Loss of Flower does not affect transferrin receptor 1 endocytosis but reduces the recycling of LAMP1 at the immune synapse

We assessed Fwe function in clathrin-mediated endocytosis by comparing the uptake of TfR1 in CTLs treated with pitstop2, an inhibitor of clathrin-mediated endocytosis [16], to that in Fwe KO CTL and wild type (WT) CTL. TfR1 uptake was assessed using FACS analysis with an FITC-labeled TfR1-antibody (Fig. 1A). In the pitstop2 treated CTLs, an accumulation of FITC signal at the cell surface even after 30 min incubation indicates a reduction in endocytosis. In WT and Fwe KO cells, the FITC fluorescence was reduced over time, indicating ongoing endocytosis. TfR1-FITC internalization was similar in control and Fwe KO CTL (near 70% at 15 and 30 min) while it was significantly lower in pitstop2 treated WT cells (Fig. 1B). We also investigated the role of Fwe in TfR1 endocytosis with SIM imaging in anti-CD3ε stimulated and unstimulated cells. Under both conditions, we observed abundant TfR1 uptake in WT and Fwe KO cells extending the result of the FACS analysis of stimulated cells (Fig. S1A, B). These data show that Fwe is not involved in TfR1 endocytosis response in classical clathrin-mediated endocytosis.

Since it was shown that Fwe promotes Syb2 endocytosis, we investigated whether it has a general role in CG membrane protein uptake by studying LAMP1 endocytosis [2]. LAMP1 is a lysosomal membrane protein regularly used as a CG marker [17]. In resting CTL, surface expression of LAMP1 is similar in WT and Fwe KO CTL [2]. However, LAMP1 is not only associated with CG but also to large extend with other lysosomes [18]. The CG fraction of LAMP1 becomes exposed to the extracellular surface of CTL following CG fusion only after T-cell receptor engagement (Fig. S1C) [9]. To get exclusively access to this exocytosed LAMP1, we applied fluorescent anti-LAMP1-Alexa647 antibody to CTLs in contact with P815 target cells. Upon binding with LAMP1, this antibody then appears in the cytosol of the CTL, allowing imaging of LAMP1 endocytosis. Cells were fixed after 30 min of co-culture and the fluorescence was visualized. The accumulated LAMP1 fluorescent signal at the IS in KO CTLs demonstrates an endocytic deficiency in KO CTLs (48.4% ± 1.83) compared to WT cells (38.7% ± 1.64, p = 0.0002) (Figs. 1C, D and S1D). Taken together, these data indicate that Fwe, though not required for general clathrin-mediated endocytosis, promotes endocytosis of CG components.

Mouse Flower 2 is localized on vesicular structures and is delivered to the plasma membrane upon T cell target cell stimulation

Due to the discrepancy between the impact of Fwe on TfR1 and LAMP1 recycling in CTLs with and without target contact, we investigated whether Fwe is differentially localized at the PM in a stimulation-dependent manner. Fwe has been reported to be present on synaptic vesicles in Drosophila and to traffic to the plasma membrane [1]. In mouse CTL, Fwe also appears on small vesicular structures. To visualize these Fwe-positive vesicles near the synapse, Fwe KO CTLs were transfected with full-length Fwe constructs (Fwe2-mTFP and mScarletI-Fwe2-mTFP) and then applied to anti-CD3ε pre-coated coverslips to facilitate synapse formation. Fwe positive vesicles were then observed with real-time TIRFM imaging. In average, we observed the fusion of about 2 to 3 Fwe-positive vesicles per secreting cells. The granule shown in Fig. 2A reached the plasma membrane at 0.1 s and subsequently fused with a gradual dispersion of the membrane-bound Fwe-mTFP over the following two seconds (Fig. 2A, right panel). Figure 2B shows a time sequence of the mean fluorescence measured from a region of interest (ROI) placed at the position of the vesicle as it reached the plasma membrane, and of the mean fluorescence of the entire frame. Though the fluorescence at the point of fusion rapidly decayed, the total brightness changed little in the two seconds following fusion (*), indicating that the Fwe-mTFP fluorescence remained at the plasma membrane while diffusing away from the point where fusion occurred.

Fig. 2

Mouse Flower 2 is localized on small vesicles that fuse with the plasma membrane upon T cell stimulation. A Live cell TIRFM imaging of mFwe2-mTFP expressing Flower KO CTL. Snapshot images of an exemplary cell and a Flower fusion event. Overall, 20 fusion events were measured in 11 cells expressing either mFwe2-mTFP or mScarletI-mFwe2-mTFP. The images were recorded at 10 Hz. Scale bar 2 µm and 1 µm, respectively. B Fluorescence intensity time course of the mFwe2-mTFP-positive vesicle shown in A. The green line shows the fluorescence intensity at the vesicle (insert with green square), whereas the gray line corresponds to the intensity of a larger membrane area (insert with stippled gray rectangle). Fusion occurs when the green line shows maximum fluorescence intensity (*). Its decay indicates dispersion of the mFwe2-mTFP in the plasma membrane following fusion of the vesicle. Accordingly, a slight increase in fluorescence is observed in the surrounding plasma membrane. The gray bar corresponds to the time of the snapshots shown in A. C A representative electron micrograph showing part of the immunological synapse of a CTL expressing mFwe2-mTFP. Post-embedding immunogold electron microscopy was done on activated mCTLs settled on anti-CD3ε coated sapphire discs. The protein was labeled with the primary antibody, anti-tRFP against mTFP, and with a secondary gold-conjugated antibody (10 nm). The microtubule organizing center (MTOC) is present, as are Fwe-positive vesicles (yellow). Flower was identified on the plasma membrane (red arrows) and on a coated pit (black arrow). Scale bar, 0.2 µm. D Quantitative analysis of the diameter of immunogold-labeled Flower positive vesicles as shown in C. Data given as mean ± SEM (N = 4, ncells = 14, nvesicles = 99). E Quantitative analysis of endogenously expressed Flower protein in WT and KO mCTLs localized on the plasma membrane (shown in C, red arrows). Post-embedding, immune electron microscopy was done with primary anti-Flower antibody and 10 nm secondary goat anti-rabbit gold antibody on ultrathin sections cut parallel to the sapphire disc. For background control (bg) sections were stained without primary antibody. Gold particles were counted per µm. Data given as mean ± SEM. Kruskal–Wallis One-way Analysis of Variance on Ranks followed by multiple comparison (Dunn’s) was done. (N = 3; WT n = 19, KO n = 18, bg n = 21; ***p < 0.001, not significant (ns)). F Quantitative analysis of overexpressed mFwe2-mTFP protein localized on the plasma membrane. Post-embedding immunogold electron microscopy was done as described in C with primary anti-Flower antibody and 10 nm secondary goat anti-rabbit gold antibody on ultrathin sections cut vertical to the sapphire disc. Inset shows a representative image of an activated CTL, settled on anti-CD3ε coated sapphire disc with marked immunological synapse (IS, green) and non IS regions (blue). Gold particles localized on the plasma membrane, as shown in C, were counted per µm (mean ± SEM (N = 3; non IS n = 18)). Data significance was analyzed by Mann–Whitney Rank Sum Test (*p < 0.05)

We further investigated the localization of the Fwe protein in CTLs via immunogold labeling with an anti-Fwe antibody by electron microscopy. The gold particles labeled vesicular compartments (highlighted in yellow) and the plasma membrane in Fwe KO CTL expressing mFwe2-mTFP (Fig. 2C) and the endogenous localization of the Fwe protein (Fig. S2). The average diameter of the immunogold-marked Fwe-positive vesicles was 122.92 ± 5.95 nm (Fig. 2D). We compared the amount of endogenous Fwe by counting the bound gold particles at the plasma membrane of Fwe KO, WT CTL and a background control (without primary antibody). WT CTL displayed a significant higher abundance of 10 nm gold particles at the PM than KO CTL (Fig. 2E). In mFwe2-mTFP overexpressing cells, we compared the density of gold particles at the IS to that of the rest of the cell. CTLs stimulated with anti-CD3ε-coated coverslips showed significant enrichment of Fwe protein at the plasma membrane of the synapse (Fig. 2F). These results show that Fwe reaches the plasma membrane upon stimulation via fusion of Fwe-positive vesicles and that this occurs predominantly at the IS.

Four Flower transcripts are expressed in mouse cytotoxic T lymphocytes at different days of activation

The mouse Fwe gene encodes five different splice variants, which result in the expression of four protein isoforms (Fig. 3D) (www.ensemble.org, [19, 20]). Mouse Fwe 2 (mFwe2, the full-length form), is prominently expressed and present in naive CTL and at higher levels in cells activated with anti-CD3/anti-CD28 activator beads for three, four and five days (western blot analysis, Fig. 3A, B). We then determined which Fwe isoforms are expressed in mouse CTL at different activation states.

Fig. 3

Flower expression in mouse CTL at different days of activation. A Representative western blot of Flower protein expression levels in homogenates of naive and stimulated mouse WT CTL of day 3, day 4 and day 5 (20 µg per lane). The loading control was GAPDH. B Quantitative analysis of the western blots shown in A. Flower expression was normalized to the expression level of GAPDH. The data represent means ± SEM from four mice. C RT-PCR analysis of mFwe1, mFwe2.1 and 2.2, mFwe3, mFwe4 and mFwe5 transcripts (Transcript: ENSMUST00000114004.8, ENSMUST00000114007.8, ENSMUST00000114006.8, ENSMUST00000114005.9, ENSMUST00000114003.2, ENSMUST00000133807.2, respectively) in WT mouse CTL at different activation states. Total RNA from adult mouse brains was used as the positive control, and total RNA isolated from mFwe knockout spleens was used as the negative control. White stars show the positions of the expected bands in mFwe2.2 and mFwe3. SDHA was used as the loading control. D Scheme of primer localization for RT-PCR analyses on the six postulated mouse Flower transcripts mFwe1 (ENSMUST00000114004.8), mFwe2.1 (ENSMUST00000114007.8), mFwe2.2 (ENSMUST00000114006.8), mFwe3 (ENSMUST00000114005.9), mFwe4 (ENSMUST00000114003.2) and mFwe5 (ENSMUST00000133807.2). Exon coding regions are shown in black with numbers, and putative exons are shown in white with numbers. Untranslated regions are shown in gray. The red bar shows the targeted deletion of exon 2 and exon 3 in the mFwe KO mouse. E Expression of mFwe1, mFwe2, mFwe3 and mFwe4 in naive, day 3 and day 5 CTL normalized to the expression of the SDHA as housekeeping gene. The data represent means ± SD of samples from three mice per condition

To analyze the transcription of the murine Fwe splice variants in mouse CTL, RT-PCR was performed on total RNA isolated from naive and activated day three and day five CTL (Fig. 3C). The primers for the RT-PCR were designed specifically at characteristic exon–intron borders to identify and differentiate the different splice variants. Figure 3D illustrates the structure of the mouse Fwe gene, as well as the five isoforms that can be expressed in mouse CTLs. In addition, the positions of the primers used for PCR are shown. Whole-brain RNA from adult mice was used as a positive control, and total RNA from day five Fwe KO CTL was used as the negative control. The amplified DNA bands had the appropriate size and indicated that mFwe1, mFwe2, mFwe3 and mFwe4 were expressed. mFwe2.2 is barely detectable and generates the same protein as Fwe2.1. The relative mRNA expression was normalized to SDHA expression (Fig. 3C, E; N = 3). Activated mFwe KO CTLs were transfected with murine Fwe isoforms Fwe1, Fwe2, Fwe3 and Fwe4 at day five for further experiments.

Flower KO CTL show similar proliferation and activation state compared to WT cells

Since CTL proliferation is an important indicator of cell integrity [21], we compared the division state and subset transition during activation in Fwe KO and WT CTLs. CSFE staining of WT and Fwe KO CTLs was performed to monitor distinct generations of proliferating cells by dye dilution. Flow cytometry analysis of unstained and CSFE-stained naive, undivided CTLs and their proliferation states at days three, four and five, of WT and Fwe KO mice were analyzed. The result showed no difference between the proliferation states in WT and Fwe KO CTLs (Fig. S3A).

Furthermore, the degree of differentiation of WT and Fwe KO CTLs was determined by FACS (Fig. S3B, C). CD44 and CD62L surface markers were used to define naive (CD44-/CD62L-), central memory (CD44 + /CD62L +), and effector memory (CD44 + /CD62L-) CTL cells [22]. CD25 was used as activation marker. Since all subsequent experiments were performed on day five, CTL subset staining was performed on day five. Activated WT and Fwe KO CTLs presented similar profiles, which consisted mainly of the effector memory population and some remaining central memory cells (Fig. S3B, C; N = 3).

Mouse Flower isoforms differ in their ability to restore Synaptobrevin2 endocytosis.

To ascertain the domain of Fwe required to promote endocytosis of CG membrane proteins, we examined the ability of naturally occurring Fwe isoforms to rescue endocytosis in Fwe KO cells. We started by evaluating the mouse isoforms Fwe1-4, whose sequence alignments are shown in Fig. 4A. Each of these proteins was expressed as a pMAX-Fwe-mTFP fusion protein in murine Fwe KO CTLs (Fig. 4B). We used Syb2, the v-SNARE mediating CG fusion [11], as a highly specific membrane protein marker for CG endocytosis [13]. The experiment was performed as follows: Fwe KO CTLs were transfected with Syb2-mRFP together with the different Fwe constructs and placed in contact with P815 target cells to induce CG exocytosis. The mRFP tag is located in the granule lumen and is exposed to the extracellular space after CG fusion (Fig. 4C). The cells were then treated extracellularly with an Alexa647-labeled anti-RFP antibody, which readily binds specifically to the mRFP exposed at the cell surface. Following endocytosis, the fluorescently tagged anti-RFP-Alexa647 was internalized and distributed within the cytoplasm (Fig. 4C). These endocytosed Syb2 vesicles are visualized as double labeled spots. Cross reactivity of the anti-RFP antibody with mTFP was avoided due to the localization of the fluorescent protein. While the mRFP is exposed to the extracellular space upon CG exocytosis, the mTFP attached to Fwe remains intracellular (Fig. S4A). We followed this process by live-cell confocal imaging over a period of 15 min.

Fig. 4

Mouse Flower isoforms differ in Synaptobrevin2 endocytosis efficiency. A Sequence alignment of mouse Flower (mFwe) isoforms 1 to 4 using Clustal Omega. The exon/intron borders are shown in red. (*) Indicates conserved amino acids in exon 3. B Murine Flower isoform constructs. The exon structures of the four mouse isoforms are shown, as are the number of amino acids and the linker region at the C-terminus with the coupled fluorophore mTFP. C Schematic drawing of the experiment shown in D. D Series of images of Fwe KO CTLs expressing Syb2-mRFP (yellow) and each isoform of mFwe-mTFP that were acquired over 15 min. CTLs were conjugated with P815 in the presence of anti-RFP-Alexa647 antibody (magenta) in the medium. Images were acquired at 0, 2.5, 5, 7.5, 10 and 15 min. White arrows mark endocytosed organelles, and open white arrows indicate organelles at the IS. Scale bar: 5 μm. E Quantitative analysis of endocytosed Syb2-mRFP signals (based on anti-RFP-Alexa647 antibody fluorescence) at the immunological synapse (IS) for mFwe1-4, as shown in D, in comparison to WT and Flower KO mouse CTL. Time zero was defined as the appearance of the first anti-RFP-Alexa647 signal at the IS. Data given as mean ± SEM; Kruskal–Wallis One-way Analysis of Variance on Ranks followed by multiple comparison (Dunn’s) was done versus Fwe KO as control (*p < 0.05, **p < 0.01, *** p < 0.001, KO n = 19 (N = 4); WT n = 16, mFwe1 n = 21, mFwe2 n = 16, mFwe3 n = 18, mFwe4 n = 19 (N = 3)). F SIM images of a Fwe KO CTL co-expressing Syb2-mRFP and mFwe2-mTFP in contact with a P815 target cell. Anti-RFP-Alexa647 antibody was applied in the medium to label endocytic Syb2. Cells were fixed after 40 min to allow synapse formation and endocytosis to occur. MIP images and single stack images of the same cell are shown to demonstrate the precise co-localization of mFwe2 and endocytic Syb2 (white arrows)

Time-lapse snapshots of Fwe KO CTLs expressing the four murine isoforms tagged with mTFP in contact with a P815 target cell (Fig. 4D) show the accumulation of anti-RFP-Alexa647 at the contact zone, the IS. The first occurrence of endocytic Syb2 is indicated as time 0′00’’. Over time, endocytosis of Syb2 was observed as fluorescent puncta moving from the IS into the cell (Fig. 4D). We quantified the fluorescence intensity of anti-RFP-Alexa647 at the IS (30% of the cell volume) versus the entire T cell. We compared cells overexpressing either one of the Fwe isoforms with Fwe KO and WT CTLs (Fig. 4E). Starting from 100% at the time of contact, Fwe KO CTLs exhibited little loss of fluorescence from the IS area, whereas WT CTLs exhibited a steady decrease in the fraction of RFP fluorescence at the IS over the 15-min recording period. Fwe KO CTLs expressing mFwe2 presented a loss of fluorescence at the IS, similar to WT cells. KO CTLs expressing either mFwe1 or mFwe3 presented a lower reduction in anti-RFP fluorescence at the IS than that observed in either the WT control or the mFwe2 construct, but the reduction was significantly greater (p < 0.001 for mFwe1 and p = 0.047 for mFwe3) than that observed in the Fwe KO CTLs. Loss of anti-RFP fluorescence in Fwe KO CTLs expressing the mFwe4 construct did not differ from that in the Fwe KO CTL. This result indicates that deletion of the N-terminus results in loss of function, while the C-terminal deletion also reduces function but to a lesser degree.

The confocal images suggest that endocytosed Syb2 is trafficked together with Fwe. To confirm this observation, we performed high resolution SIM imaging in Fwe KO CTLs overexpressing mFwe2-mTFP and Syb2-mRFP (Fig. 4F). The cells were fixed immediately after 30 min contact with P815 target cells in presence of the anti-RFP-Alexa647 antibody. We found that the mFwe-mTFP signal partially colocalized with the endocytosed anti-RFP-positive compartments and vice versa. This result suggests a transient localization of mFwe2 on endocytosed Syb2 vesicles.

The integrity of the mouse Fwe constructs was demonstrated by western blot as all constructs showed the expected molecular mass (Fig. S4B). Furthermore, no significant differences in the expression levels of the constructs were detected, indicating that the endocytosis phenotype is not due to variable protein expression levels. The analysis was conducted by analyzing the mTFP fluorescence intensities of the expressed mouse Fwe constructs in all cells used for the endocytosis assay (Fig. S4C).

The position of the mTFP tag could affect the efficacy of the mFwe2 construct function. Therefore, we expressed mFwe2 tagged with mTFP at its N- or C-terminus. Figure S4D shows exemplary images from time series of the endocytosis assay taken with Fwe KO CTLs expressing either construct. The anti-RFP fluorescence at the IS over time was similar for both constructs (Fig. S4D, E). Thus, the position of the TFP tag did not affect the efficacy of rescue.

Mouse Flower isoforms polarize to the plasma membrane with different efficacies upon stimulation

We examined the cellular distribution of the four mouse Fwe-mTFP constructs expressed in Fwe KO CTLs. The PM was stained with an anti-CD8-Alexa647 antibody, and the ER was stained with an overexpressed ER-mScarletl construct (Fig. 5A). 12 to 14 h after transfection, the cells were stimulated with anti-CD3/anti-CD28 beads for one hour to mimic IS formation and to drive the Fwe protein into the membrane. Costaining of Fwe-mTFP with the plasma membrane (CD8) is visible for all mFwe isoforms. Scatter dot plots of Manders’ coefficients show that mFwe2 reached the plasma membrane to a higher degree (0.38 ± 0.02) than did the mFwe1, 3 or 4 constructs (Figs. 5B–D and S5A for better visualization). Mouse Fwe1 and 4 are predominantly found in the ER. Although mFwe1 and mFwe3 weakly colocalized with the plasma membrane, a partial rescue of Syb2 endocytosis was detected (Fig. 4E). Additionally, we could show that all functional mFwe isoforms (1 to 3) partially trafficked to the IS on common vesicular structures (Fig. S5B, C).

Fig. 5

Mouse Flower 2 is localized on vesicular structures and the plasma membrane in bead stimulated CTLs. A Single plane SIM images of activated Fwe KO CTLs transfected with one of the four mouse Flower-mTFP isoforms and ER-mScarletI. After 14 h, cells were bead simulated for 1 h and stained with CD8-Alexa647 as a surface marker. The cells were plated on poly-L-ornithine coated coverslips, fixed with 4% PFA and imaged. Scale bar: 5 µm. B, C The Manders’ coefficient of colocalization between Flower isoforms and CD8 (B) and the ER marker (C). Means ± SD; n = 32, 29, 30, 25, respectively, N = 2, Kruskal–Wallis test followed by Dunn’s Multiple Comparison (*** p < 0.001, ns > 0.05). D Manders’ coefficient of colocalization of CD8 and ER markers (means ± SD; n = 32, 29, 30, 25, respectively, N = 2, Kruskal–Wallis test followed by Dunn’s Multiple Comparison, * p < 0.05, ** p < 0.01, *** p < 0.001, ns > 0.05)

Since expression levels do not correlate with endocytosis rescue and there were significant differences in the cellular distributions of the Fwe constructs, we examined the oligomeric state of several selected constructs after expression in CTL. Figure S6 shows western blots of lysates (10 µg) of Fwe KO CTLs expressing mFwe1–4 prepared under non-denaturing conditions and probed with anti-DsRed antibody recognizing mTFP (native PAGE and western blotting; see Materials and Methods). The mFwe2 protein was present predominantly as a tetramer but was also clearly present as a dimer, with slightly higher-order oligomerization. Although mFwe1 and mFwe3 have different transfection efficiencies, as indicated by the weaker signal in the lane of the western blot (Fig. S6), these isoforms essentially appear as tetramers and hexamers. Mouse Fwe4 is present as tetramers, hexamers and higher-order oligomers. Together with our endocytosis rescue experiments, these results indicate that functional Fwe isoforms in mouse CTLs appear as dimers or tetramers.

Mutation of the putative AP2 binding site YRWL at the N-terminus of mFwe2 and mFwe1 prevents rescue of Synaptobrevin2 endocytosis

We found that the deletion of the C-terminal end of mFwe (mFwe1 and 3) had a weaker effect on the ability to rescue Syb2 endocytosis than the N-terminus (mFwe4). A feature of both the N- and C-terminal exons is the putative AP2 binding motif YXXΦ [23], which could support an interaction with clathrin in endocytosis. Mutation of both AP2 binding motifs prevents Syb2-mRFP endocytosis [2, 15]. Our objective was to determine the individual contribution of each AP2 site to the protein's function and compare the mutation effects with those of the N- and C-terminal deletions.

We generated constructs in which an N-terminal YRWL26−29 to AAAA, a C-terminal YARI148−151 to AAAA mutation, as well as a construct in which both sequences were mutated to AAAA in mFwe2. Western blots prepared from a lysate of Fwe KO CTL expressing these constructs ran at the expected molecular mass (Fig. S7A, left). The targeted mutation of YRWL to AAAA did not alter the oligomerization state of mFwe2, as shown by the native western blot in Fig. S7C. We overexpressed them in Fwe KO CTLs and conducted our Syb2 endocytosis assay. The expression levels of these constructs measured via the mTFP fluorescence intensities of the cells were similar, with the exception of the mFwe2 (YARI/AAAA) mutant construct, which was expressed at a greater level (Fig. S7A, right). The loss of anti-RFP-Alexa647 fluorescence at the IS is shown as time-lapse live snapshots over 15 min in Fig. 6A and is quantified in Fig. 6B. For comparison, mouse Fwe KO cells and Fwe KO CTLs expressing the mFwe1- and mFwe2-mTFP constructs, are shown as well. Mutation of the N-terminal AP2 site, or both AP2 sites, blocked the ability of mFwe2 to rescue endocytosis of Syb2-mRFP. Mutation of the C-terminal site reduced rescue, but this mutant performed significantly better than the N-terminal mutant compared with the KO control. The results for both single mutants were verified by SIM (Fig. S8 A, B). Additionally, this high-resolution imaging revealed that none of the AP2 mutant constructs accumulated at the IS, suggesting that they are not retained within the plasma membrane. This finding indicates that while the C-terminal AP2 site facilitates endocytosis, the N-terminal site is essential.

Fig. 6

Mutation of the putative AP2 binding site YRWL at the N-terminus in mFwe2 and mFwe1 inhibits Synaptobrevin2 endocytosis. A Sequential images of Fwe KO CTL expressing Syb2-mRFP (yellow) and either mFwe2-mTFP mutants (N-terminus YRWL/AAAA, C-terminus YARI/AAAA and N- and C-terminus YRWL/AAAA and YARI/AAAA) or mFwe1(YRWL/AAAA)-mTFP. CTLs are conjugated with P815 target cells in the presence of anti-RFP-Alexa647 antibody. Scale bar: 5 μm. B Quantitative analysis of the distribution of anti-RFP-Alexa647 at the immunological synapse (IS) of Fwe KO CTLs expressing mFwe1 and mFwe2 mutants as described in A compared with that of Fwe KO CTL and Fwe KO CTL expressing mFwe1 and mFwe2 isoforms. Time zero was defined as the appearance of the first anti-RFP fluorescence at the IS. (Data given as means ± SEM; KO n = 19 (identical to Fig. 4E), N = 4; mFwe1 n = 21, mFwe2 n = 16, mFwe1(YRWL/AAAA)/N-terminus n = 22, mFwe2(YRWL/AAAA)/N-terminus n = 20, mFwe2(YARI/AAAA)/C-terminus n = 22, mFwe2(YRWL/YARI)/N- and C-termini n = 21, N = 3; Kruskal–Wallis test followed by Dunn’s Multiple Comparison versus KO control (* p < 0.05, ** p < 0.01, *** p < 0.001, ns > 0.05)

To prove that it is in fact the N-terminal AP2 site that confers its function to the mFwe isoforms lacking the C-terminus, we mutated YRWL to AAAA at the N-terminal AP2 site in mFwe1 (Fig. 6A). After expression in Fwe KO CTLs, this mutation also resulted in a strong reduction of Syb2 endocytosis rescue (Fig. 6B). The mFwe1 constructs ran at the expected molecular mass in western blots and the fluorescence levels in CTL were similar, ruling out differences in protein expression (Fig. S7B). These results show that the mutation of the N-terminal AP2 site of Fwe is sufficient to prevent rapid recycling of Syb2 mRFP in mouse CTLs.

The human isoforms Fwe2 and Fwe4 and the Drosophila isoform dUbi rescue Syb2 endocytosis in Flower KO murine CTLs

Human Fwe gene generates four splice variants, hFwe1-4 (Fig. 7A). Interestingly, two of these isoforms lack a central exon and show variability in their C-termini, allowing a more in-depth analysis of Fwe domains. In detail, hFwe1 and 2 differ from hFwe3 and 4 by their exons 5 and 6 at the C-terminus. hFwe1 is identical to the full-length hFwe2 with 233 amino acids (aa), but with a deletion of 42 aa beginning at aa 67 by exon skipping and additionally with a methionine at position 65. hFwe3 and hFwe4 have identical aa sequences, with the exception of the deletion mentioned above and the M65I aa exchange in hFwe4. All four human constructs contain the N-terminal YWRL site, and hFwe3 and 4 have the YARI sequence at the C-terminal site as well. In Drosophila, there are three splice variants (Fig. 7B), a ubiquitous form, dUbi, and two additional variants, dLoseA and dLoseB, which have been reported to generate “loser” phenotypes that tend to undergo apoptosis when competing with other cells [6]. The Drosophila isoforms differ among each other only in their C-termini starting from aa 147.

Fig. 7

The human isoforms Fwe2 and Fwe4 and the Drosophila isoform dUbi rescue endocytosis in mouse Fwe KO CTLs. A Sequence alignment of human Flower isoforms hFwe1, hFwe2, hFwe3 and hFwe4 using Clustal Omega. * Indicates conserved amino acids. The exon/intron borders are shown in red. B Sequence alignment of Drosophila Flower isoforms dUbi, dLoseA and dLoseB using Clustal Omega. * Indicates conserved amino acids. The exon/intron borders are shown in red. C Time-lapse images of Syb2-mRFP fluorescence (yellow) of Fwe KO CTLs expressing each of the four hFwe-mTFP constructs (green) described in A conjugated to P815 target cells in the presence of anti-RFP-Alexa647 antibody (magenta). Scale bar: 5 μm. D Quantitative analysis of endocytosed Syb2-mRFP by anti-RFP-Alexa647 fluorescence at the immunological synapse (IS) in Fwe KO mCTLs expressing hFwe1-mTFP, hFwe2-mTFP, hFwe3-mTFP and hFwe4-mTFP constructs shown in C in comparison to those in WT and Flower KO mouse CTL. Time zero is defined as the appearance of the first endocytic signal at the IS. Data given as mean ± SEM; Kruskal–Wallis One-way Analysis of Variance on Ranks followed by multiple comparison (Dunn’s) was done against Fwe KO as control (KO n = 19, N = 4; WT n = 16 (identical to Fig. 4E), hFwe1 n = 11, hFwe2 n = 11, hFwe3 n = 11, hFwe4 n = 11, N = 3; *p < 0.05, **p < 0.01, *** p < 0.001). E Time-lapse images of Fwe KO mCTLs transfected with Syb2-mRFP (yellow) and each of the three Drosophila Fwe-mTFP isoform constructs (green) described in B. CTLs are conjugated to P815 target cells with anti-RFP-Alexa647 antibody (magenta) in the medium. Scale bar: 5 μm. F Quantitative analysis of endocytosed Syb2-mRFP fluorescence by anti-RFP-Alexa647 fluorescence at the immunological synapse (IS) in Fwe KO mCTL expressing dUbi-mTFP, dLoseA-mTFP or dLoseB-mTFP constructs, as shown in E, in comparison to WT and Fwe KO mCTL. Time zero was defined as the appearance of the first anti-RFP-Alexa647 signal at the IS. Data given as mean ± SEM; Kruskal–Wallis One-way Analysis of Variance on Ranks followed by multiple comparison (Dunn’s) was done against Fwe KO as control (KO n = 19, N = 4; WT n = 16 (identical to Fig. 4E), dUbi n = 17, dLoseA n = 16, dLoseB n = 14, N = 3; *p < 0.05, **p < 0.01, *** p < 0.001)

The human and Drosophila splice variants were expressed in mouse Fwe KO CTLs as described previously to measure their ability to rescue Syb2 endocytosis (Fig. 7C–F). A rescue, equivalent to that of the WT constructs, was achieved with both hFwe2 and hFwe4 constructs while the hFwe1 and hFwe3 constructs did not rescue Syb2-mRFP endocytosis (Fig. 7C, D). Thus, deletions of exon 3 in human isoforms prevent rescue while C-terminal deletions have little effect. Rescue with the dUbi construct exhibited Syb2-mRFP endocytosis like that of WT CTLs, whereas both the dLoseA and dLoseB constructs failed to promote endocytosis of Syb2-mRFP (Fig. 7E, F). The above results indicate that the isoforms hFwe2 and 4 as well as dUbi are capable of supporting Syb2 endocytosis in mouse Fwe KO CTLs.

We confirmed the molecular sizes of all the constructs and the similarity of fluorescence intensities in mouse CTLs after transfection (Fig. S9A, B). Western blot analysis from native PAGE hFwe4 exhibits dimerization and tetramerization while hFwe3 runs at higher molecular mass (Fig. S9C). The dUbi construct appeared predominantly as a dimer with a weak band at the tetramer level (Fig. S9D).

Point mutations at the highly conserved tyrosine, Y104A in mFwe2 and Y109A in the Drosophila Ubi isoform, inhibit Synaptobrevin2 endocytosis

The results of the rescue experiments with the human isoforms indicate that the equivalent sequence found in mouse exon 3 is critical, as shown in hFwe1 and 3 (Fig. 7D). Deletion of exon 3 of mFwe1 (mFwe1(Δex3)) resulted in a complete loss of endocytic function (Fig. S10A–C). Fwe is included in a family of small proteins (less than 200 residues) that contain a Cg6151-P domain, which is conserved from fungi to humans. The alignment of various proteins from this Cg6151-P family in 16 different species, including mouse, human and Drosophila Fwe, is shown in Fig. 8A [24, 25]. Within the aligned sequences corresponding to the human exon 3 (highlighted in gray), only a glutamate (E74 in mouse), a proline (P76 in mouse) and a tyrosine (Y104 in mouse) are conserved (Fig. 8A, highlighted in yellow and marked *). Additionally, two (E, P) of these three amino acids are comprised in a sequence that has been identified as the selectivity filter in some TRP channels in Drosophila [1]. These authors exchanged the glutamate (E79) with a Q in Drosophila WT-Flower-PB and reported a loss of endocytosis in Drosophila salivary gland cells. We targeted mutations in the full-length mFwe2 at the E74 and Y104 residues and in the equivalent Y109 in dUbi and confirmed the integrity of all the constructs (Figs. S11A and S9E, respectively). Syb2 endocytosis analysis was done as described above in Fwe KO CTLs (Fig. 8B, C). The change in anti-RFP fluorescence at the IS in Fwe KO cells expressing the mFwe2 (E74A) mutation was significantly higher than that in the Fwe KO CTL (p < 0.001) but lower than that in the mFwe2 expressing cells (p = 0.712, Kruskal–Wallis test followed by Dunn’s Multiple Comparison versus control). Mutation of Y104A in mFwe2 resulted in a dramatic loss of Syb2 endocytosis, as did mutation of Y104 to F (91.88% ± 0.76 at 15 min). High-resolution imaging of this construct confirmed this observation and revealed the localization of the construct at the IS (Fig. S8C). The equivalent Y109A mutation in Drosophila Fwe isoform dUbi prevented rescue as well (Fig. 8B, C). These point mutations in mFwe2 and dUbi had no influence on their expression levels, as the cells expressing the constructs displayed comparable mTFP fluorescence intensities (Figs. S11B and S9E, respectively). Manders’ coefficient analysis showed that the mFwe constructs with point mutations (Y104A or Y104F) reach the plasma membrane to a lesser degree than WT mFwe2 but to a greater degree than mFwe1, a mouse variant that partially rescues Syb2 endocytosis (Fig. 8D). The localization of the mutants in the ER was comparable to that of mFwe2 but significantly lower than that of mFwe1 (Fig. 8E).

Fig. 8

Point mutations in the highly conserved tyrosine residue Y104A in mFwe2 and Y109A in the Drosophila Ubi isoform block rescue of Synaptobrevin2 endocytosis. A Alignment of selected Cg6151-P-containing proteins in Clustal Omega format. The proteins include the TVP18 Golgi membrane proteins and Flower proteins from various species. The highly conserved glutamate (E), proline (P) and tyrosine (Y) residues are highlighted in yellow and marked with an *. The mouse, human and Drosophila sequences are highlighted in blue. The human Fwe equiva