Mitochondrial fission factor (MFF) is a critical regulator of peroxisome maturation

Peroxisomes are highly dynamic subcellular compartments with important functions in lipid and ROS metabolism. Impaired peroxisomal function can lead to severe metabolic disorders with developmental defects and neurological abnormalities. Recently, a new group of disorders has been identified, characterised by defects in the membrane dynamics and division of peroxisomes rather than by loss of metabolic functions. However, the contribution of impaired peroxisome plasticity to the pathophysiology of those disorders is not well understood. Mitochondrial fission factor (MFF) is a key component of both the peroxisomal and mitochondrial division machinery. Patients with MFF deficiency present with developmental and neurological abnormalities. Peroxisomes (and mitochondria) in patient fibroblasts are highly elongated as a result of impaired organelle division. The majority of studies into MFF-deficiency have focused on mitochondrial dysfunction, but the contribution of peroxisomal alterations to the pathophysiology is largely unknown. Here, we show that MFF deficiency does not cause alterations to overall peroxisomal biochemical function. However, loss of MFF results in reduced import-competency of the peroxisomal compartment and leads to the accumulation of pre-peroxisomal membrane structures. We show that peroxisomes in MFF-deficient cells display alterations in peroxisomal redox state and intra-peroxisomal pH. Removal of elongated peroxisomes through induction of autophagic processes is not impaired. A mathematical model describing key processes involved in peroxisome dynamics sheds further light into the physical processes disturbed in MFF-deficient cells. The consequences of our findings for the pathophysiology of MFF-deficiency and related disorders with impaired peroxisome plasticity are discussed.


50
Peroxisomes are highly dynamic membrane-bound organelles with key functions in cellular lipid and 51 ROS metabolism. Defects in peroxisome biogenesis and metabolic function can result in severe 52 disorders with developmental defects and neurological abnormalities (Dorninger et al. 2017;Wanders 53 2018). Peroxisome biogenesis disorders (PBDs) result from mutations in PEX genes, which encode 54 proteins essential for peroxisomal membrane biogenesis and matrix protein import. PBDs, such as 55 Zellweger Spectrum disorders, are usually characterised by a loss of functional peroxisomes. This 56 impacts on multiple metabolic pathways (e.g., peroxisomal αand β-oxidation of fatty acids, and the 57 synthesis of ether-phospholipids, which are abundantly present in myelin sheaths) and results in various 58 patient phenotypes and symptoms (Braverman et al. 2016). Peroxisomal single enzyme deficiencies 59 (PEDs) on the other hand are caused by mutations in genes encoding a specific peroxisomal 60 enzyme/protein and usually affect one metabolic pathway or function. The most prominent example is 61 X-linked adrenoleukodystrophy, which is caused by mutations in the ABCD1 gene, encoding a 62 peroxisomal ABC transporter required for the import of very-long-chain fatty acids (VLCFAs) into the 63 organelle (Raymond et al. 1993). In addition to PBDs and PEDs, a third group of disorders has been 64 identified, which is characterised by defects in the membrane dynamics and division of peroxisomes 65 rather than by loss of metabolic functions (Waterham et  The tail-anchored membrane proteins MFF and FIS1 act as adaptor proteins for the recruitment of DRP1 76 to the peroxisomal membrane and interact with PEX11β (Schrader et al. 2016). With the exception of 77 PEX11β, all proteins involved in peroxisome growth and division identified so far are also key 78 mitochondrial division factors. FIS1 and MFF are dually targeted to both peroxisomes and mitochondria, 79 and also recruit DRP1 to the mitochondrial outer membrane (Koch et al. 2005; Gandre-Babbe and van 80 der Bliek 2008; Costello et al. 2017aCostello et al. , 2018. Mitochondria also possess the adaptor proteins MiD49 and 81 MiD51, which are specific to mitochondria and can recruit DRP1 independent of FIS1 and MFF (Palmer 82 et al. 2013). GDAP1 is another tail-anchored membrane protein shared by mitochondria and 83 peroxisomes, which influences organelle fission in an MFF-and DRP1-dependent manner in neurons 84 (Huber et al. 2013). Recently, also MIRO1, a tail-anchored membrane adaptor for the microtubule-85 dependent motor protein kinesin, has been shown to localise to mitochondria and peroxisomes and to 86 contribute to peroxisomal motility and membrane dynamics (Castro et

Fibroblast Cell Culture and Transfection 123
For routine culture and morphological experiments, MFF-deficient patient skin fibroblasts and controls 124 (Shamseldin et  For photobleaching experiments, data are presented as the mean grey value for each increment. Only 215 peroxisomes which did not overlap with other peroxisomes were analysed. 216

Measurement of Subcellular Redox Dynamics 231
The procedures involved in the measurement of subcellular redox levels have been previously described 232 in detail (Lismont et al. 2017). In brief, SV40 large T antigen-transformed human fibroblasts (HUFs-T) 233 were transfected with plasmids coding for GSH/GSSG-(roGFP2) or H2O2-sensitive (roGFP2-ORP1) 234 reporter proteins targeted to various subcellular compartments [cytosol (c-), mitochondria (mt-), or 235 peroxisomes (po-)]. One day later, the cells were incubated for 30-60 minutes in phenol red-free culture 236 medium and imaging was performed to visualize both the oxidized (excitation 400 nm, emission 515 237 nm) and reduced (excitation 480 nm, emission 515 nm) states of roGFP2. During image acquisition, 238 the cells were maintained in a temperature-, humidity-, and CO2-controlled incubation chamber. For 239 cytosolic measurements, the ROI was selected outside the nucleus. The Cell^M/xcellence software 240 module (Olympus) was used to quantify the relative fluorescence intensities of roGFP2 at 400 and 480 241 nm excitation, giving a ratiometric response. 242

Measurement of Peroxisomal pH using pHRed 243
Peroxisomal pH was measured as previously described (Godinho and Schrader 2017). Briefly, MFF-244 deficient and control fibroblasts were transfected with plasmids coding for a cytosolic or peroxisomal 245 pH-sensitive red fluorescent protein (pHRed-Cyto and pHRed-PO, respectively) (Godinho and 246 Schrader 2017). Twenty four hours after transfection, cells were imaged using excitation wavelengths 247 of 458 and 561 nm. Prior to image acquisition, a controlled temperature chamber was set-up on the 248 microscope stage at 37°C, as well as an objective warmer. During image acquisition, cells were kept at 249 37°C and in a HEPES-buffered CO2-independent medium. For calibration, the cells were incubated in 250 solutions of known pH (containing 5 µM nigericin) in a confocal stage chamber. ImageJ (Schneider et 251 al. 2012) was used to calculate the 561/458 ratiometric response. 252

Statistical Analysis 253
Unless indicated otherwise, a two-tailed, unpaired t-test was used to determine statistical differences 254 against the indicated group (*, P < 0.05; **, P < 0.01; ***, P < 0.001). Boxplots are presented with the 255 bottom and top of each box representing the 25th and 75th percentile values, respectively; the horizontal 256 line inside each box representing the median; and the horizontal lines below and above each box 257 denoting the range. In the roGFP (Fig. 4B) and roGFP-ORP (Fig. 4D) box plots, these lines denote the 258 standard deviation. Bar graphs are presented as mean ± SEM. In-text data are presented as mean ± SD. 259 Analysis was performed from at least three independent experiments. 260 peroxisomes showed a punctate staining pattern typical for human fibroblasts (Fig. 1A). Mitochondria 272 in patient cells were also reported to be elongated ( 2017b). We transfected MFF-deficient fibroblasts with Myc-MFF using microporation, which allowed 282 us to monitor peroxisome morphology at early time points (2-3 hours) after transfection and therefore 283 capture the initial stages of MFF-mediated peroxisome division (Suppl. Fig. S1). Cells were processed 284 for immunofluorescence using antibodies against the Myc-tag and PEX14. Twothree hours after 285 transfection, MFF was observed to localise in spots on elongated peroxisomes (and elongated 286 mitochondria) supporting a role in the assembly of the division machinery and the formation of division 287 sites. Many MFF-expressing cells already contained short, dividing peroxisomes or fully divided, 288 spherical peroxisomes (Suppl. Fig. S1). 289 290 Suppl. Figure S1. Re electron microscopy ( Fig. 1B). In contrast to immunofluorescence, constrictions of elongated 308 peroxisomes were not observed in ultrastructural studies (Fig. 1B). Interestingly, EM revealed the 309 presence of spherical peroxisome bodies, with a single, smaller tubule protruding from the body (Fig.  310 1B). We assume that the "constricted" appearance of peroxisomes in immunofluorescence is likely due 311 to instability of the extremely long, delicate membrane structures during fixation with para-312 formaldehyde, highlighting the importance of ultrastructural studies to validate light microscopy 313 observations. Ultrastructural studies (Fig. 1B) and immunofluorescence microscopy ( Fig. 1C) show 314 that the peroxisomal membrane tubules are frequently aligned along microtubules, which may 315 contribute to tubule stability and maintenance. 316 Measurement of peroxisomes in EM micrographs revealed that peroxisome bodies are significantly 317 larger than peroxisomal tubules (mean width, body: 141 ± 37 nm, tubule: 81 ± 22 nm) (Fig. 1D). The 318 measured body width is consistent with that of spherical peroxisomes in human fibroblasts from healthy 319 individuals typically being reported to be between 50-200 nm in width (Arias et al. 1985;Galiani et al. 320 2016). Peroxisome length was also quantified based on immunofluorescence data, with a wide range of 321 lengths being present, from smaller, rod shaped peroxisomes (> 3 µm) up to very highly elongated 322 tubules (> 30 µm) (mean length, 8.73 ± 9.2 µm) (Fig. 1E). As expected with a defect in division, the 323 peroxisome number was reduced in MFF-deficient fibroblasts in contrast to controls (mean number, 324 control fibroblasts: 244 ± 116, dMFF: 34 ± 25) (Fig. 1F). Overall, we reveal that peroxisomes in MFF-325 deficient patient fibroblasts are fewer and consist of two continuous membrane domains: a spherical 326 peroxisome body with typical peroxisome size, and a thin, highly elongated tubular structure protruding 327 from this body. 328  were slightly higher than the reference range, but this does not indicate any dysfunction. The activity 354 of dihydroxyacetone phosphate acyltransferase (DHAPAT), the first enzyme of the plasmalogen 355 biosynthesis pathway located in peroxisomes, was within reference range. The intra-peroxisomal 356 processing of the peroxisomal β-oxidation enzymes acyl-CoA oxidase 1 (ACOX1) and 3-ketoacyl-CoA 357 thiolase was not altered, suggesting normal peroxisomal matrix protein import and processing activity 358 in contrast to fibroblasts from a patient with a peroxisomal biogenesis disorder (Fig. 2)  of ACOX1 is non-specific. 384

Protein import into MFF-deficient peroxisomes is impaired in tubular extensions 385
As globular peroxisomal bodies were visible in ultrastructural studies (Fig. 1B) but surprisingly less 386 visible in immunofluorescence studies with anti-PEX14, which labelled predominantly tubular 387 structures (Fig. 1A), we performed co-localisation studies with anti-catalase, a prominent peroxisomal 388 marker enzyme in the peroxisomal matrix (Fig. 3A). In contrast to PEX14, endogenous catalase was 389 found to localise primarily to the spherical peroxisome bodies, with weaker fluorescence intensity along 390 the peroxisomal tubules (Fig. 3A). Analysis of fluorescence intensity along single peroxisomes of both 391 PEX14 and catalase confirmed PEX14 fluorescence primarily along tubules with some localisation in 392 bodies, whereas catalase fluorescence was primarily detected in the peroxisomal body, with reduced 393 intensity along the tubule (Fig. 3A). Peroxisomes import matrix proteins from the cytosol via dedicated 394 import machinery at the peroxisomal membrane (Francisco et al. 2017). Matrix proteins such as catalase 395 are imported into peroxisomes via a C-terminal peroxisomal targeting signal (PTS1). These steady-state 396 observations imply that catalase is mainly imported into the spherical bodies, suggesting that those 397 represent mature, import-competent structures. To test this hypothesis, we expressed a GFP-fusion 398 protein with a C-terminal PTS1 signal SKL (GFP-SKL) in MFF-deficient cells. Cells were processed 399 for immunofluorescence after 24 hours and labelled with anti-PEX14 antibodies (Fig. 3B). Similar to 400 endogenous catalase, exogenously expressed GFP-SKL localised primarily to peroxisomal bodies, with 401 less presence in the peroxisomal tubules (Fig. 3B). This was confirmed by analysis of fluorescence 402 intensity (Fig. 3B). Immunofluorescence microscopy with the peroxisomal membrane markers PMP70 403 and PEX14 revealed co-localisation of both membrane proteins at membrane tubules (Fig. 3C). PMP70 404 also localised to the spherical bodies, where PEX14 is less prominent (Fig. 3C). These findings indicate 405 that the spherical bodies represent mature, import-competent peroxisomes, whereas the tubular 406 extensions comprise a pre-peroxisomal membrane compartment which has not yet fully acquired import 407 competence for matrix proteins or lacks the capability to retain them. To confirm these conclusions, we 408 performed FRAP experiments (Suppl. Fig. S2). Peroxisomes in MFF-deficient fibroblasts expressing 409 GFP-SKL were photobleached followed by immediate observation through live-cell imaging. After 410 photobleaching of the entire organelle (peroxisome body and short tubule), recovery of GFP-SKL 411 fluorescence was first observed in the peroxisome body, indicating that recovery is due to import of 412 GFP-SKL into the peroxisome body rather than into the tubule (Suppl. Fig. S2). We cannot completely 413 exclude that there is some matrix protein import into the tubule, which may be slow or less efficient. 414 However, our findings support our conclusion that spherical bodies are mature import competent 415 structures, whereas the tubules represent pre-peroxisomal membrane structures which have not yet fully 416 acquired import competence for matrix proteins or lack the capability to retain them. 417 hours prior to processing for immunofluorescence microscopy using antibodies against α-tubulin and 438 PEX14. Scale bars, 10 μm, magnification, 2 µm. 439 Suppl. Figure S2. The peroxisomal body is import-competent. MFF-deficient fibroblasts were 441 transfected with GFP-SKL and grown on 3.5-cm glass bottom dishes. Photo-bleaching experiments 442 were performed after 24-48 hours using a Visitron 2D FRAP system. The entire organelle (peroxisome 443 body and short tubule) was photo-bleached (0 min) and recovery of GFP-SKL fluorescence monitored 444 over a period of 10 minutes (A). Note that GFP-SKL fluorescence was observed in the peroxisome body 445 (arrowheads), but not in the peroxisome tubule, indicating that recovery is due to import of GFP-SKL 446 into the peroxisome body. (B) Quantification of fluorescence intensity. Data are presented at the mean 447 grey value for each increment along the length of the peroxisome. Scale bar, 5 μm. 448

A role of PEX14 in maintaining peroxisomal tubule stability 450
As PEX14 is part of the matrix protein import machinery ( ultrastructural and confocal studies microtubules were frequently observed in close association with the 458 entire length of peroxisomal tubules in MFF patient cells (Fig. 1B, C). Furthermore, in a previous study, 459 we showed that highly elongated peroxisomal tubules in fibroblasts are associated with microtubules, 460 and that tubule elongation is reduced in PEX14-deficient cells (Castro et al. 2018). Based on these 461 observations, we hypothesised that PEX14 may be required for the stabilisation of highly elongated 462 peroxisomal tubules. To test this, we depleted PEX14 by siRNA-mediated knock down in MFF-463 deficient cells (Fig. 3D, F, G). Peroxisomal tubules in these cells are typically stretched out in the cell, 464 allowing for easy visualisation. However, when PEX14 was knocked down, peroxisomes lost their 465 tubular morphology and appeared clustered or fragmented (Fig. 3D) (cells with clustered/fragmented 466 morphology: control siRNA: 4.7 ± 1.2%, PEX14 siRNA: 95.3 ± 3.1%) (Fig. 3G). The peculiar 467 peroxisome morphology was specific for silencing of PEX14, and was not observed after silencing of 468 PEX5, excluding an effect of impaired peroxisomal import (Fig. 3E). Furthermore, peroxisome 469 morphology was not altered after silencing of PEX11β or ACBD5 in MFF-deficient cells (Costello et  470 al. 2017b). Clustering and fragmentation of elongated peroxisomes in MFF-deficient cells was also 471 observed after depolymerisation of microtubules with nocodazole (Fig. 3H). These observations 472 suggest a role for PEX14 in facilitating and stabilising peroxisomal membrane extensions by linking 473 the peroxisomal membrane to microtubules. This may explain why PEX14 is predominantly localising 474 to the highly elongated peroxisomal membranes in MFF patient cells. 475

Peroxisomal redox state and pH levels are altered in MFF-deficient fibroblasts 476
The metabolic parameters of peroxisomes in MFF-deficient cells were normal, in particular their 477 different functions in lipid metabolism ( Table 1). As peroxisomes play a role in cellular H2O2 478 metabolism and redox homeostasis, we also investigated these parameters (Fig. 4). Firstly, we assessed 479 the glutathione disulphide (GSSG) to glutathione (GSH) ratio, a marker of oxidative balance. Therefore, 480 MFF-deficient SV40 large T antigen-transformed human fibroblasts (HUFs-T) were transfected with a 481 plasmid encoding cytosolic, mitochondrial or peroxisome-targeted roGFP2 (Fig. 4A) Analyses of the 400/480 ratiometric responses of peroxisome-targeted roGFP2 revealed that the intra-485 peroxisomal pool of glutathione is less oxidized in the MFF-deficient fibroblasts than in the control 486 cells (Fig. 4B). In contrast, no alterations in the glutathione redox state could be detected in the cytosol 487 or the mitochondrial matrix. 488 To monitor changes in hydrogen peroxide homeostasis, MFF-deficient HUFs-T and controls were 489 transfected with plasmids coding for cytosolic, mitochondrial, or peroxisome-targeted roGFP2-ORP1, 490 a H2O2-responsive variant of roGFP2 (Fig. 4C) (Lismont et al. 2019b). No changes in oxidation state 491 were observed in the cytosol and mitochondria (Fig. 4D). However, for peroxisomes, a decreased 492 400/480 nm ratiometric response was seen (Fig. 4D) probe successfully targets to peroxisomes in control and MFF-deficient fibroblasts (Fig. 4E). It mainly 498 distributes to the import-competent spherical peroxisomal bodies, but also to the membrane tubules 499 (Fig. 4E). Following calibration of the pHRed probe (Fig. 4F), the intra-peroxisomal pH can be 500 calculated based on the 458/561 nm ratiometric response. Interestingly, intra-peroxisomal pH in MFF-501 deficient fibroblasts was found to be more alkaline than in control fibroblasts (Fig. 4G) (mean 502 peroxisomal pH, control: 7.24 ± 0.30, patient fibroblasts: 8.00 ± 0.29). 503 Overall, these findings point towards alterations in the peroxisomal redox environment. Specifically, 504 we observed a decrease in the GSSG/GSH ratio and H2O2 levels in MFF-deficient fibroblasts. In 505 addition, we have shown that absence of MFF results in a more alkaline intra-peroxisomal pH. This 506 suggests that MFF-deficiency may compromise normal peroxisomal redox regulation. 507 Quantification of peroxisomal pH in control (C109) and MFF Q64* cells, converting the ratiometric 528 response to pH using the calibration curve (n =20). Scale bars, 10 µm; magnifications, 2 µm. Data are 529 from at least 2-3 independent experiments. *, p < 0.05; ***, p < 0.001; two-tailed, unpaired t test. 530 531 3.6. Highly elongated peroxisomes in MFF-deficient fibroblasts can be degraded by autophagic 532 processes 533 Autophagic processes are important for the maintenance of cellular homeostasis and the integrity of 534 organelles (Anding and Baehrecke 2017). Peroxisome homeostasis is achieved via a tightly regulated 535 interplay between peroxisome biogenesis and degradation via selective autophagy (pexophagy) 536 (Eberhart and Kovacs 2018). It is still unclear if highly elongated peroxisomes are spared from 537 pexophagy, e.g. due to physical limitations, as the elongated peroxisomes may not fit into the 538 autophagosome. Such a scenario would prevent degradation of peroxisomes and could have 539 pathophysiological consequences. 540 To examine if highly elongated peroxisomes in MFF-deficient fibroblasts can be degraded by 541 autophagic processes, we first induced pexophagy by the expression of a fragment of peroxisomal 542 biogenesis protein PEX3. Expression of the first 44 amino acids of the peroxin PEX3, which can insert 543 into the peroxisome membrane, was observed to cause complete removal of peroxisomes (Soukupova 544 et al. 1999). When expressing HsPEX3(1-44)-EGFP in control fibroblasts (Fig. 5A, B), peroxisomes 545 were greatly reduced in number, with many GFP expressing cells showing almost complete loss of 546 PEX14 labelling (Fig. 5A, C109). As reported earlier, loss of peroxisomes resulted in mistargeting of 547 HsPEX3(1-44)-EGFP to the mitochondria (Soukupova et al. 1999) (Suppl. Fig. S3). Interestingly, in 548 MFF-deficient fibroblasts, expression of HsPEX3(1-44)-EGFP also caused a marked reduction of 549 peroxisomes (Fig. 5A, middle panel, B) or complete loss of PEX14 labelling (Fig. 5A, lower panel,  550  B). Increased mitochondrial mistargeting of HsPEX3(1-44)-EGFP was observed with increased loss of 551 peroxisomes ( Fig. 5A; Suppl. Fig. S3). 552 553 554 555 Suppl. Figure S3. HsPEX3(1-44)-EGFP is targeted to mitochondria when peroxisomes are lost. Human 556 control (C109) or MFF-deficient (MFF Q64* ) fibroblasts were transfected with a plasmid coding for 557 HsPEX3(1-44)-EGFP to induce peroxisome degradation and processed for immunofluorescence after 558 24 and 48 hours using antibody against mitochondrial ATP synthase (ATPB). Note the mistargeting of 559 HsPEX3(1-44)-EGFP to mitochondria (arrowheads). Furthermore, mitochondrial morphology is 560 altered including fragmentation and clustering. Scale bars, 10 µm, magnification, 2 µm. 561 562 To examine peroxisome degradation under more physiological conditions, we applied nutrient 563 deprivation. Limiting amino acids in the growth media of cells has been previously shown to induce 564 removal of peroxisomes (Sargent et al. 2016). For assessing peroxisome degradation, controls and MFF-565 deficient fibroblasts were cultured in Hanks' Balanced Salt Solution (HBSS), which lacks amino acids. 566 After 0, 24 and 48 hours, cells were processed for immunofluorescence using anti-PEX14 as a 567 peroxisomal marker (Fig. 5C). In control cells, we observed a marked decrease in spherical peroxisomes, 568 with often only a few organelles remaining after 48 hours in HBSS (Fig. 5C, D). As with HsPEX3(1-569 44)-EGFP, we also observed a decrease in peroxisomes in nutrient-deprived MFF-deficient cells, which 570 was accompanied by a significant reduction in peroxisomal length (mean peroxisomal length, 0 hours 571 HBSS: 6.08 ± 4.90 µm, 48 hours HBSS: 1.55 ± 1.43 µm) (Fig. 5C, E). The reduction in peroxisomes 572 was accompanied by a reduction in peroxisomal marker proteins (Fig. 5F). Peroxisomes and protein 573 levels recovered in control and MFF-deficient cells after switching to complete culture medium for 24 574 hours (Fig. 5C-F). Interestingly, the switch to complete growth medium resulted in the recovery of 575 elongated peroxisomes (mean peroxisomal length, 24 hours recovery: 3.84 ± 3.40 µm) (Fig. 5E), 576 indicating that peroxisomes in MFF-deficient fibroblasts are still dynamic under certain conditions. 577 Overall, these data show that highly elongated peroxisomes in MFF-deficient cells are not spared from 578 autophagic processes and are capable of being degraded. 579 peroxisomes, which are, however, largely functional and otherwise normal. We now reveal in MFF-638 deficient cells that this is not the case. We show that the elongated peroxisomes in those cells are 639 composed of a spherical body, which represents a mature, import-competent peroxisome, and of thin, 640 tubular extensions, which likely represent pre-peroxisomal membrane compartments; not yet fully 641 import-competent for peroxisomal matrix proteins. An alternative interpretation may be that the tubular 642 structures are to some degree import-competent but lack mechanisms to retain the imported matrix 643 proteins. Such a mechanism for retaining matrix proteins may be provided by membrane constriction, 644 which is impaired in MFF-deficient cells. 645 These observations are consistent with the proposed multi-step maturation model of peroxisomal 646 growth and division and with previous data on tubular membrane extensions after expression of PEX11β (Delille et al. 2010;Schrader et al. 2012Schrader et al. , 2016. In this respect, elongated peroxisomes in MFF-648 deficient cells resemble those observed after expression of a division-incompetent PEX11β, which also 649 results in elongated peroxisomes with an import-competent spherical body and a pre-peroxisomal 650 membrane expansion (Delille et al. 2010). In contrast, elongated peroxisomes in DRP1-depleted cells  651 are constricted, with a "beads-on-a string" like appearance, and the interconnected spherical 652 peroxisomes ("beads") are import-competent for matrix proteins (Koch et al. 2004). These constrictions 653 may therefore provide a mechanism to retain matrix proteins. This indicates that a defect in MFF 654 influences peroxisome division earlier than a defect in DRP1, and results in a maturation defect of 655 elongated peroxisomes, which are unable to constrict and to subsequently import and/or retain matrix 656 proteins. Re-expression of MFF in the MFF-deficient fibroblasts early on results in a spot-like 657 localization of MFF on elongated peroxisomes indicating a role for MFF in the assembly of the division 658 machinery. In line with this, it has recently been shown that MFF can act as a sensor but also potentially 659 as an inducer of mitochondrial constriction (Helle et al. 2017). We propose that MFF deficiency, which 660 impairs peroxisomal membrane constriction and proper assembly of the division machinery, blocks 661 further maturation of the pre-peroxisomal membrane compartment. 662 This means that, although the number of fully functional peroxisomes is reduced and matrix proteins 663 are largely restricted to the mature spherical bodies, membrane surface area and volume of the 664 peroxisomal compartment are increased in MFF-deficient cells (mean estimated total surface area, 665 control fibroblasts: 1.55x10 7 ± 7.29x10 6 nm 2 , dMFF: 1.15x10 8 ± 6.57x10 8 nm 2 ; mean estimated total 666 volume, control fibroblasts: 4.1x10 8 ± 1.94x10 8 nm 3 , dMFF 2.5x10 9 ± 1.45x10 9 nm 3 ) (Suppl. Fig. S4), 667 as well as the surface area to volume ratio (mean estimated SA:V, control fibroblasts: 0.038 ± 0.001, 668 dMFF: 0.046 ± 0.005) (Suppl. Fig. S4). This likely explains why biochemical functions of elongated 669 peroxisomes are overall normal under standard conditions. However, it can be speculated that sudden 670 environmental changes (e.g. an increase in peroxisomal substrates via nutrients/diet or stress conditions), 671 which require increased peroxisomal metabolic activity and number, will overwhelm the capacity of in patient cells may be less able to cope with increased expression of peroxisomal matrix enzymes or 676 PMPs. Those may accumulate in the cytoplasm and may be degraded or mistargeted (e.g. to 677 mitochondria) due to the reduced number of import-competent peroxisomes (Ebberink et al. 2012). 678 Suppl. Figure S4. Calculations of peroxisomal surface area, volume, and surface area to volume ratio. may anchor them to microtubules in order to stabilise those highly elongated, delicate membrane 700 structures and to facilitate membrane extension. The membrane topology of PEX14 is poorly defined, 701 but a recent study suggested that the N-terminal domain is protease-protected and may not be exposed 702 to the cytosol (Barros-Barbosa et al. 2019). Such a topology may be inconsistent with tubulin-binding, 703 but it is possible that different populations or complexes of PEX14 exist which may fulfil different 704 functions at the peroxisomal membrane. 705 Peroxisomes are oxidative organelles with important roles in cellular redox homeostasis (Fransen and  706 Lismont 2018). Alterations in their redox metabolism have been suggested to contribute to aging and the development of chronic diseases such as neurodegeneration, diabetes, and cancer (Fransen and  708 Lismont 2019). Using genetically encoded fluorescent sensors with ratiometric readout in live-cell 709 approaches, we revealed alterations in the glutathione redox potential within peroxisomes of MFF-710 deficient fibroblasts, which was less oxidising compared to controls. In addition, we detected reduced 711 levels of peroxisomal H2O2 in these cells. Given that the peroxisomal parameters (Table 1) and catalase 712 levels (Fig. 5F) are similar in control and MFF-deficient human fibroblasts, the possible mechanisms 713 underlying these observations remain a subject of speculation. In this context, it is interesting to note 714 that in a previous study in which mouse embryonic fibroblasts were cultured in medium containing 715 1,10-phenanthroline, a Zn 2+ -chelating compound that induces oxidative stress and disrupts peroxisomal 716 and mitochondrial function (Coyle et al. 2004;Jo et al. 2015), the intra-peroxisomal redox state in 717 tubular peroxisomal compartments was observed to be slightly lower than in spherical bodies (Lismont 718 et al. 2017). Given that (i) peroxisome-derived H2O2 can easily cross the peroxisomal membrane 719 (Lismont et al. 2019a), and (ii) the surface to volume ratio is larger in the tubular structures, this may 720 be explained by the fact that H2O2 can diffuse faster out of the tubular structures than out of the spherical 721 bodies. Alternatively, as this study indicates that matrix proteins are predominantly imported into the 722 spherical bodies and less into the peroxisomal tubules ( Fig. 3; Suppl. Fig. S2), the lower values for 723 peroxisomal redox parameters in the tubular structures may also be due to the fact that these structures 724 contain less H2O2-producing oxidases. However, in contrast to what was observed before in cells 725 cultured in the presence of 1,10-phenanthroline, no significant differences in the glutathione redox state 726 or H2O2 levels could be detected between the spherical and tubular structures in MFF-deficient cells 727 (data not shown). Importantly, the glutathione redox balance and hydrogen peroxide levels in the 728 cytosol and mitochondria were similar to controls, indicating peroxisome-specific alterations due to 729 loss of MFF-function. Peroxisome-derived H2O2 may be an important signalling messenger that 730 controls cellular processes by modulating protein activity through cysteine oxidation (Fransen and  731 Lismont 2019). However, the precise interrelationship between peroxisomal redox metabolism, cell 732 signalling, and human disease remains to be elucidated. Further insight may come from the 733 identification of primary targets for peroxisome-derived H2O2. We also revealed changes in the 734 peroxisomal pH in MFF-deficient fibroblasts, which was more alkaline than in controls. . This strongly indicates that peroxisome morphology and division is affected in a cell type-782 specific manner. 783 We recently developed a mathematical model to explain and predict alterations in peroxisome 784 morphology and dynamics in health and disease conditions (Castro et al. 2018). In this stochastic, 785 population-based modelling approach, each individual peroxisome consists of a spherical body with an 786 optional cylindrical elongation. Peroxisome shape (i.e. the body radius and elongation length) are 787 determined by (i) membrane lipid flow into the body (e.g., from the ER) (governed by rate α and lipid 788 flow constant γ), (ii) elongation growth (governed by speed v and minimum body radius rmin) and (iii) 789 peroxisome division with a rate proportional to the elongation length (governed by rate β and minimum 790 length Lmin). Peroxisome turnover is controlled by the peroxisome mean lifetime τ. We recently 791 demonstrated that this model is applicable to a range of experimental and disease conditions, e.g. loss 792 of PEX5 in Zellweger spectrum disorders (Castro et al. 2018). With wild-type parameters, peroxisomes 793 in the model are typically high in number, with only a low percentage showing elongations, all of which 794 are short (Fig. 6A). The morphological alterations of peroxisomes in MFF-deficient fibroblasts that we 795 have observed experimentally are captured by changing only one parameter, namely by reducing the 796 division rate β to almost zero (Fig. 6B). As the membrane lipid flow rate and elongation growth speed 797 remain unchanged, this results in reduced numbers of peroxisomes with significantly longer membrane 798 elongations (Fig. 6D, E). The observation that control fibroblasts display large numbers of small, 799 spherical peroxisomes, but turn into few, extremely elongated organelles upon blocking of peroxisomal 800 division, indicates that membrane lipid flow rate, elongation growth speed and division rate must be 801 high in fibroblasts under normal conditions. In contrast, low membrane lipid flow rate or elongation 802 speed in other cell types may result in a population of small peroxisomes and reduced numbers. This is 803 reflected by depletion of ACBD5, which impacts on peroxisome-ER tethering and membrane expansion, 804 resulting in shorter peroxisomes in MFF-deficient cells (Costello et al. 2017b). This morphological 805 change can also be captured in the model by reducing the lipid flow rate α in addition to the division 806 rate β (Fig. 6C-E). It is thus likely that peroxisome morphology is differently affected in various cell 807 types in MFF-deficient patients. It should also be considered that environmental changes and related 808 signalling events that trigger peroxisomal membrane expansion and division (e.g. metabolic alterations 809 and with reduced lipid flow to simulate silencing of ACBD5 (siA5) at t = 300 hours (α = 5 nm 2 /s, 819 β = 2 × 10 −15 /nm/s, v = 0.3 nm/s, τ = 4 × 10 5 s, γ = 2.5 × 10 −7 /nm 2 ). (D) Average peroxisome number at t 820 = 300 hours of simulations shown in A-C, represented as percentages relative to WT (n = 100). (E) 821 Average non-zero peroxisome elongation length at t = 300 hours of simulations shown in A-C, 822 represented as percentages relative to WT (n = 100). Scale bars, 1 µm. 823