Binding of IFT22 to the intraflagellar transport complex is essential for flagellum assembly

Abstract

Intraflagellar transport (IFT) relies on motor proteins and the IFT complex to construct cilia and flagella. The IFT complex subunit IFT22/RabL5 has sequence similarity with small GTPases although the nucleotide specificity is unclear because of non‐conserved G4/G5 motifs. We show that IFT22 specifically associates with G‐nucleotides and present crystal structures of IFT22 in complex with GDP, GTP, and with IFT74/81. Our structural analysis unravels an unusual GTP/GDP‐binding mode of IFT22 bypassing the classical G4 motif. The GTPase switch regions of IFT22 become ordered upon complex formation with IFT74/81 and mediate most of the IFT22‐74/81 interactions. Structure‐based mutagenesis reveals that association of IFT22 with the IFT complex is essential for flagellum construction in Trypanosoma brucei although IFT22 GTP‐loading is not strictly required.

Introduction

Cilia (also known as flagella) are important organelles needed for cell motility, morphogenesis, sensory perception, and several signaling pathways, such as sonic hedgehog and PDGFRα signaling (Huangfu et al, 2003; Kohl et al, 2003; San Agustin et al, 2015; Salinas et al, 2017; Schneider et al, 2005). Cilia are tail‐like appendages protruding from the cell surface of nearly every eukaryotic cell type and are found on various unicellular organisms and on almost all cells in the mammalian body. For example, the protist Trypanosoma brucei, commonly known as the parasite causing sleeping sickness, carries a motile flagellum that is required for development and disease pathogenesis (Ralston et al, 2009; Langousis & Hill, 2014; Rotureau et al, 2014).

A microtubule‐based axoneme extending from the centriole‐like basal body at the ciliary base is the central shape‐giving element of cilia. The organelle is surrounded by the ciliary membrane, which is a continuous outgrowth of the plasma membrane but hosts a unique composition of lipids and membrane proteins (Emmer et al, 2010; Serricchio et al, 2015). To date, more than 600 different proteins have been identified to reside in the ciliary compartment (Pazour et al, 2005). Cilia construction as well as maintenance of the organelle in almost all organisms relies on a conserved active transport process termed intraflagellar transport (IFT; Kozminski et al, 1993; Rosenbaum & Witman, 2002). Intraflagellar transport particles are thought to be responsible for the selective transfer of ciliary cargo proteins from the cytoplasm through the diffusion barrier at the transition zone. IFT is dependent on the motor proteins kinesin II for anterograde (base to tip; Cole et al, 1993, 1998; Prevo et al, 2015) and dynein 2 for retrograde (tip to base) movement (Pazour et al, 1999; Porter et al, 1999; Signor et al, 1999) of cargo proteins and turnover products. The IFT complex is required for construction of the flagellum and likely serves important functions in ciliary cargo selection and transport (Bhogaraju et al, 2013b) and can be divided into biochemically distinct IFT‐A and IFT‐B sub‐complexes, consisting of 6 and at least 16 individual proteins, respectively (Piperno & Mead, 1997; Cole et al, 1998; Taschner & Lorentzen, 2016b). The IFT‐B complex is organized into two stable sub‐complexes, the 10‐subunit IFT‐B1 (IFT22, IFT25, IFT27, IFT46, IFT52, IFT56, IFT70, IFT74, IFT81, IFT88; Lucker et al, 2005; Follit et al, 2009; Ishikawa et al, 2014; Taschner et al, 2014) and the 6‐subunit IFT‐B2 complex (Taschner et al, 2016). While inactivation of IFT‐B complex components or the kinesin motor typically leads to defects in cilium construction due to disrupted anterograde IFT (Pazour et al, 2000; Absalon et al, 2008), IFT‐A protein or dynein deletions produce phenotypes associated with malfunctioning retrograde transport (Pazour et al, 1999; Blacque et al, 2006). Mutations in IFT components and other ciliary proteins are the cause for a wide range of genetic diseases and developmental abnormalities, known as ciliopathies (Reiter & Leroux, 2017).

For the assembly of large complexes that bind diverse cargo proteins, most IFT proteins are composed of protein–protein interaction domains such as coiled‐coils, β‐propellers, and tetratricopeptide repeats (Taschner et al, 2012). However, the two IFT complex members IFT22 (Rabl5) and IFT27 (Rabl4) show significant sequence homology to small GTPases of the Rab family, which are key regulators of vesicular membrane‐trafficking (Stenmark, 2009; Itzen & Goody, 2011). IFT22 and IFT27 share low sequence identity (< 15%) and may play regulatory roles in IFT (Schafer et al, 2006; Qin et al, 2007; Adhiambo et al, 2009; Bhogaraju et al, 2011). In mammalian cilia, IFT27 was shown to be required for exit of the BBSome complex and associated ciliary cargoes (Keady et al, 2012; Eguether et al, 2014). Recently, Rabl2 was identified as a potential third Rab‐like member of the IFT complex as it was shown to regulate IFT initiation and to associate with the IFT74/81 sub‐complex (Lo et al, 2012; Kanie et al, 2017; Nishijima et al, 2017). Interestingly, IFT22 and IFT27 also associate with the IFT74/81 sub‐complex (Taschner et al, 2014) suggesting that IFT22, IFT27, and Rabl2 may be located within close proximity in the IFT B1 complex. IFT22, IFT27, and RabL2 are unusual Rab GTPases as they lack the C‐terminal prenylation motif commonly found to associate Rab GTPases with membranes.

Previous studies classified IFT22 as an atypical small GTPase with a high degree of sequence variance from classical Rab proteins, particularly in sequences assigned to the conserved nucleotide‐binding pocket (Schafer et al, 2006; Adhiambo et al, 2009). IFT22 lacks the conventional G4 motif and contains a highly diverse G5 motif required for interaction with the guanine base of GTP/GDP (Rensland et al, 1995; Vetter & Wittinghofer, 2001; Itzen & Goody, 2011). Hence, it is unclear if IFT22 can specifically bind guanine nucleotides. Interestingly, in vivo studies in several ciliated organisms revealed functional differences of IFT22 between species. Mutation of the Caenorhabditis elegans (Ce) IFT22 homolog (called IFTA‐2) does not affect cilium formation or IFT, but worms show deficiencies in the DAF‐2 (insulin‐IGF‐1‐like) signaling pathway, leading to an extended lifespan and abnormalities in dauer stage formation (Schafer et al, 2006; Blacque et al, 2018). In contrast, RNAi knockdown experiments of IFT22/Rabl5 in Trypanosoma brucei (Tb) led to a retrograde IFT inactivation phenotype that is characterized by short flagella filled with IFT material (Adhiambo et al, 2009). In Chlamydomonas reinhardtii (Cr), IFT22 was shown to control the cellular levels of both IFT‐A and IFT‐B proteins and to regulate availability of particles participating in IFT (Silva et al, 2012). Intriguingly, IFT22 homologs are missing in the genomes of Giardia intestinalis and Tetrahymena thermophila, although IFT is present in these ciliated organisms, whereas Drosophila melanogaster lacks both IFT22 and IFT74/81 homologs (van Dam et al, 2013).

In this study, we provide insights into nucleotide specificity of IFT22, the molecular basis of incorporation into the IFT complex and a dissection of the in vivo function of IFT22 using structure‐based mutations in T. brucei. We show that IFT22 specifically binds G‐nucleotides and present the crystal structures of GTP‐ and GDP‐bound IFT22, which identify a new, unusual binding mode for G‐nucleotides in the absence of the classical G4 motif. The crystal structure of the trimeric IFT22/74/81 complex provides a molecular basis for IFT22 incorporation to the IFT complex via the switch regions of IFT22 and a heterodimeric coiled‐coil region of IFT74/81. In vivo experiments using structure‐based IFT22 mutants in T. brucei demonstrate that association of IFT22 with IFT‐B1 is essential for ciliogenesis.

Results

IFT22 specifically binds GDP/GTP

Due to the unusual G4/G5 regions, it was unclear if IFT22 is a selective guanine nucleotide‐binding protein or if IFT22 may bind other purine nucleotides such as ATP (Espinosa et al, 2009; Taschner et al, 2012). To address nucleotide specificity, we overexpressed and purified TbIFT22 and removed nucleotides retained during the purification by urea treatment and refolding (Appendix Fig S1A–C). We then measured the affinities of apo TbIFT22 for GTP and GDP in titration experiments with fluorescently labeled non‐hydrolyzable GTP/ATP derivatives (mant‐GMPPNP/mant‐AMPPNP) or GDP (mant‐GDP). TbIFT22 bound the GTP analog with a K_d of 2 μM and GDP with a K_d of 20 μM (Fig 1A, left and middle panels). These weak μM affinities are in the same range as reported for GTP/GDP‐binding by IFT27 (Bhogaraju et al, 2011) and suggest that nucleotide exchange does not necessarily require a guanine nucleotide exchange factor (GEF), as it is the case for some large GTPases (Uthaiah et al, 2003). No binding was observed for the ATP analog (Fig 1A, right panel). We therefore conclude that IFT22 is a specific guanine nucleotide‐binding protein.

Figure 1. IFT22 associates with guanine nucleotides through an unusual G5‐dependent mechanism.

1. IFT22 nucleotide‐binding experiments. Fluorescence measurements using increasing amounts of TbIFT22 and TbIFT22/74_342–401/81_397–450 core complex incubated with mant‐labeled GDP (mant‐GDP) or non‐hydrolysable GTP/ATP analogs (mant‐GMPPNP/mant‐AMPPNP). The fluorescence intensity is plotted as a function of protein concentration. Data were fitted to a single‐site binding equation for determination of the dissociation constant (K_d). K_d values and standard deviations are calculated from three independent experiments.
2. Structural comparison of GTP‐bound HsRab8A (light purple) and TbIFT22 (green) depicted in cartoon representation. Nucleotides are shown as sticks and Mg²⁺ as balls. Unstructured regions of TbIFT22 are represented with dotted lines. The zoomed‐in view shows a superposition of the nucleotide‐binding pocket. While classical GTPases form hydrogen bonds between a conserved aspartate of the G4 motif (NKxD) and the guanine base, IFT22 instead utilizes D175 located in the G5 loop.
3. Topology diagrams of a classical Rab GTPase and of IFT22. Positions of the conserved nucleotide‐binding G‐motifs (G1–G5) as well as switch regions are indicated.
4. Top: Cartoon representation of IFT22 (gray) with positions of two nucleotide‐binding mutants highlighted. GTP is shown in stick representation and Mg²⁺ as a ball. D175 (blue) is the unusual residue binding the guanine base (see also Fig 1B), while S19 (pink) is a conserved residue required for coordination of the Mg²⁺ cation and is commonly mutated to an asparagine to prevent nucleotide binding. Bottom: Nucleotide‐binding experiments of IFT22 nucleotide‐binding mutants D175E, D175A, and S19N with fluorescently labeled nucleotides. Only the S19N mutation (light pink) abolished IFT22 nucleotide‐binding ability completely.
5. Superposition of GTP (green)‐ and GDP‐bound (light green) TbIFT22 structures. Switch regions are marked in yellow and dotted lines indicate disordered loops not modeled in the structures.

We also measured the affinities for mant‐GMPPNP and mant‐GDP of IFT22 in context of the TbIFT22/74/81 core complex (TbIFT22/74_342–401/81_397–450), which demonstrated a modest increase in affinities when compared to IF22 alone (Fig 1A, left and middle panels). To confirm these results, IFT22 or IFT22/74/81 core complexes from Tb, M. musculus (Mm) or C. reinhardtii (Cr) were incubated with excess of GTP and the content of bound nucleotides analyzed after size‐exclusion chromatography (SEC) using an HPLC‐based system (Appendix Fig S1E). IFT22/74/81 core complexes from all three species bound GTP, albeit to a different degree. The core complex from T. brucei incorporated the highest percentage of GTP, followed by C. reinhardtii and last M. musculus. Notably, TbIFT22 bound less GTP than the TbIFT22/74/81 core complex and no nucleotide could be detected for MmIFT22, which likely reflects that MmIFT22 has lower affinity for GTP than the Tb and CrIFT22 proteins. These results show that GTP‐binding is a conserved property of IFT22 across species and confirm that the IFT22/74/81 core complex has higher affinity for GTP than IFT22 alone.

Next, we analyzed the intrinsic GTPase activity of TbIFT22 and detected very low but measureable hydrolysis rates for both TbIFT22 and the TbIFT22/74/81 core complex (Appendix Fig S1F) comparable to reported intrinsic hydrolysis rates of other small GTPases (Simon et al, 1996; Scheffzek & Ahmadian, 2005; Bhogaraju et al, 2011). Thus, if GTP turnover is required for the cellular function of IFT22, a GTPase activating protein (GAP) is required to stimulate nucleotide hydrolysis.

Structures of IFT22 with GTP or GDP reveal the molecular basis of guanine specificity

To address the molecular basis of nucleotide binding by IFT22, we crystallized TbIFT22 with co‐purified GTP and determined the structure at 2.3 Å resolution (Fig 1B and Table 1). To obtain a GDP‐bound structure, TbIFT22 was treated with urea, dialyzed to remove bound nucleotides, and refolded in the presence of GDP. The IFT22‐GDP structure was determined at 2.5 Å resolution, and the electron density clearly supports the presence of GDP (Appendix Fig S1G, compare left and middle panels). As expected, IFT22 exhibits the overall fold of a Rab GTPase, containing a mixed six‐stranded β‐sheet surrounded by α‐helices (Fig 1B, right image). However, in contrast to classical GTPases that contain five α‐helices, IFT22 lacks the α4 helix between β5 and β6 (Fig 1C). When the IFT22 structure is compared to protein structures currently available in the protein data bank using the Dali server (Holm & Sander, 1993), IFT22 is most similar to structures of other Rab family GTPases with Homo sapiens (Hs) Rab8A as the closest match (PDB ID: 4lhw), superposing with a root mean square deviation (rmsd) of 2.4 Å (see Fig 1B).

Table 1.

Data collection and refinement statistics

	TbIFT22‐GTP	TbIFT22‐GDP	TbIFT22/7479–401/811–450‐GTP (SeMet)
PDB code	6IA7	6IAE	6IAN
Data collection
Wavelength (Å)	1.00000	0.97891	0.97899
Resolution range (Å)	48.37–2.30 (2.38–2.30)	48.52–2.49 (2.57–2.49)	82.67–3.20 (3.40–3.20)
Space group	P 61	P 61	P 21
Unit cell (Å, °)	a = 55.85	a = 56.02	a = 68.56
	b = 55.85	b = 56.02	b = 228.30
	c = 263.45	c = 263.09	c = 115.71
	α = 90	α = 90	α = 90
	β = 90	β = 90	β = 96.76
	γ = 120	γ = 120	γ = 90
Total reflections	205,477 (17,635)	321,430 (27,213)	2,191,763 (226,031)
Unique reflections	20,689 (1,996)	16,300 (1,558)	113,747 (18,991)
Multiplicity	9.9 (8.8)	19.7 (17.5)	19.3 (11.9)
Completeness (%)	99.7 (96.7)	99.3 (96.1)	100.0 (100.0)
Mean I/sigma	16.7 (0.9)	18.5 (0.9)	11.5 (0.7)
CC1/2	0.999 (0.497)	0.998 (0.363)	0.999 (0.395)
Refinement
Number of reflections	20,588	16,236	57,455
Protein residues	295	305	1,637
Number of atoms	2,311	2,354	12,090
R‐work	0.212 (0.292)	0.220 (0.359)	0.241 (0.441)
R‐free	0.264 (0.349)	0.244 (0.385)	0.280 (0.447)
Ramachandran favored (%)	93.5	94.0	96.3
Ramachandran outliers (%)	0.0	0.35	0.25
RMS bonds (Å)	0.005	0.006	0.007
RMS angles (°)	0.90	1.0	1.1
Average B‐factors (Å2)	64	72	143

Statistics for the highest resolution shell are shown in parentheses.

The high degree of sequence divergence of IFT22 when compared to other Rab GTPases (Appendix Fig S2) translates into a highly unconventional nucleotide‐binding mode in IFT22. The classical G4 NKxD motif, which is missing in IFT22, features an aspartate residue (D124 in HsRab8A, see Fig 1B, detailed view) that interacts with the base of the nucleotide thus providing specificity for guanine over adenine (Rensland et al, 1995; Paduch et al, 2001). In the absence of a G4 aspartic acid, TbIFT22 uses D175 from the unusual G5 motif to form a bifurcated hydrogen bond with the guanine moiety (see detailed view in Fig 1B). While the classical G4 motif is positioned in a loop connecting β5 with α4, Asp175 is located between β6 and α5* (see Fig 1C). IFT22 homologs from D. rerio (Dr) and mammals (Mm, Hs) have a glutamate residue in the position of TbIFT22 D175 (Appendix Fig S2) indicating a potentially similar binding mechanism in those species. Cr and C. elegans (Ce) IFT22 contain an alanine and a glycine, respectively, at the position of TbIFT22 D175 making it unclear how or if they achieve specificity for G‐nucleotides.

To evaluate the importance of D175 in GTP/GDP‐binding, we purified D175E and D175A mutant forms of TbIFT22 and carried out titrations with mant‐GMPPNP or mant‐GDP (Fig 1D). While the TbIFT22_D175A mutant did not show any detectable nucleotide‐binding, the TbIFT22_D175E mutant bound mant‐GMPPNP and mant‐GDP with K_D values of 18 μM and 139 μM, respectively, which is approximately one order of magnitude lower affinity than wild‐type TbIFT22. Interestingly, TbIFT22_D175A did bind mant‐GMPPNP when in context of the IFT22_D175A/74/81 core complex with a K_D of 102 μM, which is approximately two orders of magnitude lower affinity than what we observed for the wild‐type IFT22/74/81 core complex (Fig 1A and D). These results are in agreement with higher GTP affinity of the IFT22/74/81 core complex compared to IFT22 alone. Our data demonstrate that D175 is important for nucleotide binding in TbIFT22 and suggest that E175 can contact the guanine moiety of GTP/GDP although in a less favorable manner than D175 likely due to steric problems caused by the longer side‐chain. We also introduced the classical S19N mutation in TbIFT22 that prevents Mg²⁺ coordination and thus abolishes nucleotide binding. As expected, our titration data show that TbIFT22_S19N does not associate with mant‐GMPPNP (Fig 1D).

A hallmark of small GTPases are the switch regions that typically undergo major conformational changes between the active GTP‐bound and the inactive GDP‐bound states, which allow for binding of effectors (Vetter & Wittinghofer, 2001; Mourão et al, 2014). Surprisingly, no major conformational changes were observed when comparing the GTP‐ and GDP‐bound states of IFT22 and the switch regions are unstructured in both GDP‐ and GTP‐bound TbIFT22 structures (Fig 1E). While switch I and II of GDP‐bound GTPases are known to be rather flexible and often unstructured, active GTP‐bound forms usually exhibit ordered switch regions that provide a stable interaction surface for downstream effector binding. The switch regions of GTP‐bound TbIFT22 are thus not in a pre‐ordered conformation ready to associate with effectors. However, the observation that the IFT22/74/81 core binds nucleotides with higher affinity than IFT22 alone does suggest that IFT74/81 may interact with and stabilize the nucleotide‐binding pocket of IFT22.

Structure of the IFT22/74/81 complex

To elucidate how IFT22 is incorporated into the IFT‐B1 complex and determine if IFT74/81 is an effector of IFT22, we set out to obtain a structure of IFT22/74/81. IFT74 and IFT81 are both predicted to contain mostly coiled‐coil structures (Fig 2A) and share 26% sequence identity in T. brucei suggesting that IFT74 and IFT81 are distant homologs. IFT74 and IFT81 interact directly with each other to form a binding platform for the IFT‐B1 components IFT22, IFT25/27, and IFT46/52 (Lucker et al, 2005; Taschner et al, 2011, 2014). In addition to the coiled‐coil regions, IFT74/81 contains an N‐terminal tubulin‐binding module contributed by both proteins (Bhogaraju et al, 2013a). Since IFT22/74/81 core complexes (TbIFT22/74_342–401/81_397–450, Appendix Fig S1D) did not yield crystals, we co‐expressed and purified longer constructs of TbIFT74/81 with TbIFT22, spanning the N‐terminal predicted coiled‐coil domains. The positively charged IFT74 N‐terminus is prone to degradation and was consequently removed resulting in the TbIFT74_79–401 construct. Whereas complexes lacking the IFT81 CH domain did not crystallize, the GTP‐bound TbIFT22/74_79–401/81_1–450 complex containing the IFT81 CH domain crystallized and the structure was determined at 3.2 Å resolution by experimental phasing (Appendix Fig S3A–C and Table 1).

Figure 2. Structure of the Tb IFT22/74/81 complex.

1. Domain organization of IFT81, IFT74, and IFT22. Numbers refer to the T. brucei protein sequence and indicate different constructs used in this study. The part of the IFT81/74 sequence shown in shaded colors is not part of the construct used for structure determination. Coiled‐coil boundaries are depicted based on the structure (ccI‐ccVI) or prediction from the PCOILS webserver (cc). (CH = calponin homology, cc = coiled‐coil).
2. Crystal structure of TbIFT22/74_79–401/81_1–450 in two perpendicular orientations shown in cartoon representation. GTP is shown as a stick model. IFT22 is depicted in green, IFT74 in orange, and IFT81 in gray. Coiled‐coils are labeled ccI‐ccVI.
3. Zoomed‐in view of the N‐terminal TbIFT81 CH domain (gray) superposed onto the CrIFT81 CH domain (brick‐red). Basic tubulin‐binding residues are highlighted in yellow and light orange, respectively.
4. Zoomed‐in view of the IFT22‐binding site on IFT74/81 ccVI with ordered switch regions of IFT22 depicted in yellow.

The TbIFT22/74_79–401/81_1–450 crystal structure reveals an elongated coiled‐coil complex with the IFT81 CH domain and the IFT22 GTPase located at opposite ends (Fig 2B). Rather than forming one long coiled‐coil, the IFT74_79–401/81_1–450 structure can be subdivided into six separate heterodimeric coiled‐coil regions (ccI to ccVI) separated by short loop regions (Fig 2A and B). Boundaries of these coiled‐coils do not match particularly well with predicted coiled‐coils from the PCOILS webserver (Alva et al, 2016). IFT74 and IFT81 interact intimately and share a large buried surface interface of 8,300 Å², constituting both interactions within the coiled‐coils and between different heterodimeric regions (Fig 2B and Appendix Fig S4). Whereas ccI and ccVI protrude from either end of the complex to interact with the IFT81 CH domain and IFT22, respectively, the central four coiled‐coil regions, ccII‐ccV, form a highly compact structure held together by interactions between ccII‐ccIII, ccIII‐ccIV, and ccII‐ccIII‐ccV (Appendix Fig S4). IFT74/81 ccII‐ccV appears to form a rather unique compressed spring‐like structure. Searches using the Dali server did not reveal any structures similar to IFT74/81 ccII‐ccV in the protein data bank.

The position of the N‐terminal IFT81 CH domain is fixed to IFT74/81 ccI through contacts with the 15‐residue linker region and the bent C‐terminal helix of the IFT81 CH domain (Appendix Fig S5E). This C‐terminal helix, which is bent in IFT81 CH domains (Appendix Fig S5A and B), adopts a straight conformation in the two MT‐binding CH domain containing proteins NDC80 and EB1 (Slep & Vale, 2007; Ciferri et al, 2008; Appendix Fig S5C and D). This observation provides a molecular rationale for the different architecture of the IFT74/81 and the NDC80/NUF2 complexes (Alushin et al, 2010; Appendix Fig S5F). Interestingly, many of the positively charged Arg/Lys residues previously shown to mediate αβ‐tubulin cargo binding in the CrIFT81 CH domain (Bhogaraju et al, 2013a) are found in structurally conserved positions in the TbIFT81 CH domain (Fig 2C). These tubulin‐binding residues point toward ccII‐ccIV perhaps suggesting that tubulin cargo could be sandwiched in the gap between the CH domain and ccII‐IV (Fig 2B and C).

Noteworthy, upon IFT74/81 association the switch regions of IFT22 become ordered and mediate binding to ccVI of IFT74/81 (Fig 2D). As observed for the IFT22/74/81 core complex, IFT22/74_79–401/81_1–450 also co‐purified with GTP (confirmed by HPLC) and was set up for crystallization with a molar excess of GTP at 4°C. Although the guanine base only displays partial electron density, the ribose and tri‐phosphate moieties have clear electron density confirming that GTP is bound in the nucleotide‐binding pocket of IFT22 (Appendix Fig S1G, right panel). The observation that the switch regions are ordered in the GTP‐bound IFT22/74/81 complex structure but not in GTP‐bound IFT22 shows that binding of IFT22 to IFT74/81 induces a fixed conformation of switch I and II (Fig 2D). There are no direct contacts between GTP and IFT74/81 suggesting that the increased nucleotide affinity of the IFT22/74/81 core complex compared to IFT22 alone (Fig 1A and E) is an indirect effect of fixing the switch regions in a conformation with higher nucleotide affinity.

The switch regions of IFT22 interact with a conserved surface patch contributed by both IFT74 and IFT81

Analysis of the TbIFT22/74/81 complex structure reveals a relatively small (710 Å² buried surface) mixed hydrophobic/hydrophilic interface of between IFT22 and IFT74/81 (Fig 3A and B). Switch I and II contribute most of the IFT22 residues to the interface with IFT74/81 with a few additional residues contributed from the β‐sheet of the core GTPase fold (Fig 3B). Both IFT74 and IFT81 interact with IFT22 although IFT81 contributes about twice as many residues to the interface with IFT22 as IFT74 does (Fig 3A and B, and Appendix Fig S3D). A high degree of evolutionary conservation of residues in the interface between IFT22 and IFT74/81 (Fig 3A and Appendix Fig S3D) suggests that IFT22 associates with IFT74/81 in a similar manner in other ciliated organisms. To confirm this notion, we show that the IFT22‐IFT74/81 interaction interface between Chlamydomonas and Trypanosoma is conserved to such a degree that TbIFT22 efficiently pulls down a purified CrIFT25/27/74/81 complex, thereby forming a stable pentameric IFT‐B1 chimera (Fig 3C). The prevention of nucleotide‐binding via the TbIFT22_S19N mutant reduced the amount of CrIFT25/27/74/81 pulled down by His‐tagged TbIFT22 to background levels (Fig 3C). We conclude that IFT22, using mainly switch I and II, interacts with IFT74/81 to form an evolutionarily conserved IFT‐B sub‐complex.

Figure 3. Molecular basis of IFT22 association with IFT74/81 ccVI .

1. Cartoon representation of the IFT22‐binding site on IFT74/81 ccVI (top left) and surface conservation representation of different orientations of IFT22/74/81 (top right and bottom). IFT74/81 ccVI displays a highly conserved patch at the IFT22‐binding interface (black dashed circle). A 180° rotation of IFT22 (bottom right) exhibits a likewise conserved patch at the IFT74/81‐binding interface (black dashed circles; position of the IFT74/81 helices is marked with light gray lines). Conserved residues are marked and labeled according to the Tb sequence. Conservation coloring is based on Clustal Omega multiple sequence alignments with H. sapiens, M. musculus, D. rerio, T. brucei, C. reinhardtii, and C. elegans sequences (see Appendix Fig S2 and Appendix Fig S3D) and ConSurf conservation grades (Landau et al, 2005; Sievers et al, 2011).
2. Detailed view of the IFT22‐IFT74/81 binding site in two perpendicular orientations showing interacting residues in stick representation. Residues provided by IFT22 switch I and II are shown in yellow, whereas IFT22 non‐switch region interactions are colored in green. IFT74 is shown in orange and IFT81 in gray. IFT22 residue where mutation abolishes interaction with IFT74/81 are highlighted in red (R43 and A86, see also Fig 4 and Appendix Fig S3E).
3. SDS–PAGE gel of a Ni²⁺‐NTA pull‐down using His‐tagged TbIFT22 (WT and mutants) and untagged CrIFT25/27/74/81. WT TbIFT22 is able to pull down the Cr tetrameric complex, thus forming a chimeric IFT‐B1 pentamer, while both the A86R and S19N mutant fail to bind the complex.

Source data are available online for this figure.

To verify the interaction observed in the IFT22/74/81 structure and devise mutants for in vivo analysis, we constructed 7 different point mutations of TbIFT22 residues and evaluated complex formation by co‐expression and pull‐down of the TbIFT22/74/81 core complex (Appendix Fig S3E). Two point mutations of well‐conserved IFT22 residues abolished complex formation, namely R43E and A86R. R43 is located adjacent to switch I and A86 is located in switch II (highlighted in red in Fig 3C). Mutation of A86 to an arginine inserts a long, charged amino acid side‐chain predicted from the structure to result in steric clashes with the IFT74 helix of ccVI and prevents complex formation with IFT74/81 (Fig 4D and Appendix Fig S3E). The R43 side‐chain forms a salt bridge with E432 of IFT81 and the R43E charge reversion disrupts core complex formation, whereas a R43A mutation leads to weaker binding without completely disrupting complex formation (Appendix Fig S3E). These results identify residues vital for the interaction of IFT22 with IFT74/81 and corroborate the interaction observed in the crystal structure (Fig 3).

Figure 4. IFT74/81 interaction analysis using different IFT22 mutants.

1. SDS–PAGE gel of a Ni²⁺‐NTA pull‐down using His‐tagged IFT74_342–401/81_397–450 and untagged IFT22 (WT and mutants). Pull‐downs were done from cell lysates of co‐expressed proteins. Lanes 1–4 show similar total expression levels of the different co‐expressed constructs (input samples). Lanes 5–8 show pull‐down elutions. The IFT22_S19N (inactive GTPase mutant) and the IFT22_A86R mutant (IFT74/81‐binding mutant) did not interact with the IFT74_342–401/81_397–450 complex in the pull‐down experiment.
2. SDS–PAGE gel of a Ni²⁺‐NTA pull‐down using the His‐tagged IFT74_79–401/81_1–450 and untagged IFT22 (WT and mutants). Pull‐downs were done from cell lysates of co‐expressed proteins. Lanes 1–4 show similar total expression levels of the different co‐expressed constructs (input samples). Lanes 5–8 show pull‐down elutions. The IFT22_S19N (inactive GTPase mutant) and the IFT22_A86R mutant (IFT74/81‐binding mutant) did not interact with the IFT74_79–401/81_1–450 complex in the pull‐down experiment.
3. SDS–PAGE gel of a Ni²⁺‐NTA pull‐down of full‐length IFT25/27/74/81 complex with His‐tagged IFT22 (WT and S19N mutant). The nucleotide‐binding deficient mutant IFT22_S19N interacts with IFT25/27/74/81 in the pull‐down experiment.
4. SDS–PAGE gel of a Ni²⁺‐NTA pull‐down of full‐length IFT25/27/74/81 complex with His‐tagged IFT22 (WT and A86R mutant). The IFT74/81‐binding mutant IFT22_A86R fails to pull down the full‐length tetrameric complex.

Source data are available online for this figure.

IFT22 GTP‐binding is not a strict requirement for IFT complex formation or Trypanosoma flagellum construction

Effectors preferably interact with the switch regions of the active GTP‐bound state of small GTPases (Vetter & Wittinghofer, 2001), typically with three orders of magnitude higher affinity than with the GDP‐bound state (Leung & Rosen, 2005). The IFT22/74/81 structure presented here demonstrates that GTP‐bound IFT22 interacts with IFT74/81 mainly using the switch regions (Fig 3B), which suggests that IFT74/81 is an effector of IFT22. However, the fact that GDP‐ and GTP‐bound TbIFT22, in the absence of IFT74/81, adopt very similar structures where switch I and II are disordered (Fig 1E) shows that GTP‐bound IFT22 does not adopt a structure pre‐ordered for IFT74/81‐binding, but that the switch region conformation observed in the IFT22/74/81 structure is a result of an induced fit upon complex formation.

To further analyze the nucleotide requirement for IFT22 association with IFT74/81, we investigated different nucleotide‐binding mutants (Fig 4). First, we examined the IFT22_D175A mutant that binds GTP with about two orders of magnitude lower affinity than wild‐type IFT22 (Fig 1D). Surprisingly, the wild‐type and the IFT22_D175A mutant pulled down similar amounts IFT74/81 (Fig 4A and B). However, as HPLC experiments showed that the IFT22_D175A/74/81 core complex still co‐purified with GTP, we also tested the IFT22_S19N mutant where Mg²⁺ binding is disrupted and GTP‐binding completely abolished in our titration experiments (Fig 1D). TbIFT22_S19N was unable to assemble into a core complex with TbIFT74_342–401/81_397–450 (Fig 4A) but could still interact to some degree with the TbIFT74_79–401/81_1–450 complex used for structure determination (Fig 4B) and almost to the same level as wild‐type IFT22 in case of the TbIFT25/27/74/81 complex (Fig 4C). These results may suggest that the helices of IFT74/81 ccVI are not stably associated in the absence of the C‐terminal parts of IFT74/81. Indeed, we observed pronounced degradation of IFT81_1–450 in context of the IFT74_79–401/81_1–450 complex when expressed in the absence of IFT22 (Appendix Fig S3F). We conclude that GTP‐binding by IFT22 is not absolutely required for the interaction with IFT74/81 but appears to modulate the affinity of the interaction.

To examine the impact of the IFT22 nucleotide binding in vivo, a functional complementation assay was developed. RNAi knockdown of IFT22 in trypanosomes results in the assembly of short flagella accumulating IFT proteins as expected (Adhiambo et al, 2009). This phenotype was rescued by the expression of an RNAi‐resistant version of IFT22 fused to GFP (GFP::IFT22_rescue) that localizes and traffics normally in the flagellum (for details, see supplemental text, Movies EV1 and EV2, and Appendix Fig S6). The IFT22_D175A mutation was investigated in this context. Western blot analysis showed that GFP::IFT22_D175A was expressed and resistant to silencing in contrast to the endogenous IFT22 protein (Appendix Fig S6C). IFT22_D175A traffics normally in the flagellum in the presence (Movie EV3) and absence (Movie EV4) of the endogenous IFT22 protein. The fluorescent signal of GFP‐IFT22_D175A was strong in the flagellum as also observed by kymograph analysis (Appendix Fig S6D). Finally, Immunofluorescence (IFA) showed that these cells display the classic distribution of IFT proteins. However, they assemble flagella of slightly shorter length (Appendix Fig S6E and F). Next, we addressed the question if the GTPase cycle plays a role during IFT in T. brucei. The IFT22_S19N mutant is unable to bind GTP (Fig 1D) and displays weaker binding to IFT74/81 in vitro (Fig 4). A GFP‐tagged RNAi‐resistant version of IFT22_S19N was expressed in trypanosomes. Western blot analysis confirmed efficient silencing of the endogenous IFT22 protein, whereas GFP‐IFT22_S19N remained present (Fig 5A). IFT22_S19N was found at the base of the flagellum and trafficked normally within the organelle in the presence (Movie EV5) or the absence (Movie EV6) of endogenous IFT22, as formally demonstrated by kymograph analysis (Fig 5C). We noticed that IFT trains tended to pause and change speed more frequently in the latter case, suggesting a mild disruption of IFT. Although at first sight most cells looked normal, a minority of cells possessed clearly shorter flagella. This was confirmed by IFA analysis with an axonemal marker (Fig 5B, second column, arrowheads) and some accumulation of IFT172 occurred in these cells (Fig 5B, last column). The length of the flagellum was measured and revealed that ~10% of the population had shorter flagella than normal (Fig 5D). Moreover, statistical analysis (Anova test) revealed a significant difference (P < 0.001) in the length of the flagellum between non‐induced cells expressing IFT22_S19N compared to the non‐mutated version (Fig 5D). This indicates a dominant‐negative effect of IFT22_S19N on the length of the flagellum. This was also observed in induced conditions when comparing with the IFT22 rescue cell line. From these results, we conclude that nucleotide binding by IFT22 is not an absolute requirement for IFT or flagellum construction in T. brucei, although minor perturbations of IFT and the formation of shorter flagella do suggest a somewhat impaired function of the IFT22_S19N mutant compared to wild‐type IFT22.

Figure 5. The IFT22_S19N mutant shows a mild ciliogenesis phenotype in vivo .

1. Western blot analysis of the IFT22 ^RNAi + GFP::IFT22 _S19N cell line probed with the anti‐IFT22 antibody (bottom) and with an anti‐paraflagellar rod (PFR) as loading control (top).
2. IFA of the indicated trypanosome cell lines using the mAb25 (marker for the axoneme, left panels) and an anti‐IFT172 antibody (marker for IFT, right panels). The top panels show the phase contrast images merged with DAPI (cyan) that stains nuclear and mitochondrial DNA. Scale bars correspond to 5 μm.
3. Kymographs showing the movement of the GFP::IFT22_S19N in the presence (left) or the absence (right) of the IFT22 endogenous protein. Note the improved signal‐to‐noise ratio in the absence of endogenous IFT22. P and D mark the proximal and distal ends of the flagellum, respectively. Scale bars are 5 s (time, vertical) and 5 μm (length, horizontal)
4. Dot plot representation of flagellum length in the indicated cell lines and conditions. For each condition, 100 flagella were measured and statistical analysis was performed with the ANOVA test. Significant differences are indicated with a star (P < 0.0001).

Source data are available online for this figure.

Association of IFT22 with IFT74/81 is essential for flagellum assembly in trypanosomes

To investigate if association of IFT22 with IFT74/81 is a requirement for IFT and flagellum construction in trypanosomes, we further investigated the TbIFT22_A86R mutant that was unable to assemble into IFT22/74/81 core complexes (Fig 4A). TbIFT22_A86R did not interact with TbIFT74_79–401/81_1–450 in pull‐down experiments (Fig 4B). In case of the full‐length TbIFT25/27/74/81 complex that still interacted with TbIFT22_S19N (Fig 4C), the amount pulled down by TbIFT22_A86R was close to background levels demonstrating that the IFT22_A86R‐IFT74/81 interaction was severely impaired. We next investigated the consequences of the TbIFT22_A86R mutation for flagellum formation in vivo in T. brucei cells. An RNAi‐resistant version of GFP::IFT22 carrying the A86R mutation was expressed in trypanosomes in the tetracycline‐inducible IFT22 ^RNAi cell line. Western blot analysis showed the expected size for the fusion protein as well as efficient and specific silencing of the endogenous version of IFT22 (Fig 6A). In both non‐induced and induced conditions (leading to knockdown of the endogenous IFT22 protein), TbIFT22_A86R does not display IFT, fails to localize to the flagellum, and accumulates throughout the cytoplasm (Movies EV7 and EV8). Phase contrast microscopy showed the emergence of cells with tiny flagella filled with IFT172 protein or even no flagella (Fig 6B, last column). Transmission electron microscopy (TEM) analysis was performed on tetracycline‐induced IFT22_rescue and IFT22_A86R cell lines. It revealed that the base of the flagellum was properly inserted in the flagellar pocket but that the flagella were very short and contained excessive amount of electron‐dense material (Fig 6C, first two columns). The transition zone was properly assembled and displayed normal morphology (Fig 6C, third column) including the typical collarette that surrounds its proximal part (Trépout et al, 2018). By contrast, sections through the flagella revealed abnormal microtubule organization and excessive IFT material (Fig 6C, last two columns). This corresponds to the typical phenotype for IFT22 RNAi silencing (Fig 6C, second row) (Adhiambo et al, 2009) and confirms that the IFT22_A86R protein cannot rescue the phenotype. These results demonstrate that association of IFT22 with the IFT complex via IFT74/81 is crucial for proper flagellum organelle assembly in trypanosomes.

Figure 6. The IFT22_A86R mutant displays a severe retrograde IFT phenotype in vivo .

1. Western blot analysis of the IFT22 ^RNAi + GFP::IFT22 _A86R cell line probed with the anti‐IFT22 antibody (bottom) and with an anti‐BiP as loading control (top).
2. IFA with the indicated cell lines using the mAb25 (marker for the axoneme, central panels) and an anti‐IFT172 antibody (marker for IFT, bottom panels). The top panels show the phase contrast images merged with DAPI (cyan) that stains nuclear and mitochondrial DNA. The arrowheads indicate the presence of short flagella stained with the Mab25 antibody in the IFT22 ^RNAi cells. Scale bars correspond to 5 μm.
3. Sections of IFT22 ^RNAi + GFP::IFT22 _rescue (top panels) or IFT22 ^RNAi + GFP::IFT22 _A86R cells (bottom panels) were analyzed by transmission electron microscopy. Sections through the flagellar pocket, the transition zone and the flagellum are shown. Scale bars are 500 nm (flagellar pocket sections) or 200 nm (transition zone and flagellum sections). The white arrow indicates an endocytic vesicle budding off the flagellar pocket, whereas the black one points at an IFT train. (Axo = axoneme, TZ = transition zone, BB = basal body).

Source data are available online for this figure.

Discussion

Here we show that IFT22 specifically binds G‐nucleotides through an unusual mechanism with μM affinity and has a low intrinsic GTP hydrolysis rate. GTP hydrolysis rates and affinities for nucleotides are comparable to the ones reported for IFT27, another small GTPase of the IFT complex (Bhogaraju et al, 2011), and indicate the need for a GAP, but not necessarily a GEF protein for realization of a complete GTPase cycle (Rensland et al, 1995; Vetter & Wittinghofer, 2001; Itzen & Goody, 2011). Studies in mouse and trypanosomes demonstrated that GTP‐binding by IFT27 is needed for association with the IFT particle and that IFT27 mutants unable to bind GTP are excluded from the cilium (Eguether et al, 2014; Huet et al, 2014), suggesting that the IFT complex is an effector of IFT27. Interestingly, Eguether and colleagues found that the MmIFT27^T19N GTP‐binding mutant retains some affinity for IFT‐B in the absence of endogenous IFT27 and can enter the cilium and partially rescue the IFT27 knockout phenotype. These observations suggest that GTP‐loading of IFT27 is not a strict requirement for IFT complex association but rather modulates the affinity of the interaction of IFT27 with IFT74/81. In sensory neurons of C. elegans worms, the IFT22_T42N mutant designed to preferentially bind GDP over GTP lost its ciliary localization and was delocalized through the cytoplasm (Schafer et al, 2006). As the IFT22_T42N mutant was expressed in the presence of the endogenous IFT22 protein, the observed mis‐localization of IFT22_T42N could be a result of lower affinity for the IFT complex compared to the GTP‐loaded wild‐type version. Consistent with this, we observe that the GTP‐binding by TbIFT22 is not strictly required for IFT complex formation but does appear to modulate TbIFT22‐IFT74/81 complex formation. In vitro pull‐down experiments suggest that GTP‐bound TbIFT22 has higher affinity than the nucleotide‐free TbIFT22_S19N mutant for IFT74/81. The role of GTP hydrolysis by IFT27 and IFT22 (if any), possibly assisted by yet to be identified GAPs, remains to be identified.

Given that IFT22 co‐purified with the IFT‐B complex from trypanosomes (Franklin & Ullu, 2010), the phenotypic defect in retrograde transport upon RNAi knockdown was rather unexpected (Adhiambo et al, 2009). Here, we formally demonstrate that the phenotype is specific because it can be rescued by the expression of an RNAi‐resistant version of the gene. Similar results were obtained for IFT25 and IFT27 that associate with the IFT‐B complex and whose inhibition results in defects in entry of IFT dynein and IFT‐A proteins in the trypanosome flagellum, possibly explaining the retrograde phenotype (Huet et al, 2014, 2019). The molecular basis for the retrograde IFT phenotype observed when IFT22 is knocked down or prevented from interacting with the IFT particle, and thus, entering the cilium is currently unknown although it is possible that IFT22, IFT25, and IFT27 cooperate in flagellum IFT dynein import. Curiously, IFT25 and IFT27 are not required for the construction of mouse primary cilia or cilia in the trachea (Keady et al, 2012; Eguether et al, 2014) but are essential for formation of the sperm flagellum (Liu et al, 2017; Zhang et al, 2017), suggesting that the requirement for some IFT proteins could be variable from one cell type to the other, even in the same organism. To our knowledge, the function of IFT22 has only been investigated in T. brucei and C. elegans. It will be interesting to see whether it behaves like IFT25/27 or more classic IFT proteins in mammalian cells.

IFT81 Short‐Rib Polydactyly Syndrome mutation may affect IFT22 incorporation into the IFT complex

Although no patient mutations in IFT22 have been reported to date, a recent study identified a series of mutations in IFT81 causing Short‐Rib Polydactyly Syndrome (SRPS) (Duran et al, 2016). One of the disease mutations reported was an in‐frame deletion of amino acid L435, which corresponds to L443 in trypanosomes and is a well‐conserved residue positioned directly in the interaction interface with IFT22 (Appendix Fig S7). L435 deletion could result in an overall IFT81 protein instability, but given its structural position and interaction with IFT22 residues (Appendix Fig S7) a likely molecular rationale for the observed ciliopathy phenotype is that the L435del leads to dissociation or weakened binding of IFT22 to IFT74/81. Unfortunately, no cultured cells were available for the L345del mutant and Duran and colleagues could thus not provide experimental data regarding expression levels and stability of this IFT81 mutant protein. Interestingly, many of the SRPS‐causing mutations affect proteins required for retrograde IFT such as dynein‐2 components (Dagoneau et al, 2009; Taylor et al, 2015) or the IFT‐A component IFT121 (Mill et al, 2011), which result in retrograde IFT inactivation phenotypes. Intriguingly, knockdown of IFT22 in trypanosomes also causes a retrograde IFT phenotype (Adhiambo et al, 2009) consistent with the notion that weakened IFT22 association with the IFT complex could underlie the SRPS disease phenotypes observed in the patient with the IFT81 L345del mutation.

Materials and Methods

Recombinant protein expression and purification from E. coli

Wild‐type and mutant (A86R, S19N, D175A/E) IFT22 proteins from T. brucei and Mus musculus were expressed as tobacco etch virus (TEV) cleavable N‐terminal His₆ fusion proteins in E. coli BL21(DE3) grown in TB‐medium at 37°C. Overexpression was induced at 18°C at an OD₆₀₀ of 1.8 with 0.5 mM IPTG. Cells were lysed by sonication in a buffer containing 50 mM Tris pH 7.5, 150 mM NaCl, 10% (v/v) glycerol, 10 mM imidazole, 2 mM MgCl₂, 5 mM β‐mercaptoethanol, 1 mM PMSF, and 25 μg/ml DNaseI, and the extract was cleared by centrifugation (4°C, 75,000 g, 30 min). In a first step, proteins were purified via a Ni²⁺‐NTA affinity column (5 ml, Roche). In order to remove N‐terminal His₆‐tags, proteins were incubated with TEV protease overnight at room temperature and dialyzed against 50 mM NaCl buffer for subsequent ion‐exchange chromatography (5 ml HiTrap Q sepharose, GE Healthcare). For further purification, proteins were subjected to size‐exclusion chromatography (SEC) after concentrating to 20–30 mg/ml in a buffer containing 10 mM HEPES pH 7.5, 150 mM NaCl, 2 mM MgCl₂, and 1 mM DTT using a HiLoad Superdex 75 column (GE Healthcare). In general, both Tb and Mm (but not Cr) IFT22 were highly soluble and could be concentrated up to 2 mM (approx. 50 mg/ml). Proteins were stored at −80°C in SEC buffer. IFT22/74/81 core complexes of Tb, Mm and Cr as well as the N‐terminal TbIFT22/74_79–401/81_1–450 sub‐complex were co‐expressed in E. coli BL21(DE3) with each protein on a separate plasmid using N‐terminal His₆‐tagged IFT74 and IFT81 constructs and untagged IFT22. The same purification procedure was followed for IFT22. Expression and purification of CrIFT25_1–136/27/74_128‐C/81 was done as described previously (Taschner & Lorentzen, 2016a).

The IFT74 sequence from extracted genomic trypanosome DNA contains an insertion corresponding to an additional 6 amino acids to the published sequence in the TriTrypDB database (Tb927.7.3370, 596 residues; Aslett et al, 2010) in the N‐terminal part of the protein. This RPGSQM insertion is a repetitive sequence that is present in three consecutive copies in the annotated TbIFT74 sequence, but in four copies in our TbIFT74 construct. Since this part of the protein is predicted to be unstructured, we assume that this is a natural protein variant that does not affect IFT74 function. The IFT74 residue numbering in this publication will refer to this 602 residue‐protein version.

Expression of selenomethionine derivatives

Selenomethionine derivative proteins were obtained from co‐expression cultures of TbIFT22/74_79–401/81_1–450 grown in M9 minimal medium supplemented with 60 mg/l selenomethionine. Overnight expression was induced at an OD₆₀₀ of 1.0 with 0.5 mM IPTG, and the temperature was shifted to 20°C. The purification procedure was followed as for the native proteins.

Recombinant protein expression and purification from insect cells

Coding sequences of Tb IFT25, IFT27, IFT74_79‐C, and IFT81 were cloned as TEV‐cleavable N‐terminal His₆ fusion proteins into the multiple cloning sites of pFL vectors, with IFT25/27 and IFT74_79‐C/81 being located on the same vector, respectively (IFT27 into MCS1 via SmaI/SphI and IFT25 into MCS2 via EcoRI/XbaI; IFT81 into MCS1 via SmaI/SphI and IFT74 into MCS2 via EcoRI/XbaI). Recombinant baculoviruses were produced as described previously (Taschner et al, 2014). TbIFT25/27 and TbIFT74_79‐C/81 heterodimeric complexes were co‐expressed at 26°C in 6 l of HighFive insect cells (Invitrogen) infected with pre‐determined amounts of recombinant viruses. Cells were harvested after 72 h and lysed by dounce homogenization in a buffer containing 20 mM HEPES pH 7.5, 250 mM sucrose, 10 mM KCl, 1.5 mM MgCl₂, 5 mM β‐mercaptoethanol, and one pill of protease inhibitor cocktail (complete, EDTA‐free, Roche). Nuclei were removed as described in Taschner et al (2016). Protein purification was done as outlined for proteins expressed in E. coli, except for using a HiLoad Superdex 200 or Superose 6 column in the SEC step.

Crystallization of GTP/GDP‐TbIFT22 and TbIFT22/74_79–401/81_1–450

TbIFT22 was set up for crystallization at 15.2 mg/ml in SEC buffer by sitting‐drop vapor diffusion in 0.2 μl drops obtained by mixture of equal volumes of protein and crystallization solution. Crystals appeared after 2 days at 4°C as fine needle clusters after mixing with 20% (w/v) PEG3350, 50 mM NaCacodylate pH 6.5, and 200 mM calcium acetate and turned into three‐dimensional hexagons over the course of 10 days. No excess of GTP was added to the protein or the crystallization solution since TbIFT22 was purified bound to GTP from E. coli. For crystallization of the GDP‐loaded state, refolded nucleotide‐free TbIFT22 was set up at 15.6 mg/ml in SEC buffer supplemented with 7 mM GDP by sitting‐drop vapor diffusion in 0.2 μl drops obtained by mixture of equal volumes of protein and crystallization solution. Crystals grew with a similar shape transition as described above at 4°C after mixing with 15% (w/v) PEG6000, 50 mM NaCacodylate pH 7.0, and 200 mM CaAcetate. Both GTP‐ and GDP‐TbIFT22 crystals were cryoprotected in mother liquor containing 15% (v/v) glycerol prior to flash freezing in liquid nitrogen. Crystals of the TbIFT22/74/81 complex (native and selenomethionine derivate, Appendix Fig S3C) were obtained from protein concentrated to 25 mg/ml by sitting‐drop vapor diffusion at 4°C in 0.2 μl drops (0.1 μl protein solution containing 2 mM GTP + 0.1 μl crystallization solution) supplemented with 40 nl freshly prepared microseeds. Crystals grew after mixing with 15% (v/v) glycerol, 7.5% (w/v) PEG4000, and 100 mM HEPES pH 7.5 and were cryoprotected in reservoir solution containing 33% (v/v) ethylene glycol prior to flash freezing in liquid nitrogen.

Data collection and crystal structure determination

For the structures of the small GTPase, diffraction data were collected at the PXIII (for GTP‐TbIFT22) and PXII (for GDP‐TbIFT22) beamline at the Swiss Light Source (SLS) in Villigen, Switzerland, and were processed with XDS (Kabsch, 2010) prior to scaling with Aimless of the CCP4 package (Winn et al, 2011). The structure of GTP‐TbIFT22 was solved at 2.3 Å resolution by molecular replacement (MR) with an ensemble of three different superposed Rab GTPases found by HHpred search (PDB IDs: 1vg8, 2y8e, 3oes) using the program Phaser (Storoni et al, 2004). The asymmetric unit contained two molecules of IFT22 and analysis with Xtriage detected twinned data. The model was completed by iterative cycles of model building in COOT (Emsley et al, 2010), followed by refinement in PHENIX (Adams et al, 2010) using NCS restraints and applying the twin law h, ‐h‐k, ‐l. The GDP‐TbIFT22 structure was determined at 2.5 Å resolution using the previously solved GTP‐bound structure as a search model for MR. X‐ray diffraction data for the TbIFT22/74/81 complex structure were collected at the PXII beamline at SLS, indexed with XDS, and scaled with the CCP4 program Aimless. The structure was determined from selenomethionine substituted protein crystals. Single anomalous dispersion data were recorded at the Se peak wavelength, and AUTOSOL as part of the PHENIX package was used to locate Se sites and calculate experimental phases and electron density. The structure was modeled and refined at 3.2 Å resolution from a dataset derived from a selenomethionine substituted protein crystal, since native crystals diffracted significantly worse. Two copies of the TbIFT22/74_79–401/81_1–450 complex are present in the asymmetric unit. The 3.2 Å model was built in COOT and refined in PHENIX using NCS and secondary structure restraints. The two copies were very similar in most parts, but showed significant conformational differences in the C‐terminal IFT22‐binding coiled‐coils of IFT74/81 (ccVI). While we could build IFT22 into the electron density map of one copy of the complex, we were unable to build IFT22 with confidence in the second copy of the complex. Data collection and refinement statistics are summarized in Table 1.

Affinity pull‐down experiments

For purified proteins, Ni²⁺‐NTA affinity beads were pre‐incubated with buffer containing 150 mM NaCl, 50 mM Tris pH 7.5, 2 mM MgCl₂, and 10 mM imidazole. 40 μg of purified His‐tagged proteins was bound to 30 μl of beads in a total volume of 500 μl at 4°C. After 1 h, beads were washed twice with 1 ml buffer to remove excess protein and were incubated with 400 μg of untagged interaction partner in 500 μl total volume for another hour. Beads were washed three times with 1 ml buffer to remove unbound protein. Bound proteins were eluted from the beads with 50 μl buffer containing 500 mM imidazole. In the case of Ni²⁺ pull‐downs from cell lysates, the proteins were co‐expressed from separate plasmids in E. coli BL21(DE3) cells. 20 μl of each culture was taken and supplemented with SDS loading dye as “total expression samples”. Cell pellets from 10 ml overnight culture were resuspended in 1.5 ml lysis buffer (50 mM Tris pH 7.5, 150 mM NaCl, 10% (v/v) glycerol, 10 mM imidazole, 2 mM MgCl₂), and cells were lysed by sonication (1 min, 1 s pulse/1 s pause). Cell extracts were cleared by centrifugation (4°C, 16,000 g, 30 min), and the supernatant was incubated at 4°C with 20 μl Ni²⁺‐NTA affinity beads pre‐incubated with lysis buffer. After 1 h, beads were washed three times with 1 ml buffer and bound proteins were eluted with 50 μl buffer containing 500 mM imidazole. Eluate contents were analyzed by SDS–PAGE.

Protein denaturation by urea and refolding for nucleotide removal

TbIFT22 and the TbIFT22/74/81 core complex were refolded in order to remove bound GTP, since more gentle methods such as EDTA or SAP treatment were not successful (Appendix Fig S1B). Proteins were diluted to 0.5 mg/ml and dialyzed in a dialysis tube against buffer containing 50 mM Tris pH 7.5, 150 mM NaCl, 10% (v/v) glycerol, and 8 M urea overnight at 4°C. After 18 h, dialysis tubes were transferred to fresh buffer without urea for protein refolding and dialyzed for another 24 h. The buffer was exchanged twice to remove residual urea. After refolding, proteins were concentrated and subjected to SEC. Successful nucleotide removal was verified by HPLC.

HPLC nucleotide analysis

Nucleotide species of purified proteins and their hydrolysis states were verified at 20°C by reversed phase high‐performance liquid chromatography (HPLC) using a Vydac 218TP C₁₈ column with a Securityguard filter cartridge system (Phenomenex) attached. Nucleotides were separated by isocratic elution at 20°C with a buffer composed of 100 mM potassium phosphate pH 6.5, 10 mM tetrabutylammonium bromide, and 7.5% (v/v) acetonitrile and elution detected at 254 nm.

Nucleotide‐binding experiments

Nucleotide affinities of TbIFT22 (WT, mutants, and core complex) were determined by fluorescence spectrophotometric measurements (PerkinElmer LS50B) of 2′(3′)‐O‐(N‐methylanthraniloyl)‐labeled (mant‐labeled) nucleotides (Jena Bioscience). Increasing concentrations (2–200 μM) of nucleotide‐free protein (confirmed by HPLC) were incubated with 1 μM mant‐GDP/‐GMPPNP/‐ADP/‐AMPPNP for 30 min in a buffer containing 50 mM Tris pH 7.5, 100 mM NaCl, and 5 mM MgCl₂ in 60 μl volumes. Emission spectra of the samples were monitored at 20°C in a quartz cuvette from 400–500 nm (excitation at 355 nm). Intrinsic protein fluorescence and mant‐nucleotide background fluorescence were substracted from the data. Emission maxima of the mant fluorophore at 448 nm were plotted against protein concentrations. Curve fitting and dissociation constant (K_d) determination was done with GraphPad Prism 6.0 software using a binding equation that describes a single‐site binding model.

GTPase assay

GTPase activities of TbIFT22 and the TbIFT22/74/81 core complex were measured at 20°C with the EnzChek Phosphate Assay Kit (ThermoFisher). Reactions were initiated by adding 1 mM GTP to the protein mixed with kit solutions according to the manufacturer's recommendations. The release of inorganic phosphate (P_i) upon GTP hydrolysis followed by an enzymatic reaction was monitored over 20 min. The change in absorption at 360 nm was detected every minute using a PerkinElmer Lambda19 UV spectrometer. As a negative control, intrinsic GTP hydrolysis in buffer was followed. Rate quantifications were done with the help of a linear standard curve for P_i generated with defined concentrations of KH₂PO₄ from 10 μM to 200 μM after 20‐min incubation.

Trypanosome cultures and transfection

Procyclic T. brucei cell lines were derivatives of the strain 427, grown in SDM79 medium containing 10% fetal calf serum and hemin (Brun & Schönenberger, 1979). Generation of the inducible IFT22 ^RNAi (RABL5 ^RNAi) cell line has been described previously (Adhiambo et al, 2009). In this cell line, a 447‐nucleotide long fragment of IFT22 was cloned in the pZJM vector (Wang et al, 2000). The two T7 promoters face each other and can be induced in the presence of tetracycline, leading to the production of double‐stranded RNA (dsRNA). To express RNAi‐resistant versions of IFT22, the entire nucleotide sequence of IFT22 was modified by substituting the last and when possible the second nucleotide of the codon to render the transcript insensitive to RNAi (Huet et al, 2014) hence retaining the original amino acid sequence. The construct was tagged with GFP, and the resulting plasmid was called pPCPFReGFPIFT22^RNAiRes. GeneCust Europe carried out the chemical synthesis, and additional point mutations were introduced to generate the S19N, A86R, and D175A versions. The plasmids were linearized with NsiI to target integration in the PFR2 locus (Adhiambo et al, 2009) following transfection using the Nucleofector Technology (Lonza, Italy; Burkard et al, 2007).

Immunofluorescence

Cultured cells were spun at 580 × g to remove the supernatant and then washed in SDM79 medium without serum. Cells were spread onto poly‐L‐lysine‐coated slides, dehydrated, and fixed in methanol at −20°C for 5 min. Slides were rehydrated in 1× phosphate‐buffered saline (PBS) for 15 min. Primary antibodies were diluted in PBS + 0.1% bovine serum albumin (BSA), and slides were incubated for 60 min at 37°C. The mAb25 mouse monoclonal antibody that recognizes the TbSAXO1 protein found along the entire length of the axoneme (Pradel et al, 2006) was used as a flagellar marker. The anti‐IFT22/RABL5 is a polyclonal mouse antiserum recognizing IFT22 (Adhiambo et al, 2009), and the monoclonal anti‐IFT172 antibody is a classic marker for IFT‐B proteins (Absalon et al, 2008). Slides were washed three times in PBS before incubation with specific secondary antibodies, diluted in PBS + 0.1% BSA, for 60 min at 37°C. Sub‐class specific secondary antibodies were used for double labeling and detection. Secondary antibodies were coupled to either Cy3 or Cy5 (Jackson ImmunoResearch Laboratories, West Grove, PA) or Alexa 488 (Invitrogen). Slides were washed again and stained with DAPI (2 μg/μl; stains nucleus and kinetoplast) and mounted using ProLong (Invitrogen). Experiments were performed at least twice to confirm the results.

A DMI4000 Leica microscope equipped with a 100× 1.4 lens (Leica, Germany) was used for observing slides, and images were captured using an ORCA‐03G camera (Hamamatsu). Images were analyzed using ImageJ v1.49 (National Institutes of Health, USA). Flagellum length was measured using the mAb25 signal and the measuring tool of ImageJ. A total of 50 (TbIFT22_D175A; Appendix Fig S6F) or 100 (TbIFT22_S19N; Fig 5D) flagella were measured per experiment. Populations were compared using the ANOVA test with the appropriate tool in Kaleidagraph 4.5.2 (Synergy Software).

Live cell imaging

Cultured cells were spread onto a slide, covered with a coverslip, and observed using the DMI4000 Leica Microscope. Videos were acquired using an Evolve 512 EMCCD Camera (Photometrics, AZ) driven by the Micro‐Manager Acquisition software (Molecular Probes, CA) to record videos at 100‐ms exposure. Analysis of acquired videos was performed using ImageJ v1.49. Kymograph extraction was performed using the KymographTracker plugin in Icy 1.9.5.1 (BioImage Analysis Unit, Institut Pasteur, France). Kymographs give a 2D graphical representation of the spatial position of IFT trains over time. The x‐axis corresponds to the length of the region of interest (ROI), while the y‐axis corresponds to the elapsed time. The ROI was traced semi‐automatically as a path in a maximum intensity enhanced projection of a time‐lapse image sequence (200 frames at 10 fps) by clicking control points in the intensity projection such that the curve followed a high pixel‐value trail.

Transmission electron microscopy

For TEM, cells were treated essentially as in Fort et al (2016). Cells were directly fixed in suspension with 2.5% (final) glutaraldehyde and washed in PBS, and the pellet was post‐fixed for one hour with 2.5% EM grade glutaraldehyde and 1% paraformaldehyde containing 0.1% tannic acid. The samples were washed three times in phosphate buffer for 5 min, and the pellet was resuspended in 1% osmium tetroxide in phosphate buffer. After 1 h, the pellet was rinsed five times with water and incubated overnight at 4°C in 2% uranyl acetate. It was rinsed five times in water before graded dehydration series in acetone (10–30–50–70–90–100% for 15 min each). Graded replacement with the Agar‐100 resin was then carried out (25–50–75–100% for 15 min each) followed by three successive incubation in 100% Agar‐100. The resin was polymerized for 2 h at 100°C. Ultrathin sections (50–70 nm thick) were collected on formvar/carbon‐coated nickel grids using a Leica EM UC6 ultra‐microtome and stained with uranyl acetate (2%, w/v) (uranyl acetate dihydrate, Electron Microscopy Sciences) and lead citrate (80 mM, buffer made in‐house). Observations were made on a Tecnai BioTWIN 120 cryo electron microscope (FEI), and images were captured with a MegaView II camera (Arecont Vision, France) and processed with AnalySIS and Adobe Photoshop CS4 (San Jose, CA).

Western blot

Cells were washed once in PBS. Laemmli loading buffer was added to the cells, and samples were boiled for 5 min. 20 μg of protein was loaded onto each lane of a Criterion™ XT Bis‐Tris Precast Gel 4–12% (Bio‐Rad, UK) for SDS–PAGE separation. XT‐Mops (1×) diluted in deionized water was used as a running buffer. Proteins were either transferred onto nitrocellulose membranes at 100 V over 1 h or by using the Bio‐Rad Trans‐Blot Turbo™ blotting system (25 V over 7 min). The membrane was blocked with 5% skimmed milk for one hour and then incubated with the anti‐RABL5/IFT22 primary antibody diluted in 0.05% PBS‐Tween (PBST). The anti‐RABL5 polyclonal antibody was diluted 1/500. As a loading control, the anti‐BiP (marker for an endoplasmic reticulum protein; Bangs et al, 1993) diluted 1/1,000 and anti‐PFR (L13D6; Kohl et al, 1999) diluted 1/50 were used. Both primary antibodies were diluted in 0.05% PBST containing 1% milk. After primary antibody incubation, three washes of 5 min each were performed in 0.05% PBST followed by secondary antibody incubation. Anti‐mouse secondary antibody coupled to horseradish peroxidase, diluted 1/20,000 in 0.05% PBST containing 1% milk, was used, and the membrane was incubated for 1 h. The Amersham ECL Western Blotting Detection Reagent Kit (GE Healthcare Life Sciences, UK) was used for final detection of proteins on the membrane.

Accession numbers

The coordinates and structure factors have been deposited in the Protein Data Bank under the accession codes 6IA7, 6IAE, and 6IAN.

Author contributions

SW carried out the purification of all proteins, performed biochemical assays, and designed mutants. SW determined the crystal structures with the help of JB and EL. JJ, SS, and CF carried out the in vivo analysis in trypanosomes under supervision of PB. EL supervised the biochemical and structural experiments. SW, EL, and PB wrote the paper.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Movie EV1

Movie EV2

Movie EV3

Movie EV4

Movie EV5

Movie EV6

Movie EV7

Movie EV8

Source Data for Appendix

Review Process File

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 5

Source Data for Figure 6

The EMBO Journal (2019) 38: e101251