Development and validation of RAMP

For use in the RAMP system, we constructed plasmids encoding full-length forms or part of organellar proteins such as the lysosomal/late-endosomal lysosome-associated membrane protein 1 (LAMP1), mitochondrial translocase of the outer membrane 20 (TOM20), peroxisomal peroxin 3 (PEX3), or endoplasmic reticulum (ER) VAMP-associated protein (VAP)-A, appended at their cytosolically exposed N- or C-termini with the 38-amino–acid SBP and a fluorescent protein (green fluorescent protein [GFP] or mCherry) (Fig 1A). These plasmids were co-transfected with plasmids encoding the motor, neck, and part of the coiled-coil domains from the plus-end–directed kinesin-1 KIF5B (denoted KIF5B*) [20] or minus-end–directed kinesin-14 KIFC1 (denoted KIFC1*) [21], both fused to streptavidin (strep) and another fluorescent protein (mCherry or GFP) (Fig 1B). These kinesin constructs lacked their cargo-recognition tail domains, so they could only bind to organelles via the SBP–streptavidin interaction (Fig 1C). Different combinations of organellar and motor constructs were expressed by transient co-transfection into cultured cells, resulting in accumulation of organelles at the plus or minus ends of microtubules, depending on the motor construct used (Fig 1D). This accumulation could be reversed by the addition of biotin (Fig 1D).

Initial tests of the RAMP system were performed for lysosomes in HeLa cells (Fig 2 and S1–S3 Figs). In these cells, lysosomes are distributed throughout the cytoplasm, although with a higher concentration in the perinuclear area of the cell. This distribution is dynamic, as lysosomes can move bidirectionally along microtubules radiating from the microtubule-organizing center (MTOC) [22,23] (S1A Fig). Plus-end–directed (anterograde) and minus-end–directed (retrograde) movement of lysosomes is driven by kinesin and dynein–dynactin motors, respectively [2,24]. We observed that singly expressed LAMP1-SBP-GFP localized to lysosomes, as detected by costaining for the endogenous lysosomal marker late endosomal/lysosomal adaptor and MAPK and MTOR activator 4 (LAMTOR4) [25] (Fig 2A and 2F). Singly expressed mCh-KIF5B*-strep, on the other hand, exhibited a largely cytosolic localization, albeit with a slight accumulation at cell protrusions, where the plus ends of microtubule bundles are more abundant (Fig 2B, arrow). Expression of LAMP1-SBP-GFP or mCh-KIF5B*-strep alone did not affect the distribution of lysosomes (Fig 2A and 2B) or the structure of the microtubule network (S1B and S1C Fig). When LAMP1-SBP-GFP and mCh-KIF5B*-strep were expressed together, however, we observed a dramatic redistribution of both proteins, as well as lysosomes labeled for LAMTOR4, to the cell periphery, particularly at cell protrusions (Fig 2C and 2F). This perturbation did not affect the integrity of the microtubule network (S1D Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Redistribution of lysosomes by expression of RAMP constructs.

(A) Colocalization of LAMP1-SBP-GFP with endogenous LAMTOR4 in HeLa cells. Arrows indicate colocalization. (B) mCh-KIF5B*-strep is largely cytosolic and does not disrupt the distribution of lysosomes stained for endogenous LAMTOR4. Arrow shows localization of a small amount of mCh-KIF5B*-strep in a cell protrusion. (C) Coexpressed LAMP1-SBP-GFP and mCh-KIF5B*-strep accumulate on LAMTOR4-stained lysosomes at peripheral cell protrusions. (D) Strep-KIFC1*-mCh is cytosolic and does not disrupt the distribution of lysosomes stained for LAMTOR4. (E) Coexpressed LAMP1-SBP-GFP and strep-KIFC1*-mCh accumulate on LAMTOR4-stained lysosomes in the perinuclear area of the cell. Nuclei were stained with DAPI. Cell edges are outlined. Scale bar: 10 μm. Rightmost panels are 4.5× magnifications of the boxed areas in the merged images in each row. (F) Box-and-whisker plots represent the fractional distance distribution (f = 95%) of lysosomes in the conditions from panels A, C, and E (see S4 Fig and Methods section for details). Summary data available as Supporting Information (S1_Data.xlsx). GFP, green fluorescent protein; KIF, kinesin superfamily; LAMP, lysosome-associated membrane protein; LAMTOR4, late endosomal/lysosomal adaptor and MAPK and MTOR activator 4; mCh, mCherry; RAMP, reversible association with motor proteins; SBP, streptavidin-binding protein; strep, streptavidin.

https://doi.org/10.1371/journal.pbio.3000279.g002

Singly expressed strep-KIFC1*-mCh was also largely cytosolic and did not affect the distribution of lysosomes (Fig 2D) or microtubules (S1E Fig). However, upon coexpression with LAMP1-SBP-GFP, both constructs colocalized to lysosomes clustered in the pericentrosomal area of the cells (Fig 2E and 2F) without perturbing the microtubules (S1F Fig). Similar results were obtained upon coexpression of LAMP1-SBP-GFP with either mCh-KIF5B*-strep or strep-KIFC1*-mCh in other nonpolarized cell lines such as COS-7 and HT-1080 (S2A and S2B Fig).

The above results demonstrated that the coupling of lysosomes to KIF5B* or KIFC1* via the SBP–streptavidin interaction was sufficient to reconstitute active kinesin motors and thus redistribute lysosomes to the periphery or the center of the cell, respectively.

Control experiments showed that endosomes containing transferrin receptor (TfR) remained distinct and were not redistributed under these conditions (S2C and S2D Fig), indicating that the effects of RAMP were specific for the target organelle. In addition, we observed that both the peripherally and centrally clustered LAMP1-SBP-GFP–positive lysosomes in HeLa cells accumulated the fluorescent acidotropic probe LysoTracker Blue (S3A and S3D Fig) and fluid-phase endocytic marker Alexa Fluor 647-dextran (AF647-dextran) (S3B and S3E Fig) and hydrolyzed the fluorogenic protease substrate dye-quenched bovine serum albumin (DQ-BSA) (S3C and S3F Fig). Thus, the redistribution of lysosomes induced by RAMP was specific and did not grossly alter their functional properties.

Addition of 50 μM biotin to cells coexpressing LAMP1-SBP-GFP and mCh-KIF5B*-strep rapidly dissociated mCh-KIF5B*-strep from the lysosomes and reversed the accumulation of both proteins at peripheral sites (Fig 3A and S1 Movie). Similarly, biotin addition dissociated strep-KIFC1*-mCh from the centrally clustered lysosomes, allowing lysosome redistribution towards the cell periphery (Fig 3B and S2 Movie). Quantitative image analysis (S4 Fig) showed that, in both cases, lysosomes recovered their normal steady-state distribution approximately 20–30 minutes after the addition of biotin (Fig 3C). This recovery rate was slower than that predicted by computational methods used to study bidirectional cargo transport undergoing tug-of-war between motor molecules (Fig 3D, and snapshots of the simulations shown in S6A and S6B Fig) (see also S1 Text) [26,27]. The discrepancy between in vivo and in silico experiments could be explained by the contribution of biotin uptake and SBP–streptavidin dissociation rates to the overall kinetics of recovery. Taken together, all of the above results demonstrated the validity of RAMP as a method to redistribute lysosomes in a reversible manner and provided a quantitative framework to analyze the experimental data.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Analysis of the reversal of lysosome redistribution after addition of biotin.

(A,B) HeLa cells were co-transfected with plasmids encoding (A) LAMP1-SBP-GFP and mCh-KIF5B*-strep (see S1 Movie) and (B) LAMP1-SBP-GFP and strep-KIFC1*-mCh (see S2 Movie), resulting in the accumulation of lysosomes at the periphery or center of the cells, respectively. At 24 h after transfection, biotin was added, and the return of lysosomes to their normal distribution was recorded by live-cell imaging. Scale bars: 10 μm. (C) Quantification of the reversal of lysosome redistribution from S1 and S2 Movies (average fractional distance (f = 75%) of lysosomes as a function of time after biotin addition), in comparison to lysosomes expressing only LAMP-SBP-GFP (also check S4 Fig and Methods for details). Values are the mean ± SD from three independent experiments. (D) Results from simulations of a computational model based on a stochastic model of tug-of-war (S1 Text) show the exponential recovery, reaching normal distribution in a WT context of lysosomes undergoing tug-of-war with the endogenous KIF5B and KIF1B kinesins and dynein [28]. Values are the mean ± SEM from 500 independent cargo initial configurations for each condition. Summary data available as Supporting Information (S1_Data.xlsx). GFP, green fluorescent protein; KIF, kinesin superfamily; LAMP, lysosome-associated membrane protein; mCh, mCherry; SBP, streptavidin-binding protein; strep, streptavidin; WT, wild type.

https://doi.org/10.1371/journal.pbio.3000279.g003

A previously used BicD2 fragment has a dominant-negative effect

Previous methods to redistribute organelles towards microtubule minus ends used fusions to an N-terminal fragment from the dynein–dynactin cargo adaptor bicaudal D homolog 2 (BicD2) [9,11,29,30]. This fragment (herein denoted BicD2*) comprises the region that interacts with dynein and dynactin but lacks the C-terminal cargo-binding domain [31]. We observed that an mCh-BicD2*-strep fusion protein (S5A Fig) indeed caused clustering of LAMP1-SBP-GFP-containing lysosomes at the cell center (S5B Fig). Furthermore, upon addition of biotin, lysosomes were released from their central location (S5B Fig and S3 Movie). However, we noticed that rather than returning to their normal steady-state distribution, the released lysosomes became clustered at cell protrusions (S5B Fig and S3 Movie). Expression of mCh-BicD2*-strep alone also caused peripheral accumulation of lysosomes labeled for endogenous LAMP1 and dispersal of endosomes labeled for TfR (S5C and S5D Fig). These experiments thus demonstrated that the N-terminal fragment of BicD2 behaves as a dominant-negative mutant that causes lysosome dispersal per se, probably by outcompeting an endogenous adaptor of lysosomes to dynein–dynactin such as Rab-interacting lysosomal protein (RILP) [32], JNK-interacting protein 3 (JIP3) [33,34], or JIP4 [35]. For this reason, our experiments to cluster organelles at the cell center were done using constructs based on KIFC1* rather than BicD2*.

Analysis of axonal and dendritic transport of lysosomes using RAMP

Next, we extended the testing of RAMP to lysosomes in neurons. Unlike HeLa cells, neurons are highly polarized cells that have specialized peripheral domains such as axons and dendrites with distinct compositions of cytoskeletal structures and organelles [36–38]. The microtubule cytoskeleton, in particular, exhibits a remarkably polarized organization. In the axon, microtubules are long and unidirectional, with their plus ends pointing towards the axon terminal [39,40]. In contrast, dendritic microtubules are short and have a mixed orientation [39–41]. Axonal and dendritic microtubules also have different patterns of tubulin post-translational modifications and microtubule-associated proteins (MAPs) [42,43]. The distinct properties of axonal and dendritic microtubules allow for specialized forms of organelle movement that are not present in other cells. For example, KIF5/kinesin-1 members predominantly move cargo towards axon terminals, while KIF1/kinesin-3 members can move cargo in different neuronal domains [44–47]. How the neuron acquires and maintains this highly polarized organization of microtubules and organelles is a topic of intense interest in neurobiology. The development of RAMP provides a new, to our knowledge, tool to unravel the mechanisms that underlie this organization.

To examine the characteristics of lysosome transport in the axon and dendrites, we started by coexpressing LAMP1-SBP-GFP and mCh-KIF5A*-strep in rat hippocampal neurons in primary culture. In these experiments, we used a motor construct based on KIF5A because this KIF5 paralog is specifically expressed in neurons [48]. We observed that coexpression of LAMP1-SBP-GFP and mCh-KIF5A*-strep resulted in accumulation of lysosomes at the tips of axons but not dendrites (Fig 4A, axon tips indicated by arrows), in accordance with the preference of KIF5A for axonal microtubules [29,44,45,49]. Similar results were obtained using a motor construct based on the ubiquitously expressed KIF5B (S7A Fig). Upon addition of biotin, lysosomes that had accumulated at the axon tip by the action of mCh-KIF5A*-strep were rapidly mobilized in the retrograde direction (lines with positive slopes in the kymographs), with an initial rate of 0.6 μm/s (Fig 4B and S7B Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Analysis of polarized transport of neuronal lysosomes using RAMP.

(A) DIV5 rat hippocampal neurons were transfected with plasmids encoding LAMP1-SBP-GFP alone (first column) or together with plasmids encoding mCh-KIF5A*-strep (second column) or strep-KIFC1*-mCh (third column). Cells were cultured in the presence of NeutrAvidin to remove biotin from the medium. The following day, neurons were fixed and imaged for GFP and mCherry fluorescence. The top two rows show single neurons in the absence of biotin. The bottom strips show 50 μm of straightened axon terminal and 35 μm of the dendrite portion from each of the above neurons. (B,C) DIV5 neurons were transfected with plasmids encoding LAMP1-SBP-GFP and mCh-KIF5A*-strep (B) or strep-KIFC1*-mCh (C), leading to the accumulation of lysosomes at axon tips and soma, respectively. The following day, live neurons were stained with a CF640R-conjugated antibody to neurofascin to label the AIS and thus identify the axon. Biotin was added, and neurons were then imaged live for all fluorescent probes. The top two rows show frames at 0 and 60 minutes after the addition of biotin. The bottom panels show 50 μm of straightened portions and kymographs of the distal axon (B), proximal axon (C), or dendrites (C), as indicated, beginning at 0 and 60 minutes after the addition of biotin. In the kymographs, positive slopes represent retrograde transport, negative slopes anterograde transport, and vertical lines stationary foci. Arrowheads mark the trajectory of the axon. Arrows indicate the axon tip. Scale bars: 20 μm. Summary data available as Supporting Information (S1_Data.xlsx). AIS, axon initial segment; DIV5, day in vitro 5; GFP, green fluorescent protein; KIF, kinesin superfamily; LAMP, lysosome-associated membrane protein; mCh, mCherry; RAMP, reversible association with motor proteins; SBP, streptavidin-binding protein; strep, streptavidin.

https://doi.org/10.1371/journal.pbio.3000279.g004

We also examined the effect of coexpressing LAMP1-SBP-GFP with mCh-KIFC1*-strep. We observed that this manipulation resulted in accumulation of lysosomes in the soma, with concomitant depletion of lysosomes from the axon and reduction in the number of lysosomes in the dendrites (Fig 4A, right panels). Nevertheless, the remaining dendritic lysosomes had a dotted distribution along the entire dendrite (Fig 4A). These findings demonstrated that the KIFC1* motor can drive movement in both axon and dendrites but with different end results, consistent with the distinct organization of axonal and dendritic microtubules. Also, in this case, treatment with biotin rapidly restored lysosome transport in both the axon (initial average rate: 1.1 μm/s) and the dendrites (initial average rates: 0.6 μm/s anterograde, 0.7 μm/s retrograde) (Fig 4C and S7B Fig). Altogether, these studies demonstrated the suitability of RAMP for analyzing the polarized transport of lysosomes in different neuronal domains.

Shape changes of mitochondria accumulated in the cell periphery by RAMP

To test the applicability of RAMP to other organelles, we extended our analyses to mitochondria in HeLa cells (Fig 5). Coexpression of TOM20*-SBP-GFP (Fig 5A) with mCh-KIF5B*-strep or strep-KIFC1*-mCh also caused accumulation of mitochondria at the periphery or the center of the cell, respectively (Fig 5B and 5C; see Fig 5D for quantification). Interestingly, both fluorescence (Fig 5B) and electron microscopy (Fig 5E) showed that mitochondria became rounder when accumulated at the cell periphery by mCh-KIF5B*-strep. Quantification of different morphological parameters from the fluorescent images confirmed that the peripherally distributed mitochondria had a rounder shape, as evidenced by their shorter length, as well as lower aspect ratio and form factor (Fig 5F) [50]. In addition, this analysis revealed that peripherally distributed mitochondria were more numerous and had a smaller area (Fig 5F) [50]. Electron microscopy also showed that, despite their altered shape, the peripherally accumulated mitochondria retained their internal structure, including typical cristae (Fig 5E). They also retained a negative membrane potential, as determined by staining with MitoTracker Red CMXRos [51–53] (S8A Fig). Similar changes in mitochondrial morphology were observed upon peripheral redistribution by coexpression of TOM20*-SBP-GFP with three other constructs comprising the coiled-coil and motor domains from the plus-end–directed kinesin-3 members KIF1A and KIF13A and kinesin-2 member KIF17, all appended with mCherry and streptavidin (S8B and S8C Fig). Since other than mediating plus-end–directed transport, KIF1A, KIF13A, and KIF17 are structurally and functionally distinct from KIF5B, these results indicated that the change in mitochondrial morphology is not specific to KIF5B but is likely dependent on the strong anterograde pulling force and relocation to the cell periphery induced by kinesins [54]. Mitochondria that accumulated at the cell periphery by expression of TOM20*-SBP-GFP with mCh-KIF5B*-strep retained their rounder shape even after reversal of their accumulation by the addition of biotin (Fig 5G, S4 Movie). In contrast, accumulation of mitochondria at the cell center by expression of strep-KIFC1*-mCh (Fig 5C) and reversal of this accumulation by addition of biotin (Fig 5H, S5 Movie) had no effect on the overall shape of mitochondria.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Redistribution of mitochondria to the cell periphery alters their shape.

(A) Colocalization of TOM20*-SBP-GFP with 25 nM MitoTracker Red CMXRos in HeLa cells incubated for 30 minutes at 37°C, 5% CO₂. (B) Coexpression of TOM20*-SBP-GFP with HA-KIF5B*-strep causes peripheral redistribution of mitochondria and alters mitochondrial shape. (C) Coexpression of TOM20*-SBP-GFP with strep-KIFC1*-mCh causes central redistribution of mitochondria without affecting their morphology. In A–C, nuclei were stained with DAPI. The rightmost images in A–C are 3.5× magnified views of the boxed areas. Scale bars: 10 μm. (D) Box-and-whisker plots representing the fractional distance distribution (f = 95%) of mitochondria in the conditions from panels A, B, and C (see S4 Fig and Methods section for details). Summary data available as Supporting Information (S1_Data.xlsx). (E) Electron microscopy of cells expressing TOM20*-SBP-GFP alone (top row) or with HA-KIF5B*-strep (bottom row), where rounder mitochondria with normal internal organization is shown. Images on the right are magnifications of boxed areas on the left. Left column scale bar: 4 μm; right column scale bar: 0.6 μm. (F) Quantification of the phenotype observed in (B), using mitochondrial shape descriptors [50] (see Methods section for more details). Summary data available as Supporting Information (S1_Data.xlsx). (G,H) HeLa cells coexpressing TOM20*-SBP-GFP with mCh-KIF5B*-strep (G) (S4 Movie) or strep-KIFC1*-mCh (H) (S5 Movie) were incubated with biotin and analyzed by live-cell imaging. Scale bar: 10 μm. Rightmost images are 4.5× enlargements of the boxed areas. GFP, green fluorescent protein; HA, hemagglutinin; KIF, kinesin superfamily; mCh, mCherry; SBP, streptavidin-binding protein; strep, streptavidin; TOM, translocase of the outer membrane.

https://doi.org/10.1371/journal.pbio.3000279.g005

Mitochondrial morphology is maintained by a dynamic equilibrium between fusion and fission. Mitochondrial fusion is achieved by mitofusins 1 and 2 [55], two GTPases that tether mitochondria and allow membrane fusion [56,57]. Mitochondrial fission is coordinated by a much more complicated mechanism involving the ER [58], actin polymerization [59], myosin [60], ER–plasma membrane contacts [61], and several dynamin‐like proteins and their adaptors [62–64]. RAMP could thus be used to investigate how mitochondrial positioning affects these machineries.

Manipulation of the distribution of ER and peroxisomes using RAMP

We also used RAMP to manipulate the distribution of the ER (Fig 6) and peroxisomes (S9 Fig) in HeLa cells. When GFP-SBP-VAP-A* was expressed alone, it localized to the ER network (Fig 6A), and its coexpression with hemagglutinin (HA)-tagged KIF5B*-strep or strep-KIFC1* shifted its overall distribution towards the cell periphery or the cell center, respectively (Fig 6B and 6C). In all cases, mCh-SBP-VAP-A*–positive structures costained for the endogenous ER protein calnexin, indicating that RAMP caused redistribution of the ER and not just mCh-SBP-VAP-A* (Fig 6B and 6C). The redistribution of the ER to the center or periphery of the cell, however, was partial and did not lead to complete disruption of the ER network. The residual structures could be resistant to redistribution because they are tightly tethered to the nucleus, the plasma membrane, or cytoskeletal structures. ER exit sites (ERESs) [65] also underwent redistribution after RAMP (Fig 6D), demonstrating that ER-associated structures can also be manipulated with this tool. As for other organelles, addition of biotin restored the normal distribution of GFP-SBP-VAP-A* (Fig 6E and 6F; S6 and S7 Movies).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Redistribution of the ER with RAMP.

(A) Confocal microscopy of HeLa cells expressing GFP-SBP-VAP-A* shows localization of the fusion protein to a network characteristic of the ER shown by costaining for endogenous calnexin, a luminal ER-resident protein. Insets correspond to 1.6× magnified views of the boxed areas. (B,C) HeLa cells were co-transfected with plasmids encoding mCh-SBP-VAP-A* and HA-KIF5B*-strep (B) or strep-KIFC1*-HA (C) and stained for endogenous calnexin. At 24 h after transfection, cells were rinsed with PBS and immediately fixed with paraformaldehyde. Notice that the ER network mostly follows the redistribution caused by RAMP with mCh-SBP-VAP-A*. Insets correspond to 1.2× magnified views of the boxed areas. (D) HeLa cells were co-transfected with plasmids encoding mCh-SBP-VAP-A* and HA-KIF5B*-strep or strep-KIFC1*-HA and stained for endogenous Sec31A. At 24 h after transfection in absence of biotin, cells were rinsed with PBS and immediately fixed with paraformaldehyde. Notice that the ERESs also follow the redistribution caused by RAMP. In A–D, nuclei were stained with DAPI. (E,F) HeLa cells coexpressing GFP-SBP-VAP-A* and mCh-KIF5B*-strep (E) or strep-KIFC1*-mCh (F) were analyzed by time-lapse microscopy after addition of biotin (see S6 and S7 Movies, respectively). Scale bars: 10 μm. ER, endoplasmic reticulum; ERES, ER exit site; GFP, green fluorescent protein; HA, hemagglutinin; KIF, kinesin superfamily; mCh, mCherry; RAMP, reversible association with motor proteins; SBP, streptavidin-binding protein; Sec31A, secretory protein 31A; strep, streptavidin; VAP, VAMP-associated protein.

https://doi.org/10.1371/journal.pbio.3000279.g006

PEX3*-SBP-GFP alone colocalized with endogenous peroxisomal membrane protein 70 (PMP70) but not early endosome antigen 1 (EEA1) (S9A Fig). Coexpression of PEX3*-SBP-GFP with HA-KIF5B*-strep or Strep-KIFC1*-HA caused accumulation of peroxisomes at the cell periphery or the cell center, respectively (S9B and S9C Fig; S9D Fig for a quantification of the redistributions), as previously shown for lysosomes (Figs 2 and 3) and mitochondria (Fig 5). Treatment with biotin dissociated the kinesin constructs and allowed the peroxisomes to return to their normal steady-state distribution (S9E and S9F Fig; S8 and S9 Movies).

Ethics statement

Animal procedures, including preparation of rat hippocampal neurons, were conducted under protocol #16–011 approved by the NICHD Animal Care and Use Committee, in adherence to the NIH Guide for the Care and Use of Laboratory Animals.

Molecular cloning

DNA sequences encoding fragments from mouse KIF5A (KIF5A* = 1–559), mouse KIF5B (KIF5B* = 1–555), mouse KIF1A (KIF1A* = 1–405), human KIF13A (KIF13A* = 1–639), and human KIF17 (KIF17* = 1–509) were cloned into pmCherry-C1 (Clontech, Takara Bio, Mountain View, CA, USA). These fragments comprise the complete motor and neck domains and part of the coiled-coil domain required for dimerization and movement after recruitment to the target organelle [44]. DNA sequences encoding a 6×(glycine–serine) linker and streptavidin were cloned downstream of the truncated kinesins to complete the anterograde motor constructs (Fig 1B and S8B Fig). DNA sequences encoding streptavidin, followed by a 6×(glycine–serine) linker and a fragment from human KIFC1 (KIFC1* = 125–673), were cloned into pmCherry-N1 (Clontech) to create a retrograde motor construct (Fig 1B). We also engineered a BicD2-based dynein–dynactin adaptor by cloning DNA sequences encoding a mouse BicD2 fragment (BicD2* = 15–595) [29], followed by a linker (RTVSR) and streptavidin into pmCherry-C1 (Clontech). In addition, we made HA-epitope–tagged motor constructs (HA-KIF5B*-strep and strep-KIFC1*-HA) by restriction-free cloning to replace the mCherry protein with the HA epitope. The different cargo constructs carrying the SBP were generated by insertion of the SBP sequence (DEKTTGWRGGHVVEGLAGELEQLRARLEHHPQGQREP) between the organellar protein or protein fragment of interest and the fluorescent reporter protein (EGFP[A206K], a monomeric GFP variant [67], or mCherry) (Fig 1A). Full-length rat LAMP1 and fragments from human TOM20 (TOM20* = 1–30) [68], human PEX3 (PEX3* = 1–42) [29], and human VAP-A (VAP-A* = 118–249) cDNAs were used to target the SBP–GFP tandem to lysosomes, mitochondria, peroxisomes, and ER, respectively. The streptavidin and SBP peptide cDNAs were obtained from the bicistronic expression plasmid encoding the ER hook (streptavidin–KDEL) and the reporter ManII-SBP-GFP, respectively, described by Boncompain and colleagues [18]. Both plasmids were gifts from F. Perez (Institut Curie, Paris, France). All constructs were verified by DNA sequencing.

Antibodies and other reagents

Rabbit anti-LAMTOR4 (cat. 13140) (1:200) was from Cell Signaling Technology (Danvers, MA, USA). Mouse anti-α-tubulin (DM1A; cat. T6199) (1:1,000) and D-biotin (cat. 47868) were from Sigma-Aldrich (St. Louis, MO, USA). Rabbit anti-TfR (cat. ab84036) (1:200) and rabbit anti-PMP70 (cat. ab3421) (1:500) were from Abcam (Cambridge, UK). Rabbit anti-Sec31A (cat. 17913-1-AP) (1:500) was from Proteintech (Rosemont, IL, USA). Mouse anti-calnexin (clone C8.B6; cat. MAB3126) (1:1,000) was from EMD Millipore (Burlington, MA, USA). Mouse anti-pan-neurofascin (external) (clone A12/18; cat. 75–172) (1:50) was from the University of California, Davis/NIH NeuroMab Facility (Davis, CA, USA). Mix-n-Stain-CF640R (cat. 92245) was from Biotium (Fremont, CA, USA) and was used to label the antibody against neurofascin. Mouse anti-LAMP1 (H4A3) (1:500) was from the Developmental Studies Hybridoma Bank (University of Iowa, Iowa City, IA, USA). Mouse anti-EEA1 (cat. 610457) (1:1,000) was from BD Biosciences (San Jose, CA, USA). All secondary antibodies (donkey anti-rabbit and anti-mouse conjugated with Alexa Fluor 488, 555, or 647) (1:1,000) were from Invitrogen (Carlsbad, CA, USA). NeutrAvidin Protein (cat. 31050), LysoTracker Blue DND-22 (cat. L7525), AF647-dextran (cat. D22914), MitoTracker Red CMXRos (cat. M7512), and DQ Red BSA (cat. D12051) were from Thermo Fisher Scientific (Waltham, MA, USA).

Cell culture and transfection

HeLa, COS-7, and HT-1080 cells were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) (Corning, Corning, NY, USA) supplemented with 10% fetal bovine serum (Corning), 100 IU/mL penicillin and 100 μg/mL streptomycin (Corning), and 2 mM L-glutamine (Corning) in a 37°C incubator with 5% CO₂ and 95% air. Primary cultures of rat hippocampal neurons were prepared at embryonic day 18 and cultured as previously described [69]. Cells were seeded in 24-well plates with glass coverslips or 8-well chambered cover glass (Cellvis, Mountain View, CA, USA). Plasmids were transfected using Lipofectamine 2000 (Thermo Fisher Scientific) following the manufacturer’s instructions. Unless otherwise stated, cells were analyzed 24 h after transfection.

Fluorescence microscopy

For fixed-cell immunofluorescence microscopy experiments, coverslips were washed in PBS and fixed for 12 minutes in 4% paraformaldehyde and 4% sucrose in PBS at room temperature. Following fixation, coverslips were washed twice for 5 minutes in PBS, permeabilized for 15 minutes in 0.2% (v/v) Triton X-100, and blocked for 30 minutes in 0.2% porcine gelatin (Sigma-Aldrich) in PBS. Staining was carried out for 30 minutes at 37°C with different primary antibodies diluted in 0.2% gelatin in PBS (immunofluorescence buffer). Coverslips were then washed twice with PBS and incubated for 30 minutes at 37°C with Alexa-Fluor–conjugated secondary antibodies diluted in immunofluorescence buffer. Coverslips were next washed twice with PBS and mounted on slides using Fluoromount-G with or without DAPI (Electron Microscopy Sciences, Hatfield, PA, USA). Images were obtained on an inverted confocal laser scanning microscope (LSM880; Carl Zeiss, Oberkochen, Germany) fitted with a 63× 1.4 NA objective. Image analysis was performed using ImageJ (Rasband, W.S., ImageJ, U. S. National Institutes of Health, Bethesda, MD, USA, https://imagej.nih.gov/ij/, 1997–2016).

Live-cell imaging

Live-cell imaging was conducted with an Eclipse Ti Microscope System (Nikon, Tokyo, Japan) equipped with an environmental chamber (temperature controlled at 37°C and CO₂ at 5%) and NIS-Elements AR microscope imaging software. Cells were kept in microscopy medium (DMEM without phenol red, 1% bovine serum albumin [Sigma-Aldrich] and 25 mM HEPES [Corning] [pH 7.4], sterile filtered) during the entire imagining time. Spinning disk confocal images were taken with a Plan Apo VC 60× objective (NA 1.40) and a high-speed electron-multiplying charge-coupled device camera (Evolve 512; Photometrics, Tucson, AZ, USA). The focus on the sample was maintained using the Perfect Focus system. Dual-color imaging was done by fast switching of the excitation lasers, and images from green and red channels were aligned automatically. Laser power and exposure times were optimized depending on transfection efficiency and frame-rate acquisition.

Electron microscopy

HeLa cells were grown in standard culture medium on borosilicate glass coverslips; fixed with 2.5% glutaraldehyde, 2% formaldehyde, and 2 mM CaCl₂ in 0.1 M sodium cacodylate (pH 7.4) for 60 minutes; postfixed in 1% osmium tetroxide; and stained en bloc with aqueous 2% uranyl acetate overnight. Samples were dehydrated in increasing concentrations of ethanol, penetrated with epoxy resin, and subsequently polymerized at 65°C for 60 h. Coverslips were removed by hydrofluoric acid, and blocks were cut out and remounted on a microtome holder. Thin sections (80 nm) were cut parallel to the surface of the block, mounted on copper grids covered with formvar/carbon film, and stained with uranyl acetate and lead citrate. Samples were analyzed with an FEI Tecnai T20 microscope (Thermo Fisher Scientific) at 120 kV, and images were recorded with an AMT XR81 CCD camera (AMT, Woburn, MA, USA).

Quantification of lysosome movement after RAMP reversal

Time-lapse microscopy analysis of lysosome movement after addition of biotin was performed as previously described [7], with some changes (S4 Fig). Control videos were generated by expression of only LAMP1-SBP-GFP, while RAMP experiments involved the coexpression of LAMP1-SBP-GFP with a streptavidin-tagged truncated kinesin plasmid (mCh-KIF5B*-strep or strep-KIFC1*-mCh). Unless otherwise indicated, addition of biotin corresponded to time 0. Frames from live-cell imaging were analyzed using ImageJ. Z-stacks for each frame were acquired and maximum intensity projections were generated before analysis. Individual cells were masked to exclude contributions from nearby cells. The Radial Profile Extended plug-in for ImageJ was used to calculate the radial profile (fluorescence intensity of the green channel as a function of radial distance—bin size was pixel size—where the center was set at the center of the nucleus) for each frame of the video. Next, distances were normalized, adjusting to “1” as the largest circle used in the radial profile generation (i.e., the most external circle that coincided with the border of the cell) and “0” as the center of the nucleus. Finally, from these profiles, we retrieved the radius required to include a fraction f of lysosomes by finding the minimal fractional distance at which the desired fraction of the total intensity was reached. We observed that in most of the videos analyzed, maximum difference between initial (time 0, no biotin) and final (normal distribution) average fractional distances was achieved when including 70%–95% of the total fluorescence (S4D Fig). This average fractional distance was plotted against time, and curves to analyze the recovery of the organelle distribution were used in Fig 3.

Modeling of tug-of-war

The model, parameters, and simulations are described in detail in the Supporting Information (S1 Text).

Analysis of lysosome movement in neurons

Movement of vesicles containing LAMP1-SBP-GFP protein was analyzed using ImageJ. Lines (1 pixel wide) were traced from the soma to axonal tips or from the soma to dendrite terminals, and the selection was straightened for all frames. Kymographs were generated by reslicing the straightened lines, followed by Z-projection. Anterograde or retrograde movement was manually determined from the kymographs by looking at the slopes of the generated lines, with negative and positive slopes representing anterograde and retrograde movement, respectively. Speed was calculated by manually measuring the initial and final position from lines in the kymographs to determine the path length in μm, divided by the elapsed time in seconds.

Analysis of mitochondrial morphology

Mitochondrial morphology was analyzed as previously described [50]. In brief, confocal images for control (TOM20*-SBP-GFP only) and peripheral mitochondria (TOM20*-SBP-GFP and mCh-KIF5B*-strep coexpression) were acquired in HeLa cells transfected and fixed 24 h after transfection. The green channel (mitochondria) of these images was digitized using the ImageJ plug-in provided by Merrill and colleagues [50]. The following procedure was applied to each image: subtraction of background using a rolling radius of 50, followed by despeckle noise removal, contrast-limited histogram equalization (CLAHE) for enhancing local contrast (blocksize = 9 and maximum = 4) and a bandpass filtering (small = 0 and large = 256; suppress = none). Following autothresholding (method = Li), the “Analyze Particles” routine was used as originally provided with the macro to deliver the measurements. Six descriptors were obtained and analyzed: 1) length; 2) number of mitochondria; 3) area² (average size of mitochondria weighted towards larger mitochondria); 4) aspect ratio (ratio of the major and short axis of an ellipse fitted to each mitochondria; the closer to 1 [minimum value], the rounder the mitochondria); 5) form factor (FF) (describes the mitochondria’s shape complexity; the minimum value of 1 indicates a perfect circle); and 6) area-weighted FF (variant of FF that averages with a bias towards larger mitochondria or mitochondrial networks). As suggested by the authors [50], mitochondrial morphology can be considered changed if at least three of the metrics show significant differences, and the trends are apparent to a blinded observer even when analyzing raw images.

Material availability

The RAMP plasmids have been deposited in Addgene (IDs: 120163 to 120175). All other plasmids used in this study are available from the corresponding author upon request.

Statistical analysis

Statistical analysis was performed using Prism 6 (GraphPad). Data are represented as mean ± SD (Fig 3C, and S7B Fig) or mean ± SEM (Fig 3D). Box-and-whisker plots show median, first and third quartiles, and maximum and minimum values (Figs 2F, 5D and 5F and S5D and S9D Figs). Summary data available as Supporting Information (S1 Data). p-values were determined using two-tailed Student t test for unpaired data or one-way ANOVA followed by Tukey’s test for multiple comparison of data in two or more groups (* p ≤ 0.05; ** p ≤ 0.01; *** p ≤ 0.001; **** p ≤ 0.0001. Nonsignificant differences were not highlighted for clarity in the figure).