Eukaryotes compartmentalize cellular functions within distinct organelles, and regulation of organelle position is critical for cell health. Organelles are transported bidirectionally by motor and adaptor proteins along microtubules1, which are modulated by multiple post-translational modifications that comprise part of the ‘tubulin code’. Although this code has been implicated in cargo selection and directed organelle movement2, how it is decoded to mediate transport and control distribution remains largely unknown.

Endoplasmic reticulum (ER) comprises structurally and functionally divergent membrane compartments that include interconnected tubules, perinuclear matrices and sheets, and the nuclear envelope3,7. The ER is a compelling candidate for exploiting the complexity of the tubulin code, since it spreads throughout the cytoplasm in association with microtubules and makes abundant organelle contacts8,9,10. Most studies of ER shaping and organelle contacts have emphasized peripheral tubular ER. How the denser perinuclear ER is shaped and asymmetrically distributed remains largely unknown, although three ER membrane-bound proteins—CLIMP63, p180 and KTN1—localize prominently to perinuclear ER and are considered sheet-forming proteins11. Even so, depletion of CLIMP63 may paradoxically lead to the expansion of ER matrices or sheets in the periphery11,12, and perinuclear ER matrices or sheets remain abundant even upon simultaneous knockdown of all three proteins, prefiguring more complex functional roles11.

CLIMP63, p180 and KTN1 position ER

We used CRISPR–Cas9 to knock out these proteins in human U2OS cells stably expressing the ER marker mEmerald–Sec61β (Extended Data Fig. 1a, b). As previously reported11,12, peripheral ER in CLIMP63-knockout cells is populated with increased numbers of dense matrices or sheets—a ‘dispersed’ phenotype. KTN1 knockout also disperses ER, whereas p180-knockout cells exhibit a contrasting ‘clustered’ ER phenotype, with the peripheral network remaining tubular and perinuclear ER collapsing asymmetrically into a smaller area at one side of the nucleus (Fig. 1a, Extended Data Fig. 1c, d). These morphologic changes are not secondary to alterations in levels of other ER-shaping proteins or cell cycle disruption (Extended Data Fig. 1b, e, f). Double knockout of CLIMP63 and KTN1 substantially disperses ER. Conversely, ER in CLIMP63 and p180 double-knockout cells resembles the wild type, consistent with their opposing single-knockout phenotypes. Surprisingly, p180 and KTN1 double knockout causes more ER clustering than in p180-knockout cells (Fig. 1a, Extended Data Fig. 1d), suggesting a more complex interplay. In CLIMP63–p180–KTN1 triple-knockout cells, high-density ER matrices or sheets are abundant in the perinuclear region (Fig. 1a), although perinuclear ER appears less evenly distributed compared with wild-type cells, with ‘hot spots’ (Extended Data Fig. 1g) that may reflect ER positioning defects.

Fig. 1: CLIMP63, p180 and KTN1 differentially regulate ER morphology.
figure 1

a, Representative images of wild-type (WT), knockout (KO), double knockout (DKO) or triple knockout (TKO) of CLIMP63, p180 and/or KTN1 in U2OS cells stably expressing mEmerald–Sec61β (green, ER marker) and co-labelled with DAPI (blue, nuclear marker), anti-α-tubulin (red, microtubule marker) and anti-TRAPα (magenta, rough ER marker). Perinuclear and peripheral regions (left, outlined) are enlarged on the right. Scale bar, 10 μm. b, c, ER MDR (b) and asymmetry (c) (see Supplementary Text) in cells as in a. n = 30 cells. df, Quantifications of ER morphology in wild-type or CLIMP63-knockout (d), p180-knockout (e) or KTN1-knockout (f) cells expressing the indicated CLIMP63, p180 or KTN1 constructs. Since recombinant KTN1 levels are typically very low, only cells with detectable KTN1–mApple signals were quantified. Ability (+) or inability (−) of the mutants to bind microtubules (MT) is indicated. n = 32, 30, 30, 30, 31, 30 and 30 cells (left to right) in d; n = 59, 62, 64, 61, 34, 70 and 70 (left to right) in e; and n = 32, 38, 38, 39, 40 and 36 (left to right) in f. Data are mean ± s.d. with individual data points shown. Two-tailed t-test; P values are shown.

Source data

To quantitatively assess changes in ER morphology and distribution, we devised complementary algorithms. First, we harnessed a statistical approach based on probability density estimation to analyse spatial distributions of fluorescently labelled ER and other organelles. Next, we used an experimentally derived spatial probability mass function, which quantifies fluorescence changes across an image, to calculate the radial distribution and degree of cellular asymmetry of organelles (Extended Data Fig. 2a–g, Supplementary Text). Single or double knockout of CLIMP63 and KTN1 increases ER mean distribution radius (MDR) (Fig. 1b), indicating that ER is spread more peripherally. By contrast, p180 knockout or p180 and KTN1 double knockout increases ER asymmetry (Fig. 1c). Quantification assessing the rough ER marker TRAPα instead of mEmerald–Sec61β shows similar results (Extended Data Fig. 2h, i). Microtubule MDR and asymmetry change only slightly (Extended Data Fig. 2j–m).

ER proteins bind subsets of microtubules

We assessed microtubule binding of numerous ER proteins by co-sedimentation. CLIMP63 and p180, both known microtubule-binding proteins13,14, co-sediment with microtubules as expected. KTN1 also sediments robustly with microtubules (Extended Data Fig. 3a, b). Since full-length p180 (p180L) is degraded during cell lysis (Extended Data Fig. 3a, c), we used a smaller, more stable splice variant (p180s) that lacks the numerous ribosome-binding decapeptide repeats present in p180L (Extended Data Figs. 1a, 3b); p180s was undetectable by immunoblotting in the cell lines studied, facilitating identification of recombinant protein (Extended Data Fig. 3d). For each protein, we mapped microtubule-binding domains; only wild-type proteins or mutants capable of microtubule binding restored ER morphology in corresponding knockout cell lines (Fig. 1d–f, Extended Data Figs. 3, 4). For instance, CLIMP63 missense mutants R7A, K10A and R70A did not bind microtubules or suppress ER distribution defects in CLIMP63-knockout cells, whereas CLIMP63(H69A), which binds microtubules, rescues the phenotype. A phosphomimetic CLIMP63 mutant defective in microtubule binding13 also did not rescue ER distribution defects (Extended Data Fig. 4a–d). For KTN1, only the deletion mutant that binds microtubules suppressed the abnormal ER phenotype (Fig. 1f, Extended Data Fig. 4g–i). Finally, p180s lacking the kinesin-1 binding domain still suppressed the clustered ER phenotype in p180-knockout cells (Extended Data Fig. 4j–l). Thus, despite distinct phenotypes, ER morphology changes in CLIMP63-, p180- and KTN1-knockout cells are likely to all reflect alterations in microtubule binding.

We hypothesized that these proteins bind different microtubule populations and used a proximity ligation assay (PLA) to visualize their microtubule associations in cells (Extended Data Fig. 5a–c). We depleted centrosomal microtubules using centrinone B treatment15, and Golgi-derived microtubules by knocking down AKAP45016 (Extended Data Fig. 5d). We found that microtubule association of CLIMP63 was sensitive to centrosome depletion but not Golgi microtubule depletion, whereas KTN1–microtubule association was sensitive to both; p180–microtubule association was not sensitive to depletion of either centrosomes or Golgi microtubules (Extended Data Fig. 5e–h). Admittedly, these microtubule subsets can be interdependent, and centrosome depletion can boost AKAP450-dependent microtubule nucleation at the Golgi16,17. Even so, disrupting Golgi microtubules did not alter centrosome activity17 or CLIMP63–microtubule association (Extended Data Fig. 5e, f).

We inferred that CLIMP63 preferentially binds centrosomal microtubules, KTN1 preferentially binds perinuclear microtubules derived from either centrosome or Golgi, and p180 preferentially binds more peripheral microtubules regardless of origin. In this scenario, PLA distributions for microtubules with CLIMP63 should be more asymmetric than with p180 or KTN1, and PLA distributions for p180 and microtubules should be more dispersed. Indeed, PLA signals for CLIMP63 and microtubules were more asymmetric than those for p180 and KTN1 with microtubules. However, PLA MDR for p180 with microtubules resembled that for KTN1 with microtubules in wild-type cells (Extended Data Fig. 5i). We reasoned that because ER is densely packed perinuclearly in wild-type cells, PLA signals were also mostly perinuclear, making differences challenging to identify. As a workaround, we quantified PLA distributions in CLIMP63-knockout cells, in which ER is more dispersed (Fig 1a); MDR for p180–microtubule PLA signals was larger than MDR for KTN1–microtubule PLA signals (Extended Data Fig. 5j), suggesting that p180 binds more peripheral microtubules than KTN1. Consistent with this specificity, centrosome depletion led to highly dispersed ER in wild-type but not p180-knockout cells, whereas depletion of Golgi-derived microtubules clustered ER in wild-type but not CLIMP63 and KTN1 double-knockout cells (Extended Data Fig. 5k–n).

Graded binding to modified microtubules

For regulatory specificity, microtubules undergo reversible post-translational modifications including acetylation, detyrosination and glutamylation, which together constitute key elements of the tubulin code2. Although CLIMP63, p180 or KTN1 knockout did not affect overall levels of these modifications, tubulin polyglutamylation was decreased in centrosome or Golgi microtubule-depleted cells (Extended Data Fig. 6a–c). We thus considered whether variations in tubulin glutamylation underlie binding selectivity for different microtubule populations and differential effects of the proteins on ER distribution.

CLIMP63 overexpression caused tight ER–microtubule alignment13 that is suppressed in centrosome-depleted cells, whereas p180s or KTN1 overexpression did not trigger ER–microtubule alignment (Extended Data Fig. 6d–f). Co-expression of TTLL4, which monoglutamylates microtubules18 (Extended Data Fig. 6g, h), slightly enhanced ER–microtubule alignment in p180- but not KTN1-overexpressing cells (Extended Data Fig. 6d, e). By contrast, co-expression of TTLL7, which polyglutamylates microtubules19,20 (Extended Data Fig. 6g, h), led to significant microtubule–ER alignment in both p180- and KTN1-overexpressing cells (Extended Data Fig. 6d, e). Although co-expression of TTLL7 slightly enhanced ER-microtubule alignment in CLIMP63-overexpressing cells, co-expression with TTLL4 or microtubule de-glutamylases CCP1 or CCP5 (CCP1 shortens glutamate chains, whereas CCP5 is thought to remove the branch-point glutamate21, Extended Data Fig. 6g, i) did not influence ER–microtubule alignment (Extended Data Fig. 6d, e). Since CLIMP63–microtubule associations as assessed using PLA were unaffected by overexpression of TTLL4, TTLL7, CCP1 or CCP5 (Extended Data Fig. 6j), we inferred that CLIMP63–microtubule binding is not altered by changes in microtubule glutamylation. PLA signals for KTN1–tubulin were significantly increased by TTLL7 but not TTLL4 and decreased in cells overexpressing CCP1 or CCP5 (Extended Data Fig. 6k). By contrast, PLA signals of p180–tubulin were slightly increased by TTLL4 overexpression, markedly increased by TTLL7, slightly decreased by CCP1, and significantly decreased by CCP5 (Extended Data Fig. 6l). We conclude that KTN1 and p180 respond differentially to glutamylation levels, with KTN1 preferentially associating with polyglutamylated versus monoglutamylated microtubules, whereas p180 broadly associates with mono- and polyglutamylated microtubules.

We purified fragments of p180, KTN1 and CLIMP63 containing their microtubule-binding domains (Extended Data Fig. 6m, n) and investigated binding to differentially glutamylated microtubules in vitro, using TTLL6 to generate microtubules functionalized with polyglutamate chains of various lengths primarily on α-tubulin18 (Extended Data Fig. 6o). Both p180 and KTN1 showed substantial increases in binding to microtubules polyglutamylated by TTLL6, with only background binding to unmodified microtubules (Fig. 2a, b). Moreover, as the average glutamate number <nE> on α-tubulin increased from 3.5 to 8.3, binding affinities increased in lockstep, with 2.7- and 5.6-fold increases for p180 and KTN1, respectively. Notably, p180 had a 2.9-fold stronger affinity than KTN1 for microtubules with shorter chains (Fig. 2b). Next, we interrogated how microtubule binding is affected by β-tubulin monoglutamylation induced by TTLL4 and polyglutamylation induced by TTLL718,19 (Extended Data Fig. 6p). Both p180 and KTN1 showed binding preferences toward microtubules functionalized with polyglutamates by TTLL7 (Fig. 2c, d) but weaker binding to TTLL4-modified microtubules (Fig. 2c, e). Of note, p180 exhibited higher in vitro binding (3.9-fold) to microtubules monoglutamylated by TTLL4 compared with KTN1, whereas p180 and KTN1 bound similarly to microtubules polyglutamylated by TTLL7 (Fig. 2c, e). This difference was evident even though numbers of glutamates added by TTLL4 (mean of 1.5) and TTLL7 (mean of 1.2) were similar (Extended Data Fig. 6p), indicating that KTN1 prefers polyglutamate chains introduced by TTLL7 to multiple monoglutamates introduced by TTLL4. At higher glutamylation levels, both KTN1 and p180 formed patches on the microtubule lattice (Fig. 2c), indicative of cooperative binding that may be physiologically relevant in cells when these molecules are tethered and concentrated on the ER membrane. In contrast to p180 and KTN1, CLIMP63 was less responsive to microtubule glutamylation; it lacked detectable binding to unmodified and polyglutamylated microtubules with <nE> of 2.7, exhibiting microtubule binding only when <nE> reached 3.8 (Extended Data Fig. 6q, r). Thus, hyperglutamylation can enhance CLIMP63–microtubule binding in vitro, but since overexpression of CCP1 or CCP5 did not seem to affect CLIMP63–microtubule binding in cells (Extended Data Fig. 6e, j), and centrosome depletion suppressed CLIMP63 overexpression-mediated ER–microtubule alignment (Extended Data Fig. 6f), a different tubulin modification or interaction probably mediates preferential binding of CLIMP63 with centrosome microtubules.

Fig. 2: p180 and KTN1 bind glutamylated and polyglutamylated microtubules, respectively.
figure 2

a, Representative micrographs of p180 and KTN1 fragments (cyan) binding to unmodified microtubules or microtubules glutamylated in vitro by TTLL6 (magenta). Internal reflection microscopy images for the microtubule channel were background subtracted and inverted. Average numbers of glutamate molecules (E) added to microtubules as quantified from mass spectroscopy data in each group are indicated. Scale bar, 5 μm. b, Binding of p180 and KTN1 fragments to unmodified and TTLL6-glutamylated microtubules. n = 110 (unmodified), 165 (α + 3.5 E, β + 1.3 E) and 112 (α + 8.3 E, β + 2.6 E) microtubules for p180; n = 141 (unmodified), 186 (α + 3.5 E, β + 1.3 E) and 156 (α + 8.3 E, β + 2.6 E) microtubules for KTN1. c, Representative micrographs of p180 and KTN1 microtubule interacting fragments (cyan) showing binding to unmodified microtubules or microtubules glutamylated by TTLL4 or TTLL7 (magenta). KTN1 and p180 are shown with different brightness/contrast settings for TTLL4- and TTLL7-modified microtubules, reflecting large differences in binding between mono- and polyglutamylated microtubules. Scale bar, 5 μm. d, e, Affinities of p180 and KTN1 for unmodified microtubules and microtubules glutamylated by TTLL7 (d) or TTLL 4 (e). The x-axis shows weighted averages of glutamate residues attached to α- and β-tubulin. n = 185 (unmodified) and 117 (α + 0 E, β + 1.2 E) microtubules for p180; n = 128 (unmodified) and 225 (α + 0 E, β + 1.2 E) microtubules for KTN1; n = 179 (unmodified) and 237 (α + 0 E, β + 1.5 E) microtubules for p180; n = 163 (unmodified) and 224 (α + 0 E, β + 1.5 E) microtubules for KTN1. f, CLIMP63 binds centrosomal microtubules, KTN1 binds perinuclear polyglutamylated microtubules, and p180 can bind peripheral microtubules with less glutamylation. Together, these proteins maintain proper asymmetric ER distribution, which regulates organelle distributions. MT, microtubules. Data are mean ± s.d. with individual data points shown. Kruskal–Wallis (b) or Mann–Whitney tests (d, e); P values are shown.

Source data

Glutamylation regulates ER distribution

Perinuclear microtubules harbour more polyglutamylation, whereas monoglutamylation is generally more prominent peripherally (Extended Data Fig. 7a–d), consistent with KTN1 binding preferentially to perinuclear microtubules and p180 binding to peripheral microtubules. TTLL overexpression glutamylated microtubules throughout the cell, eliminating the relatively discrete perinuclear distribution of polyglutamylated microtubules and thus drawing ER towards the cell periphery, whereas overexpression of CCP1 or CCP5 decreased binding of p180 and KTN1 to perinuclear microtubules; thus, overexpression of TTLL, CCP1 or CCP5 all lead to dispersed ER (Extended Data Fig. 7e–g). In p180 and KTN1 double-knockout cells, overexpression of TTLL4, CCP1 or CCP5 did not change ER MDR (Extended Data Fig. 7h), yet TTLL7 overexpression still had minor effects, possibly through other pathways. When TTLL7 is overexpressed, KTN1 should bind all microtubules, rather than preferring perinuclear ones. Thus, with TTLL7 overexpression, KTN1 knockout resulted in less dispersed ER (Extended Data Fig. 7i). We also knocked down CCP5, which increases tubulin glutamylation. Similar to TTLL4 overexpression, CCP5 knockdown dispersed ER (Extended Data Fig. 7j–m).

We examined several cell lines widely used in ER morphology studies to assess whether they had different microtubule glutamylation levels. Notably, COS7 cells had particularly high polyglutamylation levels (Extended Data Fig. 7c), and although polyglutamylation in COS7 cells remained relatively more perinuclear compared to monoglutamylation and microtubule distribution, the difference was much less than in U2OS cells (Extended Data Fig. 7b, d). We hypothesized that KTN1 knockout in COS7 cells would show a distinct ER phenotype, possibly mimicking TTLL7 overexpressing cells (Extended Data Fig. 7i). Indeed, although knockout of CLIMP63 or p180 in COS7 cells showed similar phenotypes as in U2OS cells, KTN1 knockout in COS7 cells led to clustered ER (Extended Data Fig. 8a–d), in contrast to dispersed ER in KTN1 knockout U2OS cells. Moreover, overexpression of CCP6 (which has similar activity to CCP1) also led to clustered ER (Extended Data Fig. 8e–h). We conclude that CLIMP63, p180 and KTN1 preferentially bind centrosomal, polyglutamylated and glutamylated microtubules, respectively, to cooperatively distribute ER (Fig. 2f).

Organelle positioning and glutamylation

Live imaging of six organelles22 simultaneously revealed that most have a distribution similar to ER (Extended Data Fig. 9a), suggesting that ER might broadly regulate organelle distribution. Notably, in CLIMP63-, p180- and KTN1-knockout cells, all organelles that we examined exhibited similar distribution changes to those of ER—more dispersed in CLIMP63- or KTN1-knockout cells and more asymmetric in p180-knockout cells (Extended Data Fig. 9b–d). Moreover, CCP1 overexpression, which disperses ER, also increased MDR for lysosomes, mitochondria and peroxisomes in wild-type cells but not in p180 and KTN1 double-knockout cells (Extended Data Fig. 9e–g). Thus, perinuclear ER morphology specifies the distributions of other organelles downstream of microtubule glutamylation.

ER and lysosome movements in autophagy

Perinuclear lysosome clustering, a signature event in early autophagy, is important for proper autophagic flux23,24. Similar to lysosomes, ER migrates perinuclearly during early autophagy, and subsequently redistributes to the periphery (Fig. 3a, b, Supplementary Video 1). CLIMP63 levels increased significantly during early autophagy, and this increase did not appear to require new protein synthesis or inhibition of lysosomal or proteasomal degradation (Fig. 3c, Extended Data Fig. 10a). CLIMP63 knockout prevented ER movement toward the perinuclear region (Fig. 3b) and suppressed autophagosome–lysosome fusion and autophagic degradation, but not lysosomal activity (Extended Data Fig. 10b–f). Since p180 and KTN1 protein levels remained unchanged (Extended Data Fig. 10a), we examined their binding to microtubules. KTN1–microtubule binding did not change upon nutrient starvation, but p180–microtubule binding increased (Fig. 3d). Consistently, ER and lysosomes in p180-knockout cells remained perinuclear (Fig. 3b), and thus p180-knockout cells showed defects in recovery of mTOR signalling24 after nutrient re-supplementation, but not in autophagic degradation (Extended Data Fig. 10g, h).

Fig. 3: ER distribution changes during autophagy.
figure 3

a, U2OS cells stably expressing mEmerald–Sec61β (green, ER marker) were starved in EBSS for 0, 0.5, or 2 h and labelled with anti-LAMP1 (red, lysosome marker). Scale bar, 10 μm. b, ER and lysosome distribution in wild-type, CLIMP63- or p180-knockout cells starved in EBSS for the indicated times. n = 55, 53, 52, 52, 53, 54, 51, 50, 51, 50, 46 and 44 cells (left to right) for both ER and lysosomes. c, Top, U2OS cells were starved in EBSS for 0–2 h and immunoblotted. Bottom, protein levels relative to α-tubulin. n = 5 experiments. d, Average intensities of PLA for p180 and KTN1 with α-tubulin at 0, 0.5, or 2 h of EBSS starvation or puromycin treatment (2 μg ml−1 for 2 h). n = 273, 254, 188, 180, 143,144, 182 cells (left to right). e, Schematic of p180 domain composition. The ribosome-binding domain includes 41 positively charged decapeptide repeats, which can potentially bind microtubules once ribosomes are dissociated. Amino acid sequences for several repeats are shown, with positively charged residues in red and negatively charged residues in green. Repeat sequences vary slightly but are all positively charged. f, Under normal cellular conditions, p180L ribosome-binding repeats are occupied by ribosomes and cannot bind microtubules. When cells are starved, ribosomes dissociate from ER, and the repeats can then bind microtubules. g, PLA for p180 and RPL3 (ribosome marker) upon starvation or puromycin treatment (2 μg ml−1 for 2 h). n = 230 cells. Data are mean ± s.d. with individual data points shown. One-way ANOVA followed by Dunnett’s multiple comparisons test for (b), (c); two-sided t-test for (d), (g); P values are shown. See Supplementary Information for uncropped western blots.

Source data

Microtubule modification levels were unaffected by starvation (Extended Data Fig. 10i). Notably, the ribosome-binding region of p180L (the major cellular isoform) includes 41 positively charged decapeptide repeats (Fig. 3e). We hypothesized that this region is occupied by ribosomes under normal cellular conditions but then ribosomes dissociate during starvation, exposing these positively charged regions that can then bind microtubules (Fig. 3f). Indeed, starvation significantly decreased p180–ribosome binding (Fig. 3g, Extended Data Fig. 10j); puromycin treatment, which dissociates ribosomes from ER in fed conditions, markedly enhanced p180–microtubule binding (Fig. 3d). In contrast to p180s, which lacks most ribosome-binding decapeptide repeats, p180L overexpression increased ER–microtubule alignment. This alignment was enhanced by starvation or puromycin treatment (Extended Data Fig. 10k–m), further indicating that ribosome-binding repeats of p180L bind microtubules upon ribosome dissociation.


Peripheral ER network morphology is maintained by hydrophobic hairpin domain proteins (reticulons and receptor expression enhancing proteins (REEPs)) that shape the tubules. The polygonal network is generated via atlastin-mediated tethering and fusion of tubules at three-way junctions and distributed via cytoskeletal interactions3,25,26. Much less is known about the dynamic organization of perinuclear ER. Although microtubules have important roles in establishing ER morphology, most studies have emphasized peripheral tubular ER27 or identification of ER proteins that bind microtubules13,14,28. Proteins including CLIMP63, p180 and KTN1 are enriched in dense, sheet-like perinuclear ER, and they each bind microtubules. However, phenotypes of cells deficient in these proteins differ considerably, raising the question of how microtubule-binding specificity is maintained. Here we demonstrate that CLIMP63, p180 and KTN1 preferentially bind different subsets of microtubules to maintain perinuclear ER in its characteristic distribution, explaining the differential effects of their absence. Furthermore, depletion of centrosome or Golgi-derived microtubules has distinct effects on the microtubule binding of these three proteins.

Microtubule diversity can be achieved via different tubulin gene products, differential interactions with microtubule-associated proteins and numerous post-translational modifications2. Modifications are dynamic and rapidly reversible, but evidence for how they affect microtubule-related functions has been limited2. We have shown here that KTN1 preferentially binds perinuclear polyglutamylated microtubules with long glutamate chains, whereas p180 binds glutamylated microtubules with either short or long chains. By contrast, CLIMP63 has a higher threshold for response to microtubule glutamylation. We cannot exclude that increased affinity at higher glutamate numbers for TTLL7-modified microtubules stems from additional chains that TTLL7 initiates on tubulin tails, and not only from introduction of longer chains. Conversely, p180 is more sensitive to any increase in glutamate numbers on the tubulin tail and robustly binds both mono- and polyglutamylated microtubules. This differential effect on microtubule binding according to glutamylation state has previously been observed for the microtubule-severing ATPase spastin29 and may represent a general feature of this modification, enabling fine tuning of molecular interactions. Thus, a small difference in the number of glutamates added to tubulin side chains may exert a substantial qualitative effect on ER distribution. Other ER-localized, microtubule-binding proteins30 are likely to contribute to overall cellular ER positioning. Indeed, even in p180 and KTN1 double-knockout cells, TTLL7 overexpression still disperses ER, suggesting the involvement of other ER proteins. Moreover, tubular ER selectively moves along acetylated microtubules27, further indicating that ER distribution is broadly sensitive to microtubule modifications.

The ability of cells to dynamically control ER distribution through differential microtubule modifications has important functional implications. For instance, p180 regulates microtubule remodelling in axons31, and axonal microtubules are highly glutamylated2. Thus, p180 may affect microtubule remodelling by differentially recognizing glutamylated axonal microtubules. Of note, although dysregulation of ER shaping and microtubule polyglutamylation lead to different neurodegenerative diseases32,33, these diseases share some similar cellular phenotypes, including mitochondrial distribution defects and axon degeneration, suggesting possible convergence.

When ER positioning is disrupted, distributions of other organelles are affected. Microtubules have key roles in organelle distribution1, and their ability to selectively distribute organelles relies on a tubulin code. Our results indicate that ER distribution is mediated via specific membrane-bound proteins with differential binding to different levels and types of microtubule glutamylation, broadly affecting distributions of most other organelles. ER thus interprets the tubulin code to regulate movement and positioning of cellular organelles. Rather than imbuing each organelle with its own sensing and response mechanisms, cells achieve organizational efficiency by using ER as a first-line sensor and responder. This role is exemplified during nutrient starvation, when cells increase CLIMP63 protein levels to move ER towards the perinuclear region, which also clusters lysosomes for efficient autophagic degradation. Then, cells harness enhanced p180–microtubule binding to redistribute ER and lysosomes for a proper reset. There are likely to be other ER proteins that also decipher the tubulin code, with important implications for ER function in health and disease.


Plasmids and reagents

GFP-mCherry-LC3 was a gift from Juan S. Bonifacino. mEmerald-Sec61β, pcDNA3.1_CLIMP63-HA, mApple-SiT (Golgi apparatus), mKO-SKL (peroxisome), YFP-KDEL (ER) and CFP-LAMP1 (lysosome) were constructed as described previously6,7,18,21. pCMV6_p180s-myc-Flag (RC218816), pCMV6_KTN1-myc-Flag (RC219832), pCMV6_TTLL4-myc-Flag (RC205206) and pCMV6_CCP1-myc-Flag (RC220826) were obtained from Origene Technologies. pcDNA3.1_TTLL6-Flag (OHu07095), pcDNA3.1_CCP5-Flag (Ohu28493), pcDNA3.1_CCP6-Flag (OHu24335) and pcDNA3.1_p180L (OHu24745) were obtained from GenScript. Mutants of CLIMP63, p180s and KTN1 were generated in a pcDNA3.1(+) vector with a C-terminal HA-epitope tag. Specifically, Ser-to-Glu mutants (S3E, S17E, S19E) of CLIMP63 were synthesized by GenScript. pcNDA3.1-KTN1, pN1-KTN1-mApple, pC1-sp-mScarlet-KTN1, pN1-KTN1-mNeonGreen, 2×Strep-p180s_29-381-mNeonGreen, 2×Strep-p180s_29-381-mNeonGreen and CLIMP63-mNeonGreen-2×Strep were also constructed using standard cloning procedures. 5′ and 3′ UTR of KTN1 were synthesized by Integrated DNA Technologies and ligated into the pN1-KTN1-mNeonGreen vector. Human TTLL4, TTLL6, TTLL7, CCP1, CCP5 and CCP6 inserts were also cloned into pCMV14-3×Flag and pN1-mApple. DNA oligonucleotides were synthesized by Integrated DNA Technologies.

LD540 was provided by C. Thiele34. Camptothecin (S1288), DC661 (S8808), etoposide (S1225) and MG132 (S2619) were purchased from Selleckchem. LysoSensor Green DND 189 (L7535), G418 (11811098) and puromycin (A1113802) were from Thermo Fisher Scientific. Centrinone B (CNB, CN5690) was from Tocris. GTP (G8877), ATP, Taxol (T7402), tunicamycin (SML1287), Earle’s Balanced Salts (EBSS, E2888) and Duolink In Situ PLA kit (DUO92002, DUO92004, DUO92008, DUO92013) were from Sigma-Aldrich. The Cathepsin L Activity Assay Kit (Fluorometric) (ab65306) was obtained from Abcam.

Cell culture and transfections

All cell lines were obtained from the American Type Culture Collection: HEK 293T (CRL-11268), COS7 (CRL-1651), HeLa (CCL-2), RPE1 (CRL-4000) and U2OS (HTB-96) cells. HEK 293T, COS7 and HeLa cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM, Thermo Fisher Scientific 11995065), RPE1 cells were cultured in DMEM/F12 (1:1) Medium (Thermo Fisher Scientific, 11330-057), and U2OS cells were cultured in McCoy’s 5A medium (Thermo Fisher Scientific16600108); all were supplemented with 10% fetal bovine serum (FBS, Thermo Fisher Scientific 2614007) and 1×penicillin/streptomycin/amphotericin B (Thermo Fisher Scientific 15240112) at 37 °C with 5% CO2. HEK 293T cells were transfected with Avalanche-Everyday Transfection Reagent (EZ Biosystems, EZT-EVDY-1). U2OS and COS7 cells were electroporated using Cell Line Nucleofector Kit V (for U2OS; Lonza, VVCA-1003) and Cell Line Nucleofector Kit R (for COS7; Lonza, VVCA-1001) following the manufacturer’s instructions. Amounts of key plasmids transfected are (per 1 × 106 U2OS cells or 5 × 105 COS7 cells): 0.3 μg CLIMP63-HA, 0.5 μg p180s-HA, 2 μg KTN1-mApple, 1 μg CLIMP63-mEmerald (for overexpression), 2 μg p180s-mEmerald, 4 μg p180L-mEmerald, 4 μg KTN1-mEmerald, as well as (in various vectors) 1 μg TTLL4, 1 μg TTLL6, 0.3 μg TTLL7, 4 μg CCP1, 0.5 μg CCP5, and 1 μg CCP6.

For RNAi knock down of AKAP450, two siRNAs targeting ATATGAACACAGCTTATGA and AACTTTGAAGTTAACTATCAA were synthesized by Eurofins Genomics. Cells were transfected using Avalanche-Omni Transfection Reagent (EZ Biosystems, EZT-OMNI-1) with 20 pmol siRNA for 3 days. For RNAi knockdown of CCP5, ON-TARGETplus siRNA sets targeting human CCP5 were purchased from Horizon Discovery (LQ-009468-00-0005), and 60 pmol siRNAs were transfected per 1 × 106 U2OS cells using Lonza Cell Line Nucleofector Kit V.


Primary antibodies used: mouse monoclonal anti-AKAP450 (BD Biosciences, 611518, Clone 7/AKAP450, immunoblot 1:250), rabbit polyclonal anti-Atlastin2 (Bethyl Laboratories, A303-333A, immunoblot 1:500), rabbit polyclonal anti-Atlastin3 (Proteintech, 16921-1-AP, immunoblot 1:1,000), rabbit monoclonal anti-Catalase (Cell Signaling Technology, 12980, clone D4P7B, immunofluorescence 1:800), mouse monoclonal anti-Climp63 (Enzo, ALX-804-604, clone G1/296, immunofluorescence 1:500 immunoblot 1:5,000), mouse monoclonal anti-Flag M2 (Sigma-Aldrich, F1804, clone M2, immunoblot 1:1,000), rabbit polyclonal anti-GFP (MBL, 598, immunoblot 1:5,000, immunofluorescence 1:500), mouse monoclonal anti-GM130 (BD Biosciences, 610822, Clone 35/GM130, immunofluorescence 1:200), rabbit polyclonal anti-GM130 (Proteintech, 11308-1-AP, immunofluorescence 1:200), mouse monoclonal anti-HA (Covance, MMS-101P, clone 16B12, immunofluorescence 1:500, immunoblot 1:5,000), rabbit polyclonal anti-kinectin (Proteintech, 19841, immunoblot 1:2,000), rabbit monoclonal anti-kinectin (Cell Signaling Technology, 13243, clone D5F7J, immunofluorescence 1:100), mouse monoclonal anti-Lamp1 (DSHB, clone 1D4B, immunofluorescence 1:2,000), rabbit polyclonal anti-LC3 (Cell Signaling Technology, 4108, immunofluorescence 1:200, immunoblot 1:1,000), rabbit polyclonal anti-Lunapark (Sigma-Aldrich, HPA014205, immunoblot 1:250), mouse monoclonal anti-Myc (Santa Cruz, sc-40, clone 9E10, immunoblot 1:2,000), rabbit polyclonal anti-p180 (Thermo Fisher Scientific, PA5-21392, immunofluorescence 1:500, immunoblot 1:5,000), rabbit polyclonal anti-Pericentrin (Abcam, ab4448, immunofluorescence 1:1,000), rabbit polyclonal anti-polyglutamylation (polyE) (AdipoGen, AG-25B-0030, immunofluorescence 1:200, immunoblot 1:1,000), mouse monoclonal anti-glutamylation clone GT335 (AdipoGen, AG-20B-0020, immunofluorescence 1:200, immunoblot 1:200), rabbit polyclonal anti-REEP2 (Proteintech, 15684, immunoblot 1:3,000), rabbit polyclonal anti-REEP3 (Abcam, ab106463, immunoblot 1:1,000), rabbitpolyclonal anti-REEP4 (Proteintech, 26650, immunoblot 1:1,000), rabbit polyclonal anti-REEP5 (Proteintech, 14643, immunoblot 1:1,000), rabbit polyclonal anti-reticulon3 (Proteintech, 12055, immunoblot 1:2,000), rabbit polyclonal anti-reticulon4 (Proteintech, 10740, immunoblot 1:1,000), rabbit polyclonal anti-RPL3 (Proteintech, 66130, immunofluorescence 1:100), rabbit polyclonal anti-TOM20 (Santa Cruz, sc-11415, immunofluorescence 1:1,000), mouse monoclonal anti-TOM20 (BD Biosciences, 612278, Clone 29/Tom20, immunofluorescence 1:1,000), rabbit polyclonal anti-TRAPα (Proteintech, 10583, immunofluorescence 1:50), rat monoclonal anti-α-tubulin Alexa Fluor 647 (Abcam, ab195884, clone YOL1/34, immunofluorescence 1:50), mouse monoclonal anti-α-tubulin (Proteintech, 66031, clone 1E4C11, immunofluorescence 1:1,000, western blot 1:10,000), mouse monoclonal anti-β-tubulin (Proteintech, 66240, clone 1D4A4, immunofluorescence 1:1,000). Alexa Fluor 405/488/568/633 conjugated goat anti-rabbit/mouse IgG (H+L) highly cross-adsorbed secondary antibodies were from Thermo Fisher Scientific. HRP-conjugated goat anti-mouse or anti-rabbit secondary antibodies were from Santa Cruz Biotechnology.

Stable cell lines

To generate U2OS cells stably expressing mEmerald-Sec61β, cells were transfected with the mEmerald-Sec61β7 and selected using 200–1,000 μg μl−1 (gradually increasing) G418 for two weeks; green-positive cells were sorted into mono-clones by flow cytometry using a MoFlo Astrios cell sorter (Beckman Coulter) and cultured in the presence of 200 μg μl−1 G418 for 2–3 weeks. Proliferated clones were verified by immunoblotting and fluorescence imaging.

CRISPR–Cas9 gene editing




Centrosome depletion

To deplete the centrosome, cells were treated with 125 μM CNB for 1 week as described15 before further analysis.

Western blotting

Cells were quickly rinsed with PBS, directly lysed with sample buffer (50 mM Tris, pH 6.8, 1 mM DTT, 10% glycerol, 2% SDS, 0.1% Bromophenol blue), and boiled for 5 min. Proteins were then resolved by SDS–PAGE using Mini-PROTEAN TGX Precast Protein Gels (Bio-Rad Laboratories) and transferred to nitrocellulose membranes using the Trans-Blot Turbo RTA Midi Nitrocellulose Transfer Kit (Bio-Rad Laboratories) following the manufacturer’s instructions. Membranes were blocked with 4% milk in TBST (20 mM Tris, pH 7.4, 150 mM NaCl, 0.1% Tween-20), and incubated with primary antibody (diluted in blocking buffer) at 4 °C overnight. After washing with TBST, membranes were incubated with secondary antibody at room temperature for 2 h, followed by intensive washing with TBST. Immunoreactive proteins were visualized with GE Healthcare LS ECL Prime Western Blotting Detection Reagent (RPN2236) and imaged using a ChemiDoc XRS+ (Bio-Rad). Band intensities were quantified using Fiji software (NIH).

Immunofluorescence and imaging

Cells were fixed with 4% paraformaldehyde in PBS (Lonza) for 30 min at room temperature and permeabilized with 0.1% Triton X-100 in PBS for 10 min. Alternatively, for immunostaining of glutamylation (GT335) and polyglutamylation (polyE), cells were fixed and permeabilized with cold methanol for 5 min at −20 °C. Then, after blocking with 3% BSA for 30 min, cells were immunostained with polyE antibody at 4 °C overnight, then with polyE and GT335 together at 4 °C overnight, followed by secondary antibody staining at room temperature for 1 h, and finally with anti-α-tubulin Alexa Fluor 647 at room temperature for 2.5 h. For staining of lipid droplets with LD540 dye, cells were incubated with 0.1 μg ml−1 LD540 in PBS for 5 min. Cells were mounted using Fluoromount-G (SouthernBiotech) and imaged using a Zeiss LSM880 confocal microscope in Airyscan mode equipped with a 63 × 1.4 NA Plan-Apochromat oil objective (Carl Zeiss). Images were acquired using ZEN software (Carl Zeiss) and processed with ZEN software or Fiji (NIH).

Quantification of ER distribution

Three-dimensional images were acquired using a Zeiss LSM880 confocal microscope in Airyscan mode and reconstructed using ZEN software (Zeiss Microscopy). Summed intensity projections were generated using floating point notation to carry precision. A custom macro in Fiji-ImageJ was used to define the centre of the nucleus and remove the signal of neighbouring cells to avoid perturbing the results. From the manually defined centre, a radius was drawn out past the furthest point on the cell and swept through 360° in 0.1° steps, taking a line profile each time and rescaling the data to correct for artifacts generated by the square shape of the pixels. The resulting data represents an (r,θ)-space representation of the cell’s fluorescence distribution. For analysis referring to ‘normalized’ data, we account for the shape of the cytoplasm by finding the radius at each angle where the nuclear envelope and the edge of the cell are located. The fluorescence data were then rescaled to a normalized axis with the cytoplasm between the nuclear envelope and the cell periphery scaled from 0 to 100%. The nucleoplasm is scaled to stretch between −25 and 0, as a control. (Note that, in this 2D implementation, the nucleoplasm also contains the regions of cytoplasm and nuclear envelope above and below the nucleus).

The MDR and asymmetry of each compartment were calculated using custom Matlab scripts as described in the Supplementary Text. Where true values are given by integrals over space, the value was estimated at the resolution limit of the microscope using a sum over the pixels.

Microtubule co-sedimentation assay

To test the microtubule-binding affinities of CLIMP63, p180 and KTN1, cells were lysed in PIPES buffer (80 mM PIPES, pH 6.8, 1 mM MgCl2, 1 mM EGTA, 100 mM NaCl, 1% Triton X-100, plus Complete protease inhibitors) for 30 min on ice. Cell lysates were centrifugated twice at 20,000g for 20 min at 4 °C. The supernatant was supplemented with 1 mM GTP and 40 μM Taxol and incubated at 4 °C or 37 °C for 30 min for tubulin polymerization before centrifugation at 20,000g for 30 min at 4 °C or 37 °C, respectively. The resulting pellets (P) and supernatants (S) were collected and subjected to immunoblot analysis. In some experiments, only the pellets and supernatants of the 37 °C samples are shown.

Proximity ligation assay

PLA (Sigma-Aldrich, DUO92101) was performed according to the manufacturer’s instructions. Samples were observed under a Zeiss LSM880 confocal microscope with a 20 × 1.0 NA objective using the Airyscan function. The total intensity of the PLA signal per cell was quantified using Fiji software.

Protein purification

Deletion fragments of p180 (short isoform NM_001042576, residues 29-381) and KTN1 (NM_001079521, residues 29–400) as well as full-length CLIMP63 were expressed as fusions with mNeonGreen-2×Strep in HEK 293T cells. 48 h post-transfection, cells were lysed in PBS (Lonza) plus 500 mM NaCl, 1% Triton X-100, and protease inhibitors and then centrifugated at 30,000g at 4 °C for 30 min. Supernatants were combined with Strep-Tactin XT beads (IBA Lifesciences) and rotated gently for 3 h. After extensive washing with lysis buffer (PBS plus 500 mM NaCl and 1% Triton X-100) and then wash buffer (IBA Lifesciences), bound proteins were eluted with Strep-Tactin XT Elution Buffer (IBA Lifesciences). Eluted proteins were subjected to multiple rounds of PBS dilution and concentration using 10 kDa protein concentrators (Sigma-Aldrich), before being aliquoted and frozen in liquid nitrogen.

TIRF-based assays for protein binding to differentially glutamylated microtubules in vitro

Unmodified human tubulin was purified from tsA201 cells as described previously36. TTLL4 and TTLL6 were expressed in Escherichia coli and purified as previously described18. TTLL7 was also expressed in E. coli and purified as previously described19 . Taxol-stabilized microtubules were polymerized out of 98.5% unmodified tubulin and 1.5% biotinylated brain tubulin36,29 (Cytoskeleton T333P). Unmodified microtubules were modified using TTLL4, TTLL7 or TTLL6 at 1:10 molar ratio of enzyme to tubulin at room temperature in 20 mM HEPES (pH 7.0), 50 mM NaCl, 10 mM MgCl2, 1 mM glutamate, 1 mM ATP, 0.5 mM TCEP, and 10 μM Taxol for 4.5 h for TTLL4, between 20 min and 2 h for TTLL7, and between 7.5 and 22 h for TTLL6. Control microtubules were incubated with the enzymes under the same conditions but with aspartate, which is not a substrate for TTLL glutamylases, instead of glutamate. Enzymes were removed through a high-salt wash as previously described29. The extent of glutamylation was determined by liquid chromatography–electrospray mass spectrometry36,29 (LC–MS). The spectra display the characteristic distributions of masses with peaks separated by 129 Da, which corresponds to one glutamate (Extended Data Fig. 6o, p). The extent of tubulin glutamylation on α- or β-tubulin was determined by calculating the weighted average of peak intensities for each tubulin species present29.

For microtubule-binding assays, microtubules were immobilized in chambers made of silanized glass37 using Neutravidin (Thermo Fisher Scientific). Next, a solution containing 60 mM Pipes (pH 6.8), 0.7 mM MgCl2, 0.7 mM EGTA, 50 mM KCl, 10 mM 2-mercaptoethanol, 10 μM Taxol, 1% F127 Pluronic, 1.4 mg/ml casein, 20 mM glucose, glucose oxidase, and catalase was flushed into the chamber, followed by the same solution containing 4.7 nM mNeon-labeled p180, KTN1 or CLIMP63. Images were acquired after allowing for equilibration for 5 min at room temperature using total internal reflection fluorescence (TIRF) microscopy at an exposure of 100 ms for the GFP channel. Unlabelled microtubules were visualized using interference reflection microscopy38. Multiple fields of view were imaged. Background corrected line scan average intensities were measured using Fiji software. Multiple chambers were quantified for each condition.

Real-time PCR

Total mRNA were extracted using TRIzol (Thermo Fisher Scientific 15596018) and Direct-zol RNA Miniprep (Zymo Research, R2052), then reverse-transcribed using the SuperScript IV First-Strand Synthesis System (Thermo Fisher Scientific, 1809105). Real-time PCR primers, designed by a free online tool developed by Integrated DNA technologies, were as follows: CCP5-RT: GACTGCCAGGAACTGCTAAA and AGGAGCTCCCGATGGTAATA; GAPDH-RT: GGTGTGAACCATGAGAAGTATGA and GAGTCCTTCCACGATACCAAAG.

Real-time PCR was performed using Applied Biosystems PowerUp SYBR Green Master Mix (Thermo Fisher Scientific, A25780) with Applied Biosystems QuantStudio 6 Flex real-time PCR instrument. Data were collected and analysed in QuantStudio Real-time PCR Software and Microsoft Excel using the \({2}^{-\varDelta \varDelta {C}_{{\rm{T}}}}\) method.

Multispectral imaging

Multispectral imaging was performed as described previously22. Images were acquired with a Zeiss LSM880 confocal microscope equipped with a 32-channel multi-anode spectral detector (Carl Zeiss) using a 63×/1.4 NA objective lens, at 37 °C and with 5% CO2. Fluorophores were excited simultaneously using 458, 514 and 594 nm lasers and a 458/514/594 nm beam splitter, with images collected onto a linear array of 32 photomultiplier tube elements in λ mode at 9.7 nm bins from 468 to 687 nm. Spectra were defined by imaging singly labelled cells for each of the fluorophore reporters, using the same acquisition and laser settings as for multiply labeled cells. Multispectral images were unmixed using the linear unmixing package in ZEN (Carl Zeiss).

Measurements of autophagosome–lysosome fusion and lysosome activity

For autophagosome–lysosome fusion assessments, U2OS cells were transfected with GFP-mCherry-LC3 for 24 h, treated with EBSS for 2 h before fixation with 4% paraformaldeyde in PBS, and imaged using a Zeiss LSM880 confocal microscope in Airyscan mode equipped with a 63 × 1.4 NA Plan-Apochromat oil objective (Carl Zeiss). A z-projection was performed using maximum projection before quantification. The mCherry-positive vesicles indicate autophagosomes already fused with lysosomes, as the GFP signal would be quenched by the acidic environment of lysosomes; vesicles with both GFP and mCherry fluorescence indicate autophagosomes not yet fused with lysosomes. Quantifications of these two types of vesicles were performed manually using Fiji software.

For lysosome acidification assays, U2OS cells were labeled with 1 μM LysoSensor Green DND 189 for 4 min and immediately imaged within one minute with a Zeiss Axio microscope using a 20×/0.4 NA objective. Images were captured with ZEN software, and total intensities of each cell were quantified in Fiji.

Cathepsin L activity assays were carried out using the Abcam Cathepsin L Activity Assay kit (Fluorometric; ab65306) following the manufacturer’s instructions; 1 × 106 cells were assayed in each sample.

Statistics and reproducibility

No statistical method was used to predetermine sample size. All groups were randomly assigned and every group represents a distincttreatment or condition. Data were not analysed in a double-blinded manner. All comparisons were performed using Graphpad Prism or Microsoft Excel software. Data are expressed as means ± s.d., P values are shown on top of the corresponding columns, as determined by one-way ANOVA followed by Dunnett’s multiple comparisons test, Mann–Whitney test, Kruskal–Wallis test or by unpaired two-sided t-test as indicated in the figure legends. When representative images are shown, at least three repeats were performed except for Extended Data Figs. 1b, 3a–g, 5a, d, 6a, m, n, 8a, e, for which repeats are not necessary because they represent sequential sequence mapping data that build upon one another or else they show representative knockdown or knockout efficiencies that can be further established by the resulting cellular phenotypes.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.