Stem Cells Given by IV to Rats Whose Spinal Cords Were Cut in Two Places—at the Upper Lesion a Detour Formed to the Outside, at the Lower One a Different Route Crossed to the Other Side

It is rare for the spinal cord to be damaged at two separate levels at once. Among 710 consecutive spinal fractures at Wakayama Medical University there were 18 such cases, 2.5% (Takami, M., et al., Eur J Orthop Surg Traumatol 2017;27:79–86). At an Indian tertiary care center, 38 cases amounted to 3.01% of all spinal injuries (Goyal, S., et al., J Clin Neurosci 2025;135:111198).

Rare, and severe when it does happen. Nine of the 18 Wakayama cases (50.0%) were complete paralysis. In 1 of those 18, the second fracture was missed at the first examination. When the upper injury produces dense paralysis, it is hard to notice what is going on below it. In the Indian series of 38, meanwhile, only 4 showed neurological improvement after surgery.

Something here does not sit right. If the two lesions are at separate levels, part of the cord ought to survive at each of them. So why is the outcome worse than for someone crushed deeply at a single level? Today’s paper takes that question into an animal experiment. It is a study in rats, and the “lines” that keep appearing are not axons but trajectories computed from diffusion data.

Publication Details

• Paper: Obara, H., et al. Biomedicines 2026;14:1979 (13 authors)
• Paper link (DOI): https://doi.org/10.3390/biomedicines14091979
• Journal: Biomedicines (MDPI, Basel, Switzerland). Received 24 July 2026, accepted 31 August, published 2 September. A legitimate open-access journal indexed in Web of Science (SCIE) and Scopus, although the catalog record at the US National Library of Medicine states that it is “not currently indexed for MEDLINE”
• Impact Factor: 4.5 (2025 Journal Citation Reports, publisher’s announcement)
• License: Creative Commons Attribution (CC BY)
• Funding and ethics: four JSPS KAKENHI grants, the North Tech Foundation, and two grants from the US Department of Veterans Affairs. Conflicts of interest declared as “none”. The animal experiments were approved by Sapporo Medical University (24-056)

Senior Author Osamu Honmou, and the Department of Neural Regenerative Medicine at Sapporo Medical University

The last author is Osamu Honmou, Professor in the Department of Neural Regenerative Medicine, Institute of Regenerative Medicine, Sapporo Medical University. From 1992 he worked at the Center for Neuroscience and Regeneration Research at Yale University School of Medicine, and after returning to Japan he has held his current post since 2011 (researcher database). What he has led is Japan’s IV-MSC program: bone-marrow-derived mesenchymal stem cells (MSC) expanded in the patient’s own serum and returned through a vein. It began with an open-label trial in cerebral infarction, passed through an investigator-initiated trial in spinal cord injury, and arrived at the conditional, time-limited approval of “Stemirac” in December 2018 (the controversy comes later).

The department runs its own clinical service inside the university hospital, a Department of Neural Regenerative Medicine, so that everything from basic research to clinical application sits within a single unit (organizational overview). Ten of the paper’s 13 authors belong to this department, two of them holding joint appointments in the Department of Orthopedic Surgery. The other pillar is Yale: co-author Jeffery D. Kocsis is Professor of Neurology and Neuroscience at Yale University School of Medicine (official Yale profile). Four authors, three of them on the Japanese side, list a second affiliation with Yale Neurology, and the funding includes grants from the US Department of Veterans Affairs.

What Wasn’t Understood Before?

Cut the spinal cord in cross-section and you find H-shaped gray matter (the cell bodies of the neurons) at the center, surrounded by white matter (axons, the bundled wiring). The white matter divides into three funiculi: the dorsal funiculus (DF) at the back, the lateral funiculus (LF) at the sides, and the ventral funiculus (VF) at the front. Split left from right, and those six compartments are this paper’s measuring stick. In rats, the dorsal CST (dCST), which carries about 95% of the axons of the descending corticospinal tract (CST), runs down the dorsal funiculus (as the paper states, citing Weidner, N., et al., PNAS 2001;98:3513–3518). In humans the main trunk travels in the lateral funiculus, so “if the dorsal funiculus is destroyed the legs stop moving” is simply wrong as a statement about people.

👦 Student: But if someone is diagnosed with “complete paralysis”, doesn’t that mean nothing at all is left?

🧬 Dr. Exotaro: That is precisely what it does not mean. Of 139 patients with spinal cord injury, 88 were clinically classified as “complete”, and in 74 of those 88 (84%) there was evidence that influence from the brain still reached across the lesion. The authors called this discomplete—complete in appearance, incomplete in fact (Sherwood, A.M., et al., J Neurol Sci 1992;110:90–98). And in a cohort of 227 cervical injuries, each additional 1 mm of preserved tissue bridging the lesion went with an average of a further 10.9% (SD 1.3) of the maximum motor score returning at 12 months (Pfyffer, D., et al., Lancet Neurol 2024;23:816–825). How much is left over decides the prognosis.

The model in this paper turns that premise around. The right side is hemisected at T7, the left side at T12. Because the intact side swaps over between the upper and the lower lesion, any descending pathway running straight from brain to lumbar cord must, by design, be severed somewhere. It is like blocking the northbound lanes at one junction and the southbound lanes further on. (This layout goes back to Courtine, G., et al., Nat Med 2008;14:69–74, which showed that cortical control of walking can return through relays of propriospinal neurons even when the direct pathway is cut; it is not these authors’ invention.) That said, the paper only calls the model discomplete; it never verifies histologically how far each hemisection actually extended. What can honestly be written is that the model is designed to cut that way—no more.

So how do you look at the roads that remain? The traditional approach was to add color. Inject an anterograde tracer into the motor cortex and only the axons extending from there are labeled, so they can be followed through sections. But there is a constraint.

👦 Student: Can’t you just stain the road you want to know about? Is that really such a problem?

🧬 Dr. Exotaro: It is a problem. You want to draw a map, and yet you cannot measure anything until you have already decided which road to investigate. It is like wanting to understand a city’s traffic jams and painting only the main highway. If recovery ran along some unanticipated back street, that street was never painted, so it will never show up on your map.

What this paper used instead is the other approach—diffusion tensor imaging (DTI). Inside a bundle of axons, cell membranes and myelin sheaths act as walls: water moves easily along the direction the bundle runs, and with difficulty across it. MRI can measure how far water moved in each direction, so for every voxel (a three-dimensional pixel) that mobility can be modeled as an ellipsoid, and lines can be drawn by linking the orientation of each ellipsoid to its neighbor’s. That is tractography, and the lines it draws are streamlines (reconstructed tracts). No staining is needed, and you do not have to decide in advance which road to examine. The weaknesses come later.

There is a second premise, on the treatment side. MSCs sent into a vein do not swim to the injured spinal cord. Cultured MSCs are larger than the pulmonary capillaries, so they come to a halt in the lung, the first organ they reach, like an insect hitting a window screen (Lee, R.H., et al., Cell Stem Cell 2009;5:54–63). And yet effects still appear. The prevailing explanation is that the cells stuck there send “letters” out to the rest of the body (a previously covered study on continuous infusion deals with this in detail).

What was not known lies beyond that point. In a spinal cord carrying lesions at two separated levels, how does a single intravenous dose of MSCs change the way the long tracts run? The abstract puts it this way: “how MSCs influence the reorganization of long tracts spanning multiple lesions remains unclear”.

What Did This Paper Find?

The Skeleton of the Experiment, and How They Walked

Male Sprague-Dawley rats, 8 weeks old. With a microblade, the left side is hemisected at T12 first, then the right side at T7. The day after injury, MSCs prepared from the femoral bone marrow of five other rats are given as a single intravenous injection into the femoral vein, 1×10⁶ cells per animal (the control group receives cell-free DMEM). Because this is not an autograft, cyclosporine A 10 mg/kg is injected subcutaneously every day. The paper’s wording is “All rats”, starting the day before the cells or vehicle were given, and the only reason offered is to avoid an immune reaction—that one point and nothing else. The groups are 5 intact controls, 5 vehicle, 5 MSC. Motor function is scored over four weeks with the BBB score (the Basso–Beattie–Bresnahan locomotor rating scale, an abbreviation unrelated to the BBB of the blood-brain barrier) by an examiner blinded to treatment group.

(mean ± standard error, n = 5 per group. The intact group held at 21 points)

These two numbers must not be read as “from 5.4 to 12.2”. They are two groups placed side by side on day 28, not a before-and-after change in the same animals. The BBB is a rating sheet, not a pedometer, and in the original scale (Basso, D.M., Beattie, M.S., Bresnahan, J.C., J Neurotrauma 1995;12:1–21) the band from 8 to 13 points is where an animal moves from dragging its hindlimbs to placing the sole of the foot down and stepping with weight support. 5.4 sits in the band below that, but 12.2 does not reach the intact score of 21 either. Because it is a rating sheet, there is no guarantee that the one point from 5 to 6 weighs the same as the one point from 12 to 13; this is not a measure to be talked about in multiples.

Nine Hours at 7 Tesla, and 36 Possible Destinations

The excised cords are fixed for two weeks and cut into 2 cm blocks, one centered on T7 and one on T12. Imaging is on a 7.0-tesla small-animal MRI: b value 800 s/mm², 30 diffusion gradient directions, slice thickness 0.5 mm, 55 slices, roughly nine hours per specimen. This is only possible because the specimen is dead, and the price is that the same animal can never be imaged a second time. That is why this study has a single time point, at four weeks. The analysis is the deterministic streamline method in DSI Studio (17 May 2024 build). Tracking is guided by FA (fractional anisotropy, how well aligned the direction of diffusion is), the angular threshold is 90 degrees, and one million seeds are placed per region of interest (ROI). The ROIs, too, were placed by someone blinded to treatment group.

The crucial point is that a number like “160,000” does not mean 160,000 axons. The software is closer to a program that generates a road map automatically from satellite photographs; the lines it draws are products of an algorithm. The paper defines them that way itself, in the methods section. A streamline is “a three-dimensional trajectory computed by following the principal diffusion direction, not an individual axon”, and “tract counts are relative, semi-quantitative measures and do not represent absolute axon numbers”. Caveats to the same effect are placed in the results section and in the limitations section as well.

So how is a “detour” counted? Cross-sections are taken rostral and caudal to each lesion, and the six compartment ROIs are drawn on them. For T7, that means T6 at 2.5 mm rostral to the lesion center and T8 at 2.5 mm caudal. 2.5 mm above and 2.5 mm below, a 5 mm window in total, which in the rat thoracic cord is about two segments (on the T12 side only the segments T11 and L1 are specified; no distance in millimeters is given in the paper). That yields 36 combinations. A line that passed through the same compartment is a preserved tract; one that moved to a different compartment is a redirected tract. This too is an operational classification, and the paper states explicitly that it “does not establish identity at the level of the axon”.

At T7, a Detour Formed to the Outside

The total tract count at the center of the T7 lesion was 268,468 ± 3,815 in intact animals, 80,800 ± 6,326 in vehicle, and 160,026 ± 9,226 in MSC (p < 0.001 for MSC versus vehicle). Among the lines running all the way from T6 to T8, preserved tracts numbered 138,923 ± 3,026 intact, 14,498 ± 1,189 vehicle, and 45,241 ± 2,618 MSC (p < 0.001 for every pairwise comparison). The MSC group comes nowhere near intact. The accurate description is not recovery, but a loss that was blunted.

The interesting part is the redirected tracts: 4,227 ± 1,280 intact, 1,898 ± 1,073 vehicle, 17,572 ± 3,588 MSC, with the MSC group exceeding not only vehicle but the intact group as well (p < 0.01 for both). The same holds as a share of the total: 3.0 ± 0.9% intact, 7.5 ± 2.6% vehicle, 22.0 ± 3.0% MSC (p < 0.01 versus vehicle, p < 0.001 versus intact).

The T7 lesion is on the right, so the uninjured side is the left. Two pathways rose on that left side. (The triplets below follow the order in which the text states them, intact then vehicle then MSC, read in that order; the numbers themselves carry no group labels.)

• Left DF at T6 → left LF at T8: 215 ± 180, 216 ± 160, 4,552 ± 720 (p < 0.001 for both comparisons)
• Left VF at T6 → left LF at T8: 183 ± 183, 52 ± 48, 3,409 ± 808 (p < 0.001)

Lines that had been running in the dorsal and ventral funiculi of the uninjured side were swung out into the lateral funiculus, forming a route that curves around the lesion from the outside. The schematic calls this a lateral detour pathway. At the same time the paper states plainly that MSC administration did not restore preserved tracts on the injured side. What is responding is the uninjured side.

At T12, the Detour Crossed the Midline Instead of Going Lateral

At the caudal T12 lesion the left side is cut, so the uninjured side is the right. Since this is the mirror image of T7, the same detour—right DF and right VF into right LF—might have been expected, but it did not appear in the same way. The description, however, splits into two statements, and lumping them together produces an error.

• Redirection from right VF at T11 into right LF “was not observed at the T12 lesion level”.
• Redirection from right DF at T11 into right LF “was more frequent in the MSC group than in the vehicle group, but the difference was not significant”.

To write, as a single summary, that “no detour by way of the lateral funiculus occurs at T12” would be to convert two different things—a pathway that was not observed, and a pathway that stopped at a non-significant trend—into one negative finding.

What rose in its place was a different route. Lines crossing from the dorsal funiculus at T11 on the injured (left) side over the midline to the dorsal funiculus at L1 on the contralateral (right) side increased clearly in the MSC group: 185 ± 170 intact, 47 ± 36 vehicle, 1,047 ± 409 MSC (p < 0.05 versus intact, p < 0.05 versus vehicle).

Preserved tracts at the T12 level were 125,127 ± 3,343 intact, 15,349 ± 1,259 vehicle, and 42,244 ± 1,430 MSC (the text prints p < 0.0001 for MSC versus vehicle). Redirected tracts were 6,099 ± 565 intact, 3,559 ± 1,135 vehicle, and 7,759 ± 1,229 MSC, significantly more than vehicle (p < 0.05) but not significantly different from intact. As a mean value the MSC group’s 7,759 does exceed the intact group’s 6,099, yet the relationship that held at T7, of significantly surpassing the intact group, is not confirmed at T12.

One more striking figure is the preserved count in the dorsal funiculus on the uninjured (right) side at T12. Against 12,631 ± 1,020 in intact animals, vehicle managed 22 ± 22 and the MSC group 2,303 ± 1,003 (significantly more than vehicle, p < 0.01, though nowhere near intact). This is what the side that was supposedly never injured looks like at four weeks.

👦 Student: Almost zero, on the side that has no injury at all?

🧬 Dr. Exotaro: Seen from T12 the right side is uninjured, but that same right side was hemisected higher up, at T7. On a highway, this stretch of road is accident-free, and yet lanes have already been lost at an interchange upstream—the authors think that upstream cut is what is doing the work. That, however, belongs to the explanation, not to the measured results. I will separate the two next.

Observation and Explanation Are Different Things

Every number set out so far is an observation: a measured result, not an interpretation. The one exception, anticipated in the exchange just above, is the line of reasoning that “it is because the cord was cut upstream at T7”—that is not a measured number but an explanation the paper offers in its discussion. Let me keep the two apart.

The authors’ account runs as follows. The dorsal and ventral funiculi on the uninjured side at T12, that is the right side, had already lost much of their long tract content to the right-sided hemisection higher up at T7, leaving too little axonal substrate to support a redirection into the lateral funiculus. In its place, the dorsal funiculus on the injured (left) side, which had been spared at T7, joined the compensatory reorganization, producing the pattern that crosses the midline to the contralateral dorsal funiculus. They conclude that once redirection has occurred at one spinal level, the same redirection pattern is not readily established at a second lesion.

👦 Student: You mean the stock of usable detours is finite?

🧬 Dr. Exotaro: It is an attractive idea, but stop there. That is an explanation put forward in the discussion, not a tested result. There are only three groups—intact, vehicle and MSC—and no single-lesion control in which only T12 was cut. So the difference between T7 and T12 cannot be separated into “the stock was spent upstream” and “the lesion levels are simply different”; the mid-thoracic cord and the thoracolumbar junction differ both in the composition of their white matter tracts and in cross-sectional size. A puzzle has been presented. It has not been solved.

How Will the Future Change? (The Path to the Clinic)

The path to the clinic is best divided into three layers.

The first layer is treatment that already reaches patients. On 28 December 2018, intravenous administration of autologous bone marrow MSCs received conditional, time-limited approval as “Stemirac”. This is a scheme that puts a product on the market with a limit of up to seven years, at the stage where efficacy can merely be presumed. It is indicated for severe cases of traumatic spinal cord injury, with bone marrow aspiration to be performed within about 31 days of injury (explanation from Sapporo Medical University Hospital). The evidence base, however, was an investigator-initiated trial at a single center, open-label, with no control group and 13 patients. A 2026 commentary in Spine Journal describes the approval as resting on “limited and uncontrolled clinical evidence”, arguing that premature commercialization under public insurance can damage both scientific rigor and public trust (Kawaguchi, H., Spine J 2026;26:1239–1242). Full approval was applied for on 14 November 2025 and, as of September 2026, is under review (for the regulatory system, see the article summarizing stem cell therapy in Japan).

👦 Student: So does this paper show how that drug works?

🧬 Dr. Exotaro: That goes too far. This is an experiment in rats, using cells from other animals, with cyclosporine A injected every day. The product is human autologous MSCs cultured in the patient’s own serum, and no immunosuppressant is needed. The only thing that overlaps is the timeline: a single dose into a vein soon after injury. And what was shown goes only as far as the fact that the group given cells had better imaging findings and better walking; whether the lines that were drawn made the walking better is something the paper itself calls unclear.

The second layer is clinical trials. Most MSC trials in spinal cord injury worldwide are phase I or phase I/II studies of a few to a few dozen participants, and the great majority use the intrathecal route, injecting from the back into the space around the cord. Trials built around intravenous administration are, as far as can be confirmed on ClinicalTrials.gov, very few, and no example outside Japan has carried a design that completes with an infusion alone all the way to a marketed product. The transplantation of iPS cell-derived neural progenitor cells being pursued at Keio University is a different strategy: if that one is about replacing the cells that were lost, this one is about making the remaining wiring do more work.

The third layer is the preclinical work to which this paper belongs. What moved forward here is arguably not the treatment but the measurement. Even so, no patient can be scanned for nine hours at 7 tesla; clinical spinal cord DTI works at far lower resolution and far shorter acquisition times, and has to hold up against breathing and pulsation. Whether the rewiring of lines seen in an excised specimen can be seen in the clinic is a separate question.

How to Critically Read This Research (Limitations and Paths to Improve It)

Before the criticism, three respects in which this paper is above average. Blinding is stated explicitly in two places. Almost every bar chart shows individual values as dots, so the reader can check the spread across the five animals. And the limitations section volunteers the most painful items itself: streamlines are not axons, there is no histological confirmation, there is a single time point and males only, and “how these changes contribute to motor recovery remains unclear”.

The correct criticism is not that the limitations go unacknowledged. They are written down accurately, and yet the title of the paper and the middle of the discussion proceed in the language of axons with those reservations set aside. That gradient of caution is exactly what deserves reading. Six points.

1. There is zero histological corroboration, and the direction of the lines is unknowable. No anterograde tracer and no immunostaining were used, and there is no histological quantification anywhere. The “location” of each hemisection is said to have been confirmed on axial MR images, but nothing is written about checking in sections how far that hemisection extended. On top of that, what diffusion MRI sees is the ease of water movement, so whether a line runs from top to bottom or from bottom to top is unknowable in principle. The centerpiece finding at T12, the lines crossing the midline to the contralateral dorsal funiculus, might be compensation by a descending motor pathway, or it might be reorganization of ascending sensory fibers. The word “ascending” never appears once in the text, and nowhere are ascending sensory fibers discussed as a constituent of the dorsal funiculus. The roughly 95% quoted earlier is a breakdown of the CST, not a breakdown of the dorsal funiculus.

2. What the “long tracts” actually amount to is a 5 mm window. The paper states that the spacing of the ROIs allows it to look selectively at long-distance fibers and minimize the contribution of interneurons. In practice, though, the window is 2.5 mm above and 2.5 mm below the lesion, 5 mm in total, about two segments in the rat thoracic cord. All that this spacing reliably excludes are connections that begin and end within less than two segments, and there is no statement verifying how many longer pathways are mixed in. Meanwhile, the discussion of the same paper puts at the top of its list of detour mechanisms the “long descending propriospinal neurons that bypass the lesion through the lateral funiculus and relay cortical commands to the lumbar motor circuitry”. The very population the ROI spacing is supposed to exclude and the population supposed to carry the detour sit inside the same propriospinal system. And because the specimens were cut into separate 2 cm blocks for T7 and for T12, tracking that runs continuously from T6 to L1 cannot be performed at all. The analysis that would directly test the “long-tract reorganization that circumvents the lesion” promised by the paper’s title was shut out at the stage of specimen preparation.

3. The tracking parameters are tuned in the permissive direction. An angular threshold of 90 degrees is effectively the ceiling, allowing a turn of up to a right angle at every step. It is a setting that readily generates lines hopping from one compartment to another, that is, the primary outcome itself. The anisotropy threshold, too, is relative to an Otsu value computed per specimen. The nominal parameters may be identical, but the effective threshold can move with how damaged each specimen is. Since that degree of damage is also the variable that separates the groups, it can confound the between-group comparison.

4. The diffusion metrics themselves are never reported. FA is mentioned only as the quantity that guided tracking; no per-group values appear. The same goes for MD, AD and RD, and for the cross-sectional area of spared white matter and lesion volume. A tract count is the number of lines passing through the ROIs, so it rises mechanically as there is more trackable tissue and higher anisotropy. No evidence is offered that “the MSC group has more tracts” is anything more than a restatement of “the MSC group has more trackable white matter preserved”.

5. The intact spinal cord also contains thousands of “redirected” lines. 4,227 ± 1,280 at T7, and 6,099 ± 565 at T12. The boundaries between funiculi are not anatomical walls, and fibers change position as they cross segments. In other words, this category is not specific to injury. Normal fiber trajectories, where the ROI lines were drawn, reconstruction artifacts—how much each of these contributes cannot be worked out from the paper. The paper itself writes that redirected tracts “indicate only changes in reconstructed trajectories, cannot establish identity at the level of the axon, and cannot exclude reconstruction artifacts”. And at T12, the total number of redirected tracts in the MSC group is statistically indistinguishable from the intact group.

6. There is no control group to test the central explanation, and the reporting has holes as well. Whether the difference between T7 and T12 is because “the material was spent upstream” or because “the lesion levels themselves differ” cannot be separated, since there is no single-lesion control group. There is a second hole in the controls. The comparator is five intact rats, and no sham group was included that had been taken as far as shaving down the articular processes and exposing the cord. What is more, all the paper writes about the intact group is a single sentence: “5 non-SCI rats served as intact controls”. The anesthesia used for surgery (ketamine 90 mg/kg plus xylazine 4 mg/kg) and the twice-daily manual bladder expression for 14 days after surgery are both described inside the section covering the hemisection procedure, and nothing is written separately about how much of that the intact group received. Cyclosporine A, it is true, is stated to have been given daily to “All rats”. But that sentence sits in the section describing allocation to the vehicle and MSC groups, and the start of dosing is defined as from the day before the cells or vehicle were administered, a criterion that has meaning only for the SCI groups. Whether the intact group falls inside or outside this “all rats” cannot be settled from the text. The only things that can be read straightforwardly as applying to all three groups are the anesthesia at harvest (ketamine 50 mg/kg plus xylazine 10 mg/kg) and the perfusion procedure, steps that every animal necessarily passed through given that DTI was performed. Surgical trauma itself and postoperative care are not described for the group that had no surgery. The reading offered earlier, that preserved tracts are reduced even on the side that should never have been injured, rests precisely on this comparison with the intact group. Which means the contribution of surgical trauma and postoperative care has not been subtracted out of it. As for statistics, the effect sizes on the main total-count measures are enormous, so “nothing can be said with n of 5” is inaccurate. What does bite is that for distributions where the standard error approaches the mean, the vehicle group’s 22 ± 22 for instance, the figure legends for the three-group comparisons report, without exception, one-way ANOVA with Tukey’s multiple comparisons. On correction for multiplicity across the 36-pathway search, the text says nothing at all (the accurate statement is “not described”, not “not corrected”). In addition, there is no panel at T12 corresponding to the proportion of redirected tracts that supported the claim at T7, so no number is presented that compares how readily the two lesions detour on the same scale. Note also that cyclosporine A was given to both the vehicle and the MSC group, so the conditions for comparing those two are matched, but it is a comparison between animals that all received an immunosuppressant, not a comparison against an untreated natural course.

So what would have made it stronger? Two things require zero additional experiments. A scatter plot pairing BBB score against tract count animal by animal; and a sensitivity analysis that lowers the angular threshold from 90 degrees to 60 and to 45, and separately varies the anisotropy threshold (set to be drawn at random per seed from 0.5 to 0.7 of the Otsu value), to see whether the findings survive. If additional animals are permitted, then first, inject an anterograde tracer (BDA or AAV) into the motor cortex, section the very same cord that was imaged with DTI, and overlay the labeled axons and the streamlines in the same coordinate frame (unlike a study that asked what the tracer had actually stained, this paper does no staining at all). Second, image from T5 to L2 in one piece rather than cutting the cord in two, which would allow the primary outcome promised by the title to be seen directly. And third, a single-lesion control hemisected only at T12. If the switch from right DF and VF into LF occurs without a T7 lesion, the explanation is supported; if it does not, the cause of the difference lies in the lesion level itself. It should be added that neither the raw diffusion-weighted images, nor the ROI masks, nor the per-animal data tables have any stated repository. The data availability statement says only that the original contributions presented in the study are included in the article and that further inquiries can be directed to the corresponding author. For a paper whose main findings depend on a reconstruction pipeline, that too weighs heavily.

Dr. Exotaro’s Perspective

As a physician specializing in regenerative medicine, what I liked best in this paper was not a result but its self-restraint. It defines what it is counting, and then states in three separate places—results, limitations and methods—that what it counts is not axons. A study that can say plainly what it is measuring is strong even when its conclusions are modest. Conversely, it is worth knowing that the devices which make a modest conclusion look strong—the paper’s title, and the thick arrows in the schematic—live inside the same paper.

With that said, let me write down the distance as it looks from the consulting room. First, time. In this experiment the cells went in the day after injury. The treatment approved in Japan likewise sets bone marrow aspiration within 31 days of injury. This is all about the period while secondary injury is still unfolding. For someone reading this two years after their own injury, these results do not straightforwardly become their own story. In a treatment protocol, “when, and how many times” is more likely to be the design variable than dose. The chronic phase remains a separate problem, to be solved separately.

Next, the measuring stick. A BBB score of 12.2 sits in the upper part of the band where the hindlimbs can bear weight and step. But intact is 21 points, and 12.2 does not reach it. The human ASIA Impairment Scale is a different scale altogether, and how many points an animal moved cannot be translated into a human grade.

Nor is the drawing of a detour on an image, in itself, a promise to patients. A line being drawn and that line moving the right muscle at the right moment are two different things. If there is a promise, it comes down to a single point: there may be room to get more out of the wiring that remains. And since, for the moment, that is the only realistic route to making up for what was lost, it is not a small point.

If the explanation that “there is a finite stock of material available for detours, and spending it upstream leaves none downstream” is correct, the implications reach beyond spinal cord injury and impose constraints of order and timing on the treatment of multifocal central nervous system injury in general. Until a single-lesion control group is included, though, this stays a hypothesis. If I had to name one missing item, it would be a scatter plot laying BBB score against tract count animal by animal. No extra experiment is needed, and the numbers are already in hand. It would not close the blank the paper itself labels unclear—how changes in the drawn lines contribute to the recovery of walking. An association is not proof of contribution, and causality is a matter for a different experiment. Even so, it would be a first step toward that blank.

This is four weeks in rats, not a report of a person walking again. It is, though, one paper that entered a poorly lit clinical problem, injury at two levels, by way of designing how to measure it. And the thing that did not sit right at the outset—why the detours came out so differently above and below—remains open. That it is presented while still open is part of this paper’s honesty.

This article was written with the assistance of AI.