JIP3 links lysosome transport to regulation of multiple components of the axonal cytoskeleton

Lysosome axonal transport is important for the clearance of cargoes sequestered by the endocytic and autophagic pathways. Building on observations that mutations in the JIP3 (MAPK8IP3) gene result in lysosome-filled axonal swellings, we analyzed the impact of JIP3 depletion on the cytoskeleton of human neurons. Dynamic focal lysosome accumulations were accompanied by disruption of the axonal periodic scaffold (spectrin, F-actin and myosin II) throughout each affected axon. Additionally, axonal microtubule organization was locally disrupted at each lysosome-filled swelling. This local axonal microtubule disorganization was accompanied by accumulations of both F-actin and myosin II. These results indicate that transport of axonal lysosomes is functionally interconnected with mechanisms that control the organization and maintenance of the axonal cytoskeleton. They have potential relevance to human neurological disease arising from JIP3 mutations as well as for neurodegenerative diseases associated with the focal accumulations of lysosomes within axonal swellings such as Alzheimer’s disease.

N eurons face major demands arising from their extreme size, polarity and longevity. Axons in particular stand out due to their length, which requires both long-range transport for delivery of cargoes to and from distant locations combined with mechanisms to ensure structural integrity [1][2][3] . These challenges create unique vulnerabilities that are reflected in the numerous neurodevelopmental and neurodegenerative diseases arising from defects in axonal transport and maintenance [4][5][6] . The unique morphology and functions of axons requires specialized organization of multiple cytoskeletal components. Axonal microtubules, which provide the tracks on which motors can transport organelles and other cargoes over long distances, are polarized with their plus ends towards the distal axon and regulated by the binding of various microtubule binding proteins [7][8][9][10] . The membrane associated periodic actinspectrin lattice provides structural support to ensure axon integrity and non-muscle myosin II-dependent contractility coordinates the passage of organelles through the narrow confines of axons [11][12][13][14][15] .
The vast majority of axonal proteins are synthesized in the neuronal cell body and proximal dendritic regions, and are subsequently transported into axons to meet their structural, signaling, and metabolic demands and to support synaptic transmission 16 . Conversely, efficient retrograde transport from the axon periphery back to the cell body is required for the clearance of old or damaged proteins, as well as of material taken up by endocytosis, via endocytic and autophagic pathways 17,18 . This transport is primarily mediated by immature lysosomes and auto-lysosomes, which have a low content of lysosomal hydrolases and whose fate is to mature into fully degradative lysosomes in cell bodies by fusing with hydrolases-enriched vesicles delivered from the trans-Golgi network 18 . The massive accumulation of these organelles at the distal side of focal blocks of axonal transport reveals that such lysosomes are the major retrograde axonal cargo 19 .
A similar build-up of immature lysosomes is observed in axon swellings surrounding amyloid Aβ deposits at Alzheimer's disease amyloid plaques (both in human patients and in mouse models of the disease), which are putative sites of APP processing [20][21][22] . A link between accumulation of axonal lysosomes due to their impaired transport and amyloidogenic APP processing was further supported by studies of neurons and mice with loss of JIP3 function 23 . JIP3 is a motor interacting protein that is preferentially expressed in neurons and is thought to couple cargos such as lysosomes to dynein, the microtubule minus-end directed motor 24 . JIP3 loss-of-function mutations in multiple animal species result in the build-up of lysosomes within axons 23,25,26 . For example, primary cultures of mouse JIP3 knockout neurons exhibit a striking increase in the overall abundance of axonal lysosomes with focal accumulations within axonal swellings that are strikingly similar to the lysosome-filled axonal swellings observed at amyloid plaques 23 . These changes raised questions about the relationship between lysosomes and the axonal cytoskeleton with potential implications for Alzheimer's disease.
To address these questions, we used human JIP3 KO iPSCderived cortical glutamatergic neurons that we recently established as a cellular model for investigating the impact of JIP3 depletion on neuronal cell biology 27 . Surprisingly, we found that axons with lysosome-filled swellings had a massive disruption in their actin-spectrin and myosin-II lattice organization that was not restricted to just the local site of the swelling but which occurred throughout each affected JIP3 KO axon. The additional KO of JIP4 further enhanced this phenotype, consistent with an overlapping role of the two proteins 27 . Axonal swellings filled with lysosomes were not static but formed and resolved over the course of several hours. Intriguingly, their formation coincided with local microtubule disorganization. Our observations support a model wherein perturbed axonal lysosome transport induced by loss of JIP3 (or JIP3 and JIP4) is closely linked to a broad disruption of the neuronal cytoskeleton.

Results
Global disruption of the membrane periodic skeleton in axons with lysosome-filled axonal swellings. The axonal plasma membrane is supported by an organized cytoskeletal network containing actin filaments and spectrin tetramers 12,28,29 . This actin-spectrin network, which has been linked to axonal mechanical stability 3,13 and signaling 30 , was also recently shown to undergo transient local expansion to facilitate the transport of large cargoes in narrow axons 31 . The important role of this cytoskeletal scaffold in controlling axon diameter suggested that lysosome-filled axonal swellings of JIP3 KO neurons might require major rearrangement to this network. To address this question, human induced pluripotent stem cells (iPSCs), which can be differentiated into layer 2/3 cortical glutamatergic neurons (i 3 Neurons) 27,32 were used in this study. The KO of JIP3 in this human i 3 Neuron model system robustly develops lysosome-filled axonal swellings similar to those observed in primary cultures of mouse JIP3 KO cortical neurons while overcoming practical challenges arising from the neonatal lethality in the JIP3 KO mouse model (Fig. 1a-c) 23,27 .
We next tested whether the axons of i 3 Neurons develop a membrane associated periodic skeleton similar to that observed in rodent neuron primary culture models 29,33 . Stimulated emission depletion (STED) super-resolution fluorescence microscopy of control i 3 Neurons after labeling with antibodies against the C-terminus of βII-spectrin, revealed a periodic spectrin lattice (180-200 nm intervals) ( Fig. 1d-f), in agreement with previous studies in other neuron culture systems 33,34 . In the control i 3 Neurons, this periodic spectrin organization was apparent by 9 days of differentiation and then persisted as the neurons aged (out to 21 days in this study; Fig. 1d). This lattice was also observed in JIP3 KO neurons by 9 days of differentiation, a stage at which they have undergone extensive axon growth but do not yet display lysosome accumulations ( Supplementary Fig. 1a). In contrast, in older JIP3 KO i 3 Neurons that developed lysosomefilled axonal swellings, the axonal spectrin lattice was disrupted ( Fig. 1g-j). This striking disruption was not limited to the local site of lysosome accumulation but occurred throughout the entire affected axon (Fig. 1g-j). This phenotype was even observed in axons with very sparse swellings ( Supplementary Fig. 1b). No difference was observed in the overall protein levels of βIIspectrin between control and JIP3 KO i 3 Neurons (Fig. 1k, l). The apparent loss of βII-spectrin signal in the microscopy experiments, therefore, most likely reflects the dispersal and washout of unassembled spectrin under the extraction conditions employed to enable super-resolution imaging of the rings formed by βIIspectrin. Consistent with the disruption of the spectrin periodic skeleton, the periodic organization of axonal F-actin was also lost and F-actin instead accumulated at the swellings (Fig. 1m, n). Collectively, this data indicates that the axonal actin-spectrin lattice still formed in JIP3 KO i 3 Neurons but later disassembled in parallel with the development of lysosome accumulations.
Non-muscle myosin-II associates with F-actin within the actinspectrin lattice and controls axon radial contractility 11,31 . We, therefore, examined the impact of the absence of JIP3 on the subcellular localization of myosin-II (Fig. 1m, n). In control i 3 Neurons, myosin-II filaments displayed a periodic pattern with occasional gaps along the axonal shaft (Fig. 1m). This organization was completely lost in the axons of JIP3 KO i 3 Neurons that had lysosome-filled axonal swellings (Fig. 1n). In these axons, the myosin-II signal, like F-actin, was instead most prominent within the swellings.
Lysosome-filled axonal swellings coincide with sites of abnormal microtubules organization. The assembly of the actin/ spectrin-based axonal periodic scaffold is dependent on intact microtubules (Zhong et al. 34 ). Given the drastic disruption of this scaffold in the swollen axons of JIP3 KO i 3 Neurons (as well as the essential role for microtubules in long range axonal transport of organelles), we next investigated microtubule organization. Remarkably, the sites at which lysosomes accumulate in JIP3 KO axons coincided with local disorganization of microtubules, as assessed by α-tubulin immunofluorescence (Fig. 2a-c). The disorganization of microtubules was further exacerbated in older  (Fig. 2d-f). The majority of microtubules within the swellings were bent and looped, while axonal regions immediately adjacent to the swellings contained microtubules that were organized in the typical parallel bundles of control axons (Fig. 2b). This disruption of the normal parallel arrangement of microtubules throughout axons was even more striking in cultures of JIP3 + JIP4 double KO i 3 Neurons, where as previously reported 27 , lysosome-filled swellings were larger and more abundant ( Fig. 2g-i). Of note, while lysosomes strongly accumulate in JIP3 KO i 3 Neurons and JIP3 + JIP4 double KO i 3 Neurons, mitochondria and synaptic vesicles are also present in these lysosome-positive swellings, but to a much lesser degree (Supplementary Fig. 2a-f).
Microtubules undergo several distinct post-translational modifications [35][36][37] , which reflect microtubule age and the activities of multiple tubulin-modifying enzymes 38,39 . Immunofluorescent labeling revealed acetylation of axonal microtubules in both control and JIP3 KO i 3 Neurons, and there were no noticeable differences in the acetylation status of looped versus parallellyorganized microtubules in the KO neurons ( Fig. 3a-c). In contrast, when microtubule tyrosination status was examined, the looped microtubules were found to be predominantly detyrosinated ( Fig. 3d-f). While the mechanisms that underlie the relationship between looping and detyrosination of microtubules remain uncertain, it is possible that detyrosinated microtubules are more stable, potentially through inhibited binding of the depolymerizing kinesin (MCAK) to microtubules 40,41 . In addition, as CAP-Gly domain containing proteins, such as the p150 Glued subunit of dynactin, prefer tyrosinated microtubules and tyrosinated microtubules have been proposed to be preferred tracks for initiating movement of axonal LAMP1-positive organelles (lysosomes) 42,43 , accumulation of detyrosinated microtubules within axonal swellings could have an impact on lysosome transport at such sites.
Axonal swellings are highly dynamic. We next performed longterm live cell imaging of lysosomes and microtubules in JIP3 KO i 3 Neurons to examine the dynamics of the axonal swellings. For the purpose of microtubule visualization, we employed low concentrations of the SiR-tubulin dye. This probe yielded a fluorescent microtubule pattern highly similar to the pattern of anti-α tubulin immunofluorescence (Fig. 2a, b) including the appearance of microtubule loops within axonal swellings (Fig. 2b, d, e). Time-lapse imaging of JIP3 KO i 3 Neurons stably expressing LAMP1-GFP (lysosome marker) and labeled with SiR-tubulin revealed that lysosome-filled axonal swellings form and resolve over a time scale of several hours (3.77 ± 1.71 h; . The close spatial and temporal relationship between lysosome accumulation and microtubule looping within these axonal swellings prevented definitive conclusions regarding a cause-effect relationship between these events (Fig. 4b, c).
To test whether microtubule perturbations can directly influence the focal accumulation of lysosomes, we acutely treated JIP3 KO i 3 Neurons with a low dose of taxol (10 nM) for 1 h, as a higher concentration (1 µM) of taxol showed apparent toxicity in these neurons. We found that low doses of taxol efficiently mobilized large fractions of lysosomes from the axonal swellings within a period of 1 h ( Supplementary Fig. 3a, b). These results support a coordinated role for JIP3 in the maintenance of lysosome transport and regulation of microtubule behavior.
To test the specificity of the phenotypes observed in JIP3 KO i 3 Neurons, we performed genome editing on JIP3 KO iPSCs to remove the 1 bp insertion in our JIP3 KO line and thus restore JIP3 expression (JIP3 Rescue). JIP3 expression was restored in the neurons derived from these JIP3 Rescue iPSCs (Fig. 5a). Furthermore, these JIP3 Rescue i 3 Neurons did not exhibit lysosome-positive focal swellings nor any disruptions to the microtubule organization ( Fig. 5b-e). In addition, the disruptions to the axonal spectrin lattice in swollen JIP3 KO axons ( Fig. 1g-i) were not found in JIP3 Rescue i 3 Neurons, as indicated by the βII-spectrin staining ( Fig. 5f-j). Collectively, these results demonstrate that both the lysosome accumulations and the cytoskeletal disorganization arose specifically due to the loss of JIP3 rather than any off-target effect of the genome editing strategy that was employed to generate the JIP3 KO line.

Discussion
Our investigation of the relationships between lysosome axonal transport and multiple components of the axonal cytoskeleton in neurons that lack JIP3, or both JIP3 and its paralogue JIP4, revealed that focal lysosome accumulations are accompanied by major disruptions in organization of the axonal membrane associated periodic skeleton. Furthermore, the accumulation of lysosomes in focal swellings of mutant neurons coincided both spatially and temporally with local microtubule disorganization. These observations reveal major unexpected relationships between axonal lysosome transport and the organization of multiple aspects of the axonal cytoskeleton. They furthermore raise new questions about the chain of events that link lysosome transport, microtubule organization and the cytoskeletal machinery that controls axon diameter.
Although impacts of JIP3 depletion on axonal lysosome abundance and/or transport have been observed in multiple . c Percentage of i 3 Neurons containing at least one lysosome-positive axonal swelling represented as mean ± SD, pooled from four independent experiments (n ≥ 32 per experiment, 13 days of differentiation). d, e STED microscopy images of βII-spectrin immunofluorescence in the axons of control i 3 Neurons (day 17). (Scale bars, d: 5 μm; e: 1 μm). f Graph demonstrating the longitudinal distance between peaks in the βII-spectrin signal from the boxed region in (e). g Airyscan microscopy images of control i 3 Neurons show regular distribution of lysosomes (LAMP1, white) and intact periodic membrane skeletons (βII-spectrin, green). h JIP3 KO i 3 Neurons (day 15) exhibit disruption in the spectrin organization in axons positive for lysosome accumulations (scale bars, 5 μm). i Percentage of swollen axons with disrupted periodic membrane skeleton represented as mean ± SD pooled from three independent experiments (≥20 axons analyzed per experiment). Lysosomes and the periodic membrane skeleton were labeled with LAMP1 and βII-spectrin antibodies, respectively. j Graph depicting the mean βII-spectrin fluorescence intensity of JIP3 KO neurites with and without lysosomefilled axonal swellings (≥35 µm in length) represented as mean ± SD, pooled from three independent experiments (n ≥ 20 in total). Note the possible contribution of some dendrites to the "No swellings" group. k, l Immunoblots showing levels of βII-spectrin in control and JIP3 KO i 3 Neurons (day 15); ribosomal protein S6 was used as loading control (k), and their normalized expression levels is shown in (l). m STED microscopy images of myosin-II filaments (green) show a periodic distribution similar to that of F-actin (magenta) in control i 3 Neurons. n Myosin-II and actin organization is lost in JIP3 KO i 3 Neurons (day 15) and both were enriched at the lysosome-positive axonal swellings (white). Lysosomes and myosin-II filaments were labeled with antibodies against LAMP1 and non-muscle myosin-IIA respectively, while rhodamine phalloidin was used to label F-actin. Scale bars, 5 μm. p-values were calculated using two-tailed Student's t-test. species 23,25,26,44 , broader effects of JIP3 mutation on the axonal abundance of multiple organelles have also been reported 45-49 . Our new observations of the dramatic cytoskeleton disruption that correlates with focal lysosome accumulations in both JIP3 KO (and JIP3 + JIP4 double KO) neurons provide a potential explanation for the broad requirement for JIP3/4 for the axonal transport of multiple organelles that would not require a direct involvement of JIP3/4 in the transport of each class of organelles.
The focal accumulations of lysosomes coincide spatially and temporally with local disorganization of microtubules in JIP3 KO i 3 Neurons and the overall severity of this phenotype is further exacerbated in the JIP3+JIP4 double KO i 3 Nneurons. Our initial interpretation of these lysosome phenotypes was that lysosomes fall off their microtubule tracks due to a loss of JIP3+4-dependent connections to motors. The surprising discovery that these sites of lysosome accumulation were accompanied by extensive looping of microtubules suggests a more complex relationship between lysosome transport and microtubule organization. It remains to be determined whether this reflects an impact of forces exerted on microtubules by lysosome movement versus a potential signaling mechanism controlled by JIP3. It was particularly striking that the dynamic formation of local axonal swellings with lysosome accumulation and microtubule looping in JIP3 KO neurons was accompanied by a widespread disruption of the integrity of the axonal membrane associated periodic skeleton. This all-or-none effect is consistent with two distinct interpretations. One is that the focal swellings elicit changes that are propagated along the entire axonal shaft. Another is that loss of JIP3 results in an age-dependent disruption of the periodic actin/spectrin-based scaffold that facilitates formation of the focal swellings where lysosomes accumulate. Interestingly, the assembly of the axonal membrane associated periodic skeleton is dependent on microtubule integrity 34,50 , suggesting that the cytoskeletal changes observed in JIP3 + JIP4 KO neurons may be interconnected. Interaction between the periodic membrane skeleton and microtubules may occur via several linker proteins, including ankyrins, which bind both spectrin and microtubules 34,51,52 . More recently, dynein has been implicated in the maintenance of the actin-spectrin cytoskeleton at the axon initial segment through endocytosisrelated mechanisms 53 . This provides another means through which the integrity of the actin-spectrin rings can be regulated by motor proteins. The elucidation of the precise sequence of events will require further experimentation.
The phenotypes that we observed in JIP3 KO axons are reminiscent of axonal beading that arises in response to multiple forms of axonal perturbation [54][55][56] . It was recently shown that changes in the local tension of the axonal membrane leads to the propagation of such "beads" 57 . Myosin-II filaments within the periodic membrane skeleton contribute to contractility and may control tension homeostasis along the axonal shaft 31 . Hence, if the dilations are upstream events, the all-ornone disruption of the axonal periodic scaffold within swollen JIP3 KO axons may be a consequence of tensional instability along the axons.  Ribosomal protein S6 immunoblots served as loading controls. b-d Airyscan imaging of Lysotracker (green) and Sir-tubulin (magenta) in control (b), JIP3 KO (c) and JIP3 Rescue d) i 3 Neurons (day 17), respectively (scale bars, 5 μm). e Percentage of i 3 Neurons containing at least one lysosome-positive axonal swelling represented as mean ± SD, pooled from three independent experiments (n ≥ 20 per experiment, 17 days of differentiation). f-h Airyscan microscopy images of lysosomes (white) and βII-spectrin (green) in control (f), JIP3 KO (g) and JIP3 Rescue (h) i 3 Neurons (Day 17), respectively. (yellow arrowheads highlight lysosome-filled swellings; scale bars, 5 μm). Lysosomes and the periodic membrane skeleton were labeled with LAMP1 and βIIspectrin antibodies, respectively. i Percentage of swollen axons with disrupted periodic membrane skeleton represented as mean ± SD pooled from two independent experiments (≥8 analyzed per experiment). p-values were calculated using two-tailed Student's t-tests.
Loss of JIP3 could also have signaling consequences that could propagate beyond its direct subcellular site of action. In addition to interactions with motors, JIP3 (and JIP4) also intersect with signaling in the JNK pathway by acting as a scaffold that regulates the subcellular position and activity of DLK and JNK 25,58-60 . Additionally, transport defects could impede the ability of lysosomes to act as sites for nutrient and growth factor-dependent activation of the mTORC1 signaling pathway 61,62 .
In conclusion, our findings have revealed new reciprocal interrelations between lysosome transport and the structure and the dynamics of the axonal cytoskeleton. The stalling of lysosomes during their retrograde journey correlated with profound changes in axons. A priority for future work will be to determine cause-effect relationships between the multiple tightly linked changes that we have observed here. In addition to advancing our knowledge about fundamental aspects of cell function, these studies may provide new insight into mechanisms relevant to Alzheimer's disease pathology as well as to human intellectual disabilities arising from mutations in the MAPK8IP3/JIP3 gene 63,64 .
CRISPR-Cas9 mediated rescue of JIP3 KO iPSCs. A CRISPR-based homologous recombination strategy was used to correct the 1 bp insertion in the JIP3 KO iPSC line. Briefly, 1 × 10 5 JIP3 KO iPSCs were plated on Matrigel-coated 6-well plate and transfected the following day using the Lipofectamine Stem transfection reagent (Thermo Fisher Scientific) and the following two components: 3 µg of px458 plasmid (Addgene plasmid #48138) containing a small guide RNA encoded within the following sense (5′CACCGGGCGGCGTGGTGGTGTTACC3′) and antisense (5′AAAC GGTAACACCACCACGCCGCCC 3′) sequences that was designed to selectively target the JIP3 KO sequence containing the 1 bp insertion and a 140 bp single stranded DNA oligonucleotide repair template (5 µl of 100 µM stock) that overlapped the gRNA-targeted region. The sequence of the DNA template was the following: 5′GCCGCGCTGGCGGCGGCGGTGGCCGCGATGATGGAGATCC AGATGGACGAGGGCGGCGGCGTGGTGGTGTACCAGGACGACTACTGCT CCGGCTCGGTGATGTCGGAGCGGGTGTCGGGCCTGGCGGGCTCCATCT ACCG 3′. Transfected (GFP-positive) cells were enriched by fluorescence activated cell sorting (FACS) 2 days later. Sorted cells were expanded and then serially diluted to yield single cell-derived clonal populations. 40 clones were selected and screened using PCR amplification of genomic DNA flanking the sgRNA target site followed by sequencing of the amplicons using the primers described in Gowrishankar, et al. 27 . complex was added to the imaging dish, followed by media replacement with fresh cortical media on the following day. For Taxol treatment, i 3 Neurons were labeled with lysotracker as described above and then treated with 10 nM or 1 μM taxol (Sigma) for 1 h.
Subsequently, the nitrocellulose membranes were washed 3 times (10 min each) in TBST and probed by incubation for 1 h with the secondary antibodies conjugated with horseradish peroxidase. The membranes were then washed three times (15 min at room temperature each), developed using Pierce™ ECL western blotting substratum (Thermo Fisher Scientific) and imaged by a Versa-Doc imaging system (Bio-Rad).
Uncropped images of the immunoblots presented in Fig. 1k and Fig. 5a of this paper are shown in Supplementary Fig. 4.
Fluorescence microscopy. Two types of high-resolution microscopes were used in this study. The LSM 880 inverted confocal laser scanning microscope with Airyscan (Carl Zeiss Microscopy) accompanied with 63×/1.40 numerical aperture (NA) plan-apochromat differential interference contrast (DIC) oil immersion objective and 32-channel gallium arsenide phosphide (GaAsP)-photomultiplier tubes (PMT) area detector and 488 nm, 561 nm and 633 laser lines was used in this study. Images were acquired and processed using ZEN imaging software (Zeiss). The Leica TCS SP8 gated STED super-resolution confocal microscope (Leica Microsystems) is coupled with Leica harmonic compound (HC) plan apochromatic (PL APO) 100×/1.40 oil STED objective and Leica gated HyD hybrid detector. Briefly, a white light excitation laser accompanied with 592 nm, 660 nm and 775 nm depletion lasers was used in this study. Images were acquired using LAS X software (Leica Microsystems) and final images were deconvolved using Huygens deconvolution software (Huygens Essentials, Scientific Volume Imaging).
Quantification and statistical analysis. Images were pseudocolor-coded, adjusted for brightness and contrast, cropped and/or rotated using the open-source image processing software FIJI (ImageJ) 65 . Dendrites and axons were identified by visual tracking of the length of the neurite. Percentage of neurites were quantified using the FIJI plugin "NeuronJ" and/or the FIJI segmented lines + ROI manager tool to determine the total number of neurites. For quantification of βII-spectrin intensity, neurites of at least 35 μm in length were semi-automatically traced and quantified using "NeuronJ" to determine the mean fluorescence value (8-bit) of the traced segments. In the case of JIP3 KO i 3 Neurons where βII-spectrin intensity is markedly reduced, LAMP1 staining was used to trace the swollen axons as described above. The traced segments were then overlayed to the βII-spectrin channel for subsequent quantification.
Immunoblot data were processed using Image Lab software (Bio-Rad) and quantified using the "Gels" ImageJ plugin. The methods for statistical analysis and sizes of the samples (n) are specified in the results section or figure legends for all of the quantitative data. Student's t test or Mann-Whitney test was used when comparing two datasets. Differences were accepted as significant for P < 0.05. Prism version 9 (GraphPad Software) was used to plot, analyze and represent the data.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
All data generated or analyzed during this study are included in this published article (and its Supplementary Information files: Supplementary Material and Supplementary Data 1). Raw datasets generated during and/or analyzed during the current study are available from the corresponding authors on reasonable request.