plasmid was a gift from Kazuhisa Nakayama (Department of Physiological Chemistry, Kyoto University). The IFT20-GFP plasmid was a gift from Gregory Pazour (University of Massachusetts Medical School). Arl13b-mCherry, KIF3A-GFP, SSTR3-mCherry, the IFT20-GFP stable cell line, and the NPHP4-mCherry stable cell line were gifts from Kristen Verhey (University of Michigan). AP-SSTR3-GFP was a gift from Maxence Nachury (Addgene plasmid #49098), GFP-RAB8A was a gift from Maxence Nachury (Addgene plasmid #24898), BirA was a gift from Alice Ting (Addgene plasmid #20856), α-tubulin-GFP was a gift from Patricia Wadsworth (Addgene plasmid#12298), and mCherry-Gamma-Tubulin-17 was a gift from Michael Davidson (Addgene plasmid #55050). The Chlamydomonas reinhardtii β-tubulin-GFP and mNG-IFT54 cell lines were a gift from Karl F. Lechtreck (University of Georgia). The Chlamydomonas reinhardtii PKD2-GFP cell line was a gift from Kaiyao Huang (Chinese Academy of Sciences). The Chlamydomonas reinhardtii KAP-GFP cell line was a gift from Mary Porter (University of Minnesota) through the Chlamydomonas Resource Center (CC-4296).

Optical setup of SPEED microscopy

The SPEED microscopy setup includes an Olympus IX81 equipped with a 1.4-NA 100× oil-immersion apochromatic objective (UPLSAPO 100×, Olympus), a 35-mW 633 nm He–Ne laser (Melles Griot), 50-mW solid state 488-nm and 561-nm lasers (Coherent), an on-chip multiplication gain charge coupled–device camera (Cascade 128+, Roper Scientific), and the Slidebook software package (Intelligent Imaging Innovations) for data acquisition and processing. For individual channel imaging, GFP, mCherry, and Alexa Fluor 647 were excited by 488-nm, 561-nm, and 633-nm lasers, respectively. The fluorescence emissions were collected by the same objective, filtered by a dichroic filter (Di01- R405/488/561/635-25 ×36, Semrock) and an emission filter (NF01- 405/488/561/635-25 ×5.0, Semrock) and imaged with the above CCD camera operating at either 500 Hz when the two- to three-dimensional transformation was performed or 100 Hz when SSTR3/RAB8A cotracking and SSTR3-GFP directionality tracking under somatostatin stimulation.

Spatial localization of primary cilia and the transition zone

A marker was identified for the transition zone. The centroid of the fluorescent spot for NPHP4-mCherry was used, since NPHP4 localizes on the arms of the Y-shaped linkers, which have ninefold symmetry and are organized in multiple layers. In situations where the tested fluorescently labeled proteins did not localize strongly to primary cilia, Arl13b-mCherry was used as a ciliary marker. Since Arl13b-mCherry molecules are mobile in primary cilia, they can be quickly prephotobleached by focusing a 561-nm laser on the ciliary shaft before localization of the NPHP4-mCherry is determined in the transition zone. Since the transition zone may range up to 1000 nm in length along the long axes of primary cilia, only single-molecule data collected at ±300 nm from the NPHP4-mCherry centroid were used in the two- to three-dimensional transformation to ensure that only single molecules moving through the transition zone were collected. To mark the basal body, the above protocol was used except that γ-tubulin-mCherry, the type of tubulin that constitutes the basal body, was used instead of NPHP4-mCherry.

Single-molecule localization precision

To determine the 2D locations and spatial localization precision for single molecules on the imaging plane, we employed both the computer program Glimpse (https://github.com/gelles-brandeis/Glimpse) and the ImageJ plugin GDSC-SMLM (http://www.sussex.ac.uk/gdsc/intranet/microscopy/UserSupport/AnalysisProtocol/imagej/gdsc_plugins/) to fit single-molecule fluorescent spots and obtain information on the spatial locations of single molecules, the integral fluorescence intensity of single molecules, the background intensity, and the Gaussian width of single-molecule spots. These single-molecule data were further filtered based on the following rules: 1) To ensure that only single molecules within the dimensions of the primary cilium were retained for two- to three-dimensional transformation analysis, images containing a Gaussian width of single molecules that was too low or too high are excluded, as they are likely not derived from a single-molecule signal, and are probably either background noise or from multiple molecules that are excited simultaneously. 2) The signal-to-noise ratio (SNR) is calculated using the signal intensity from the integral fluorescence intensity of single molecules and the background intensity. Typically, we only select single molecules with a SNR > 10, which enables us to exclude single-molecule images that might provide inaccurate localization precision.

The localization precision for immobile fluorescent molecules and moving fluorescent molecules was defined as how precisely the central point of each detected fluorescent diffraction–limited spot was determined. For immobile molecules, the fluorescent spots were fitted to a two-dimensional symmetrical or an elliptical Gaussian function, and the localization precision was determined by the SD of multiple measurements of the central point. However, for moving molecules, the influence of particle motion during image acquisition should be considered in the determination of localization precision. In detail, the localization precision for moving substrates (σ) was determined by the following formula:

where F is equal to 2, N is the photon count, a is the pixel size of the CCD camera, b is the SD of the background in photons per pixel, and

Where s₀ is the SD of Gaussian function of the mobile molecules at the focal plane and D is the diffusion coefficient of the mobile molecule (109–112). Additionally, s₀ and N of single transiting fluorescent molecules were used as selective criteria to ensure that only single molecules with high localization precision within the dimensions of the transition zone were selected.

In our experiments, we typically use >1000–2000 photons and in-focus Gaussian widths (0.5–1.0 pixel, corresponding to single GFP or mCherry molecules located in the focal plane) for accurate localization of fluorescent molecules. Using the above equation, the localization precision is determined to be ≤10–20 nm with the following experimentally defined parameters: N > 1000–2000, a = 240 nm, b ≈ 2, s₀ = 150 ± 50 nm, Δt = 0.4–2 ms, and D is in the range 0.1–3 µm²s^–1 for the various tested substrates. For the single-molecule data used in two- to three-dimensional transformation, the average single-molecule localization precision for every experiment is summarized in Table 1.

Two- to three-dimensional density transform algorithm

The transformation process used to compute the three-dimensional spatial probability density maps of particles transiting through the TZ has been described in detail in our previous publications and is demonstrated again here in Supplemental Figure S6. In short, the three-dimensional spatial locations of molecules transiting through the NPC can be considered in either Cartesian (x, y, z) or cylindrical (x, r, θ) coordinates. In microscopic imaging, the observed two-dimensional spatial distribution of particle localizations is a projection of actual three-dimensional spatial locations onto the XY plane. Since primary cilia are symmetrical about their long central axes, the two-dimensional spatial distribution is averaged about those axes. The underlying three-dimensional spatial distributions can then be recovered by projection of the measured Cartesian (x, y) coordinates back onto the simplified cylindrical (x, r, constant) coordinates, on the basis of the expected cylindrically symmetrical distribution along the θ direction of the primary cilium.

Effects of deviations from the rotational symmetry on three-dimensional density map reconstruction

As detailed in our previous publications (Ruba 2018a, 2019) and demonstrated in Supplemental Figures S10 and S11, the effects of deviations from a normal rotational symmetry adopted by biological structures on our three-dimensional density map reconstruction have been fully considered by thoroughly exploring the possible deviations. First, we assessed how imperfect rotational symmetries affected the accuracy of three-dimensional density map reconstruction. Second, we evaluated how partial labeling of the structure affected the ability to reconstruct an accurate three-dimensional density map distribution. Based on these determinations, we found that the three-dimensional density map reconstruction algorithm based on the back projection is still sufficiently robust and valid under significant deformation, partial labeling, or incomplete rotational symmetry (Ruba 2018a, 2019).

Estimation of spatial probability percentage for transiting molecules traveling on each transport route

To calculate the spatial probability of single molecules for a protein with two transport routes, the integrated area of the density histogram for each transport route was calculated and expressed as a percentage of the total integrated area of all transport routes to estimate the probability of detecting a single molecule on that given transport route.

SSTR3 and RAB8A comovement analysis

To maximize the chance of detecting of any comovement between SSTR3 and RAB8A above control levels in live cells, we cotransfected plasmids containing SSTR3-mCherry and GFP-–RAB8A into NIH-3T3 cells and used a dual channel filter set to split the individual fluorescence illuminated by simultaneous 561 and 488 nm lasers from each construct onto individual halves of the CCD detector. After alignment of the two halves of the detector, comoving trajectories in each channel were selected for analysis when at least two consecutive single molecules appeared in each channel, within acceptable WI and localization precision bounds, and no further than 200 nm from each other during each step of the trajectory. This 200-nm requirement was implemented to control for two trajectories that were further than a biologically reasonable distance. Two hundred nanometers is the approximate diameter of the axoneme plus single-molecule localization precision and, therefore, the maximum size of a potential vesicle carrier entering primary cilia. When these requirements were implemented, the positive control (dual-labeled 100-nm Tetraspeck beads in 55% glycerol to mimic approximate diffusion constant of in vivo molecules) comovements were highly correlated and the negative control (fluorescein and JF561 dye in 92.5% glycerol) comovements were largely uncorrelated. Glycerol was used to adjust the viscosity of the medium and, thus, the diffusion coefficient of the above particles using the equation: , where D is the diffusion coefficient, k is the Boltzmann constant, T is temperature, η is viscosity, and R is the radius of the particle. The laser powers for the controls were adjusted to produce localization precisions comparable to experimental conditions. A standard curve was developed from these controls and used to characterize the results in vivo (Supplemental Figure S4M).

Calculation of diffusion coefficient and α value

Calculation of the diffusion coefficient was performed by first plotting each trajectory (>4 frames) on a mean squared displacement (MSD) versus time (t) plot. The data were fitted with the function MSD = 4Dt^α, where α is a measurement of graph skewedness. α represents directional movement (or superdiffusion), passive diffusion, or subdiffusion if its value is greater than 1.1, between 0.9 and 1.1, or less than 0.9, respectively.

Determination of cilia movement, cilia orientation, and axial position

In our experiments, we typically needed 2 min to complete imaging of the entire cilium and collection of 10 single-molecule videos (5000 frames per video and 2 ms per frame) from a primary cilium. Combination of 24-h serum-starving cell growth and incubation of cells in transport buffer before microscopy imaging made the shift of primary cilia in live cells less than 5 nm during the 10-s detection time for each video. As for the cilia orientation, an epifluorescence image of Arl13b-mCherry, SSTR3-GFP, HTR6-GFP, or GFP-RAB8A labeled cilium provided a complete image of entire cilia, which clearly indicated the ciliary base, the ciliary tip, and the cilia orientation. Then after prephotobleaching of the tracked protein to locally reduce its concentration, the point illumination of SPEED microscopy generated two-dimensional superresolution spatial distribution of protein molecules moving within a range of ∼1 μm along the ciliary axis in the shaft or TZ of primary cilia. Consequently, the precise location of the middle axis of a primary cilium was obtained by determining the peak position of these two-dimensional superresolution spatial locations in the ciliary radial dimension with fitting of the Gaussian function as well as the Gaussian fitting of the ciliary fluorescence in the image of the ciliary marker.

Determination of the location of the transport route for SSTR3 labeled externally

The length of the linker was determined by summing the estimated lengths of all components of the external label: SSTR3 N terminus:acceptor peptide:biotin:streptaviding:3xAlexaFluor647 (Howarth and Ting, 2008). The SSTR3 N terminus is 43 amino acids long with a marginal level of disorder according to IUPred. To estimate the length, the average (9.6 nm) between the fully disordered length (43 a.a., 0.4 nm/a.a.) and the fully globular diameter (2 nm) was taken (Dosztányi et al., 2005; Ainavarapu et al., 2007; Erickson, 2009). According to IUPred, the AP domain had a consistent structured prediction. Thus, the globular diameter was used (1.5 nm). Using the average length of a C–C bond, biotin’s length was estimated to be ∼1 nm. Last, the globular diameter of tetravalent streptavidin was estimated to be 5 nm. However, three of the four binding domains, on the average, are occupied by biotin-conjugated AlexaFluor647. Since the average center of fluorescence will be roughly in the middle of the tetravalent streptavidin molecule, 2.5 nm was used for its length. Thus, the total length of the external tag is ∼14.6 nm.

Determination of spatial localization precision of transport route

As detailed in Supplemental Figure S7, Monte Carlo simulations were performed where varying numbers of single-molecule locations were randomly simulated on an ideal radius (R_I). Then localization error (σ_LE) was added in the y and z dimensions by sampling an error value from a normal distribution with a SD of σ_LE. Subsequently, the 3D transformation algorithm was performed on only the y-dimensional data to model the loss of z-dimensional information during the two-dimensional microscopy projection process. The peak position of the transformed three-dimensional density histogram was then determined by Gaussian fitting to produce a measured radius (R_M), which may deviate from the R_I due to the number of simulated locations and simulated localization error. After 10,000 iterations of this process, 10,000 R_M values were collected and the resulting histogram of R_M values could be obtained. After 10,000 iterations, the mean of the R_M values converges on R_I, the ideal radius from which all the simulations were originally sampled, while the SD of the R_M values varies based largely on the number of simulated locations and σ_LE. As shown in Table 1, the localization precision and precision/radius ratio of each transport route in primary cilia is high enough to distinguish it and make biologically relevant conclusions possible.

In addition, all studies indicate that primary cilia have two states of protein transport: 1) ciliary growth state (ciliogenesis) and 2) maintenance state. The process of ciliary growth takes place over the 24 h of serum starvation. Thus, the 2-min imaging time of our approach should be sufficient to distinguish even the smallest transition states of protein transport in primary cilia.

Statistics

Experimental measurements were reported as mean ± standard error of the mean unless otherwise noted.

Code availability

The code for the simulations, two- to three-dimensional transformation, and sample and experimental data are available at https://github.com/YangLab-Temple.

Abbreviations used:

EM

electron microscopy

GCA

Golgicide A

GFP

green fluorescent protein

IFT

introflagellar transport

SPEED

single-point edge-excitation subdiffration

SSTR3

Somatostain Receptor 3

TM

transmembrane

TZ

transition zone