Longitudinal optical imaging technique to visualize progressive axonal damage after brain injury in mice reveals responses to different minocycline treatments

A high-resolution, three-dimensional, optical imaging technique for the murine brain was developed to identify the effects of different therapeutic windows for preclinical brain research. This technique tracks the same cells over several weeks. We conducted a pilot study of a promising drug to treat diffuse axonal injury (DAI) caused by traumatic brain injury, using two different therapeutic windows, as a means to demonstrate the utility of this novel longitudinal imaging technique. DAI causes immediate, sporadic axon damage followed by progressive secondary axon damage. We administered minocycline for three days commencing one hour after injury in one treatment group and beginning 72 hours after injury in another group to demonstrate the method’s ability to show how and when the therapeutic drug exerts protective and/or healing effects. Fewer varicosities developed in acutely treated mice while more varicosities resolved in mice with delayed treatment. For both treatments, the drug arrested development of new axonal damage by 30 days. In addition to evaluation of therapeutics for traumatic brain injury, this hybrid microlens imaging method should be useful to study other types of brain injury and neurodegeneration and cellular responses to treatment.

experimental design. All mice were anesthetized (1-1.5% isoflurane) and imaged prior to injury, using a multiphoton microscope to view the brain through an implanted gradient index (GRIN) lens. The numbers and locations of axons in a three-dimensional field of view (248 μm × 248 μm × 100 μm) were recorded prior to injury to establish a baseline for each mouse. Mice were re-imaged at pre-determined time points up to 60 days (Fig. 1). Prior to the Day 30 and Day 60 time point, some mice lost their head plates, and as result, in vivo images were not collected beyond that time. The number of mice included at each time point can be found in Table 1. Mice with lost head plates had their craniectomies sealed with 1% low melt agarose, a cover glass, and dental acrylic. Mice with representative Day 60 in vivo images are available in Supplemental Results (Fig. S1). The neurological severity score (NSS) and rotarod tests were conducted on days 2, 5, and 7 post-injury to assess Figure 1. Experimental Design. GRIN lenses were implanted ≥21 days before baseline in vivo, two-photon imaging. Acute behavioral tests (rotarod and neurological severity score (NSS)) were conducted on days 2, 5, and 7 following midline fluid percussion injury (FPI). Mice were imaged one week prior to midline fluid percussion injury (mFPI) and at 1 hr, and at 3, 7, 14, 30, and 60 days following injury. The open field, novel object recognition, and tail suspension test were conducted on Day 60. www.nature.com/scientificreports www.nature.com/scientificreports/ injury-induced sensorimotor deficits and functional recovery following brain injury and treatment, as previously described 29 . Novel object recognition, open field, and tail suspension test were performed on Day 60 to assess the long-term effect of minocycline on cognitive and affective behavior, respectively. All behavioral tests were performed in a room solely for mouse behavioral tests and analyses were performed by persons blinded to treatment conditions. Detailed methods for the behavioral tests performed at 60 days, along with the experimental results, are included in the Supplemental Data File.
Mice were excluded from the study if herniation occurred after injury (n = 1), if injured mice did not show deficits in the rotarod or NSS tests (n = 1), if sham-vehicle mice displayed deficits following sham injury (n = 1), or if head plates were detached before Day 7 (n = 2). These mice were not included in Table 1 or in our analyses.
GRin lens assembly and implantation. Implantable micro-optics systems were constructed in-house by first securing a custom glass window (500-µm diameter, 600-µm length, Bern Optics, Inc., Westfield, MA, USA) to a 3-mm diameter #1.5 glass cover slip (Electron Microscopy Sciences, Hatfield, Pennsylvania, USA) with NOA 71 optically clear adhesive (Norland Products, Inc., Cranbury, New Jersey, USA). Next, a drop of the optically clear adhesive was placed on the glass window and a 500 µm diameter, 1.7 mm long, just under ½ pitch, ILH, uncoated gradient index (GRIN) lens (Go!Foton, Somerset, New Jersey, USA) was attached. Adhesive was cured for 90 min under an ultraviolet lamp. At 45 min, the imaging system was rotated 180° to ensure uniform curing. The following day, the cover slip on the lens system was secured to a stainless steel M1.6 washer (ID = 1.7 mm, OD = 4 mm, McMaster-Carr, Santa Fe Springs, CA) with optically clear adhesive and cured again for 90 min, with rotation at 45 min to complete the GRIN lens assembly.
GRIN lens assemblies were implanted, as previously described 1,30,31 . Mice were briefly anesthetized with 5% isoflurane and then injected with a ketamine/xylazine (10 mg/kg ketamine, 1 mg/kg xylazine in 0.9% saline, i.p.) cocktail. This additional anesthesia step allowed surgeons ample time to prep the mouse for surgery. Mice were transferred to a stereotaxic frame (Stoelting Co., Wood Dale, Illinois, USA) equipped with ear bars and a nose cone for isoflurane delivery (Leica Biosystems. Inc., Buffalo Grove, Illinois, USA). A SomnoSuite system (Kent Scientific, Torrington, Connecticut, USA) was used to control isoflurane delivery and a PhysioSuite system (Kent Scientific) to maintain mouse temperature at 37 °C. For the duration of the surgery, isoflurane was maintained at 1.0-1.5%. To ensure the mouse was fully anesthetized, toe pinch withdrawal was monitored every 15 min. A craniectomy was created 2.25 mm lateral and 0.9 mm rostral to bregma. A lens assembly, held on a cannula insertion tool, was implanted to a depth of 1.4 mm at a rate of 0.1 mm/min until the imaging system rested immediately superior (dorsal) to the external capsule. RelyX dental acrylate (RelyX Unicem Aplicap, 3 M ESPE, St. Paul, MN, USA) was placed around the imaging system and allowed to cure for 20 min before the mouse was returned to its home cage with a heating pad (Braintree Scientific, Inc., Braintree, Massachusetts, USA). Mice were observed until a recumbent position was maintained. To minimize post-surgical pain, an analgesic was dissolved in the mouse's water (Children's Ibuprophen, 30 mg/kg/day in water) for 72 hr following the procedure. Mice health and welfare were monitored for 72 hr following surgery. No intervention was necessary in any of the mice in this study. Mice were allowed to recover at least three weeks before baseline (pre-injury) imaging 1,30 . Head plate design and attachment. Head plates were designed in SolidWorks software and printed (MakerBot Replicator 2) in-house with polylactic acid filament. They were secured to the skull with cyanoacrylate (Super Glue, Loctite) following implantation of the GRIN lens, as previously described 31 . Head plates were designed with two cutouts. One cutout served to aid in aligning the head plate with the washer portion of the GRIN lens assembly (Supplemental Fig. S2A) The second cutout allowed for a second surgery to be performed and the injury hub 32 attached to facilitate midline fluid percussion injury (mFPI, Supplemental Fig. S2B). The file for printing head plate and parts for the TRIO imaging support system is available at the National Institutes of Health 3D Print Exchange at https://3dprint.nih.gov/discover/3dpx-010720, Model ID 3DPX-010720 1 .
In vivo multiphoton microscopy. For in vivo multiphoton imaging before and after mFPI, the TRIO platform was used to restrain, anesthetize, and warm the mouse for the duration of the imaging session, as previously described 31 . A Chameleon Vision-2 multiphoton laser (80 MHz, Coherent) tuned to 890 nm was used to excite YFP fluorescence and the emitted light was filtered through a 525/40 nm filter (BrightLine Filters, Semrock, Inc.). A custom, upright, multiphoton microscope system (Intelligent Imaging Innovations, Inc., Denver, CO, USA) and a 40×/0.6 NA objective with a correction collar for up to 2 mm of glass (Nikon Instruments Inc., Japan) was used to acquire scanned images 33 of the external capsule. To acquire three dimensional (z-stack) images composed of 100 images with 1 µm step size, a dwell time of 2 µs and pixel averaging (5/scan) was used. Baseline images were acquired 3-4 weeks after lens implantation, which was one week prior to the mFPI procedure. Post-injury images were acquired at 1 hr, 3, 7, 14, 30, and 60 days after injury. Image sets were exported and stored as TIFF files on an external hard drive for offline analysis.  Sham-vehicle  4  4  4  4  3  1   TBI-vehicle  5  5  5  5  4  2   TBI-45-min mino  5  5  5  5  4  0   TBI-72-hr mino  5  5  5  5  4 3 www.nature.com/scientificreports www.nature.com/scientificreports/ Adapted mfpi surgery. The mFPI was conducted, as previously described 29,34,35 with an adaptation made for the attachment of the injury hub. Kwik-Sil (World Precision Instruments, Sarasota, FL), a biocompatible, silicone, elastomer was used to attach the injury hub in place of dental cement to protect the GRIN lens from being removed during injury hub detachment 1 . Mice were anesthetized with 5% isoflurane and then transferred to a stereotaxic frame equipped with ear bars and a nose cone for continuous isoflurane delivery. Mice were maintained at 1.5-2.0% in air using a SomnoSuite system and a thermostatically-controlled warming pad. Isoflurane was used instead of a ketamine/xylazine cocktail, as this and other injectable anesthetics, can disrupt the injury response after TBI induction 36 . A craniotomy was created using a 3-mm trephine between lambda and bregma, along the midline without disrupting the dura. An injury hub was fabricated from a Leur lock needle hub and fixed over the craniotomy using Kwik-Sil. Injury hubs were filled with saline to check for leaks and then covered with a piece of standard lab tape to prevent bedding and other particles from accessing the exposed dura. Lab tape remained on the injury hub until it was filled with saline for injury induction. Mice were placed on a heated pad and monitored until recumbent, at which point they were returned to their housing overnight. Diffuse brain injury by mFPI. Diffuse brain injury by mFPI was conducted 24 hr following the attachment of the injury hub, as previously described 37 . Group sizes are indicated in the Results section. Mice were anesthetized using 5% isoflurane for 5 min. Isoflurane was used as the anesthetic, as ketamine/xylazine and other injectable anesthetics can alter the injury response following TBI 36 . The lab tape was removed from the injury hub and the intact dura was inspected for debris. The hub was filled with 0.9% sterile saline and connected to an FPI device (Custom Design and Fabrication, Virginia Commonwealth University, Richmond, VA). When mice responded to a toe pinch, an injury was administered by releasing the pendulum. Sham-injured mice followed the same protocol, except the pendulum was not released. Righting reflexes, the total time from the initial impact until the mouse righted itself, were recorded as indicators of injury. Mice subjected to mFPI had an average righting time of 426 ± 24 s, while sham-injured animals had a righting time of 23 ± 1 s. Following injury, mice were re-anesthetized with 5% isoflurane and injury hubs were carefully removed by gently pulling off the Kwik-Sil with sterilized forceps and dura was inspected for uniform herniation and integrity of the dura 35 . The craniectomy was filled with 1% low melt agarose (Amresco, Irvine, CA) and a 3 mm diameter #1.5 cover glass (Electron Microscopy Sciences, Hatfield, Pennsylvania, USA) was placed over the opening. Gel control cyanoacrylate was carefully placed around the cover glass. Next, RelyX dental acrylic was placed around and on top of the cover glass to secure the cover glass to the skull 1 . Once cured, the mouse was transferred to the TRIO platform for the 1-hr imaging time point.

Minocycline administration protocol.
Prior to injury hub attachment, mice were randomly divided into four groups as follows: sham injury with saline treatment (Sham-vehicle, n = 6), TBI with saline treatment (TBI-vehicle, n = 6), TBI with minocycline treatment beginning at 45 min post-TBI (TBI-45-min minocycline, n = 6), and TBI with minocycline treatment commencing at 72 hr post-TBI (TBI-72-hr minocycline, n = 6). Minocycline (45 mg/kg) (Tocris Bioscience) in 100 μl of 0.9% saline (Teknova, Hollister, CA), or 100 μl of vehicle, was administered intraperitoneally three times for each group. An initial dose was given at the stated time point, and then another two doses were administered at 24 hr and 48 hr after the first dose. The same mice were used for imaging and behavioral tests. The number of mice included for imaging analysis at each time point can be found in Table 1.

Behavioral tests to assess injury.
To determine if acute sensorimotor deficits were present following mFPI, rotarod performance and NSS were evaluated at Days 2, 5, and 7 29,35,[38][39][40][41] . The NSS was primarily used as exclusion criterion. If injured mice had an NSS score of 3 or less, they were excluded (n = 1). If Sham-vehicle mice had a score of greater than 3, they were excluded (n = 1).
Rotarod. An Accurate Rotarod (AccuScan Instruments, Inc. Columbus, Ohio USA) was used to assess sensorimotor function. Mice were acclimated to the device three days before injury (acclimation phase) (Fig. 1). The mouse was placed on the stationary rod and allowed to explore the rod for 30 s. If the mouse fell off the rod before the 30 s time period, the mouse was placed back on the rod. Once the mouse had successfully explored the stationary rod for 30 s, the angular frequency of the rotarod was adjusted to a constant rate of five revolutions per min (rpm). If the mouse fell off the rod, it was placed back on the rod and the timer restarted until it could walk for 15 s. Following successful exploration, the rotarod speed was adjusted to an acceleration of 0.2 rpm/s. The trial ended once the mouse fell off the accelerating rod. The mouse was returned to its cage for 10 min and the accelerating rod portion of the test was re-run. Again, once the mouse fell, the mouse was returned to its home cage, and the acclimation phase ended. The acclimation phase was run between 12 pm and 4 pm on the day of the testing phase.
Rotarod tests were run on Days 3, 2, and 1 before injury (Fig. 1). The procedure took place at the same time each day, between 4 pm and 7 pm to minimize variability in performance related to diurnal cycles. The mice were placed on the apparatus and were tested for their ability to remain on the rotarod with an acceleration of 0.2 rpm/s. The apparatus is equipped with sensors to detect when a mouse falls and it records the latency to fall in s. Three trials were run on each testing day, with the first and second trials run one after the other and the third trial run after a 10-min break. The two longest trials were averaged for the day. The average of the two longest trials for one day before injury was used as the baseline data point. The rotarod test was conducted on Days 2, 5, and 7 following mFPI to detect injury-induced motor deficits. Since mice were not all tested on the same day, the latency to fall at Day 2, Day 5, and Day 7 was normalized to baseline latency to fall ( ) Day o r Latency to Fall Baseline Latency to Fall 2, 5 7 to account for both inter-animal and inter-day variability 30 .

Scientific RepoRtS |
(2020) 10:7815 | https://doi.org/10.1038/s41598-020-64783-x www.nature.com/scientificreports www.nature.com/scientificreports/ Neurological severity score. To evaluate neurological impairment, which affects acute sensorimotor performance following mFPI, eight behavioral tests were performed to establish a NSS for each mouse 29,35,40,41 . The tests determined the function of the cortex and cerebellum, and if any additional motor deficits were present. The mice were assessed on the presence of (1) hind limb flexion, (2) seeking behavior, (3) startle response, ability to walk across 30 cm long, square cross section, beams with progressively smaller widths of (4) 3 cm, (5) 2 cm, and (6) 1 cm width, and to (7) balance on a 0.5 cm square beam and a (8) 0.5 cm rod with a circular cross section. Each of these 8 tasks was scored either as a one for failure or zero for success. The sum of the scores was the NSS. These tests were conducted on Days 2, 5, and 7 following injury. Data are presented as a score ranging from zero to eight, with high scores indicative of impairment.
image selection and parameters. Exported 16-bit TIFF files were analyzed in ImageJ (version 1.51i) software 42 . Each stack for each time point was realigned using "StackReg, " a downloadable plug-in, to correct minor misalignment of successive images caused by motion artifacts 43 . After correcting image alignment, an average intensity z projection of 11 successive images acquired in 1-µm z steps was created for each time point. An identical field of view using a particular arrangement of axons was chosen for each time point to track the same axons over time.

Quantification of axons with undulations, varicosities and terminal bulbs.
The percentage of axons with undulations, varicosities, and terminal bulbs was determined at each time point. Individuals blind to the treatment groups, compared the post-injury time points to baseline images. Observers used the drawing and region of interest (ROI) tools in ImageJ (version 1.51i) to indicate which axons had features of damage. A box was drawn around the axon and it was classified as having evidence of an undulation, varicosities, or a terminal bulb. If an axon with varicosities developed a terminal bulb, it was only counted as a terminal bulb and not both at that time point. Axons were counted if they demonstrated a feature of damage; individual features (i.e., each swelling) were not counted, since the entire axon was not in a field of view. Axons appearing normal were counted as uninjured. The ROIs were tabulated at each time point for each mouse in Microsoft Excel. Percent was calculated as follows: × 100 Axons with feature Total axons in the field of view . Data were presented as the mean ± SD percentage of axons with each feature at each time point in each treatment group. Data were transferred to GraphPad Prism for statistical analysis 1 .

Quantification of new varicosities at each time point. Axons with varicosities were tracked over time.
At each time point, the varicosities were labeled as either newly developed or progressing from the previous time point. At each time point, the fraction of axons with newly developed varicosities was calculated by dividing the axons with new varicosities (N t ) by the total number of axons with varicosities (V t ). The percentage of new axons with varicosities was recorded (F N (t) = N t /V t *100). Data were reported as mean total percentage of axons with newly developed varicosities ± SD 1 .

Quantification of recovered axons at each time point.
Axons with varicosities were labeled as recovered if the axon returned to its baseline morphology by the subsequent time-point. For example, if an axon had varicosities at Day 3, but then recovered by Day 7, it was labeled as recovered by Day 7. The fraction of axons that recovered was calculated by dividing the number of recovered axons (R t ) by the total number of varicosities (V t ). The percentage of axons that recovered was recorded (F R (t) = R t /V t *100). Data were reported as mean total percentage of recovered axons ± SD by the respective time points 1 .
Quantification of axon fate. To demonstrate progressive axonal injury, individual axons were tracked over time. Varicosities and terminal bulbs have been observed in healthy, adult rodents 44 . If an axon had varicosities or an axonal bulb in the baseline image, the damaged axon was excluded from future analysis (approximately two axons per mouse had varicosities or a terminal bulb in baseline images). Undulated axons at the 1-hr time point were followed over the course of the experiment and their outcome was recorded as either developing varicosities or returning to its baseline state. The percentage of undulated axons that developed varicosities or recovered back to baseline morphology was measured. Axons with varicosities were tracked over time and at the end of the 30-day time period varicosities were identified as (1) developed into a terminal bulb or lost axon, (2) recovered (no sign of injury), or (3) unrecovered (varicosities remained). The percentage of axons in the field of view that developed into a terminal bulb, recovered, or remained damaged were reported.
Quantification of total varicosities and terminal bulbs or axon loss. The total percentage of axons with varicosities and terminal bulbs that developed over the course of 30 days was measured for each mouse. Terminal bulb and axon loss were measured together, but if a terminal bulb resulted in axon loss at the end of the experiment, it was not counted twice. Occasionally, axon loss would occur without the apparent development of a terminal bulb. The terminal bulb could have developed outside the field of view or passed through the field of view at a time point not imaged in this experiment.
Statistical analysis. Results for each group are shown as the mean ± SD 45 . Repeated measure analysis of variance (ANOVA) was performed to analyze the behavioral tests as the same subsets/cohorts of mice were analyzed over time. To analyze the NSS data, we used the Scheirer-Ray-Hare test since NSS are ordinal variables and do not follow normality assumption, necessary in ANOVA. Scheirer-Ray-Hare test is a nonparametric procedure to test significant differences among groups of independent variables on a continuous or ordinal dependent variable. ANOVA was used to test if there was a significant difference in the amount of damage (measured in the percentages of axons undulated, developed to varicosities, and TD) among the four groups. Power analyses for the (2020) 10 , where s p 2 measures the pooled sample variance). All tests were performed using R packages at α = 0.05 level of significance. The p-values and F statistics are reported individually in the figure legends or results section, along with the statistical analysis used for the data set. The number of mice included in the imaging analysis at each time point is listed in Table 1.

Results
Righting reflex as an indicator of brain injury severity. Mice were either subjected to moderate mFPI diffuse brain injury (n = 15) or sham injury (n = 4), in which the mice undergo the same procedures as the brain-injured group except the pendulum on the device was not allowed to strike the fluid percussion piston. Figure 2 represents the mean righting time in each experimental group (Sham-vehicle, n = 4; TBI-vehicle, n = 5; TBI-45-min minocycline, n = 5; TBI-72-hr minocycline, n = 5). All brain-injured groups had significantly longer righting times than the Sham-vehicle mice.
Minocycline reduced acute sensorimotor impairment post-injury. The NSS was conducted during the first week post-injury (Fig. 3A). On Day 2, the TBI-vehicle, TBI-45-min, and TBI-72-hr minocycline treated mice had significant deficits as compared to the Sham-vehicle mice, as expected. Both groups of minocycline treated mice had significantly improved NNS scores from Day 2 to Day 5 and Day 2 to Day 7. However, there was no significant difference in mean scores between Day 5 and Day 7 for either group.
Minocycline reduced motor deficits over one-week post-injury. The accelerating rotarod test was conducted on Day 2, 5, and 7 to assess motor deficits (Fig. 3B). The latency to fall on Day 2, 5, and 7 was normalized to baseline latency to fall to account for inter-animal variability. On Day 2 post-injury, the TBI-vehicle mice had significant motor deficits as compared to the Sham-vehicle mice. On Day 5 and 7, TBI-45-min minocycline treated mice performed significantly better than TBI-vehicle mice. The TBI-72-hr minocycline treated mice did not perform significantly different from TBI-vehicle or TBI-45-min minocycline mice at any time-point. The administration of minocycline commencing at 45 min following injury reduced motor deficits over 1 week following mFPI as compared to the TBI-vehicle treated mice.
time-course images of axonal damage following mfpi and minocycline administration. Images of the same brain region in live mice were collected over the course of 30 days to determine the time course of axonal damage following mFPI in both sham-injured (Fig. 4A-F) and brain-injured, untreated (Fig. 4G-L) mice. Images were also acquired to monitor the effect of minocycline on secondary axonal damage at both a known therapeutic time window (TBI-45-min mino, Fig. 4M-R) and at a novel, expanded therapeutic window (TBI-72-hr mino, 4S-X). In brain-injured mice, varicosities (Fig. 4,>), terminal bulbs (Fig. 4, asterisk), and disappearing axons (Fig. 4, dashed arrow) were observed. These hallmarks of DAI 46 were quantified over the course of 30 days to demonstrate the effect of minocycline versus saline treatment on the progression of axonal injury and repair.  (Fig. 5A, arrows) 11,12 and represent mechanical damage from the diffuse brain injury forces. Undulations were apparent at 1-hr post-injury. A few undulations persisted on Days 3 and 7, but by Day 14, no axons with undulated features were within the field of view (Fig. 5B). Minocycline had no effect on the development of undulations, as expected, since undulations are a direct result of mechanical forces applied during the induction of injury. Axons with undulations were tracked over time to determine if undulated axons from primary damage developed into varicosities (Fig. 5A, top panel) or returned to the baseline morphology (Fig. 5A, bottom panel). In both the TBI-vehicle and TBI-72-hr minocycline treated group, significantly more of the observed undulated axons developed into varicosities as compared to the Sham-vehicle mice. While the TBI-45-min minocycline treated mice had more observed undulations develop into varicosities as compared to the Sham-vehicle mice, it was not significant (Fig. 6A).

Development of varicosities and terminal bulbs following mfpi and minocycline administration.
Axons with varicosities were identified by the presence of periodic swellings, as compared to their baseline. At each time point, the percentage of axons in the field of view with varicosities was measured (Fig. 6B). At one hr, all brain-injured mice developed varicosities, independent of treatment. TBI-vehicle treated mice had a significantly higher percentage of axons with varicosities than the Sham-vehicle mice at Days 3, 7, and 14. The TBI-45-min minocycline treated mice were not significantly different from the Sham-vehicle mice at any time point. On Day 14, the TBI-45-min minocycline treated mice had a smaller, nonsignificant percentage of axons with varicosities as compared to the TBI-vehicle mice. The TBI-72-hr minocycline treated group had significantly more varicosities at Day 3 and Day 7 as compared to the Sham-vehicle mice. However, by Day 14, the TBI-72-hr www.nature.com/scientificreports www.nature.com/scientificreports/ minocycline treated mice had fewer varicosities as compared to the TBI-vehicle treated mice, although significance was not achieved. At Day 30, there was no significant difference between any group. Some single swellings appeared in uninjured brains that disappeared over time. Isolated swellings (i.e., not periodic) occur occasionally in adult rodents and are not associated with injury 44 ; thus, these were not included.
Terminal bulbs were identified as large swellings (Fig. 7A,B, dashed arrow) at the end of axons (dashed arrows). Over the 30-day time period, the TBI-vehicle treated mice developed significantly more terminal bulbs than the Sham-vehicle mice (Fig. 7C). Both minocycline treated groups had a significantly smaller percentage of axons with terminal bulbs or axon loss than the TBI-vehicle treated group. Minocycline mitigated further axonal damage and prevented axon loss.  www.nature.com/scientificreports www.nature.com/scientificreports/ were classified as either recovered (Fig. 8, dashed arrow), persisting (Fig. 8, solid arrow), or developing into a terminal bulb at the end of the 30-day time period. One mouse in each group lost a head plate before Day 30 (Table 1) which prevented further imaging. As expected, the Sham-vehicle treated group had a significantly smaller percentage of axons with varicosities 44 develop over the 30-day time period as compared to TBI-vehicle treated mice (p = 0.002) ( Table 2). Of the axons with varicosities (54.7% ± 9.1%) in the TBI-vehicle group, 47.5% ± 14.3% recovered, 26.6% ± 8.4% remained unresolved, and 25.9 ± 13.7% of these axons developed terminal bulbs or resulted in axon loss. The TBI-45-min minocycline mice had a non-significant higher percentage of axons with varicosities than the sham-vehicle (p = 0.10) over the 30-day time period, and this group had a non-significant smaller percentage than the TBI-vehicle mice (p = 0.08; Table 2). Notably, none of the tracked axons with varicosities in the TBI-45-min minocycline treated group developed terminal bulbs or axon loss. Occasionally, terminal bulbs and axon loss were identified without observing varicosities at a prior time point (Fig. 4, dashed arrow). The occurrence of terminal blubs and axon loss in this population of axons was significantly less than the TBI-vehicle group (p = 0.002, p = 0.049, for TBI-45-min and TBI-72-hr, respectively, (Fig. 7). Because the injury is diffuse, some axons could have developed varicosities outside of the field of view or between imaging time points. However, while none of the tracked axons developed terminal bulbs or axon loss in the TBI-45-min minocycline treated mice, about half of the tracked axons with varicosities persisted (45.8% ± 45.9%). The TBI-72-hr minocycline mice had a non-significant, higher percentage of axons with varicosities develop over the 30 days than the sham-vehicle mice (p = 0.08) ( Table 2). None of the tracked axons developed into terminal bulbs and the majority of the damaged axons recovered (78.3% ± 15.6%) back to their baseline state by 30-days. Administration of minocycline commencing at 45 min resulted in the lowest percentage of varicosity development (33.1 ± 5.3%) over the course of 30 days after injury. However, the data suggest that there was little therapeutic effect of acutely-administered minocycline on repairing axonal damage. In contrast, administration commencing at 72 hr did not protect against early development of varicosities, as would be expected with delayed administration. However, the majority of the damaged axons eventually returned to the baseline state, suggesting that later administration of minocycline promoted recovery.

Evolution of secondary damage is reduced with minocycline treatment at 30 day time
We conducted an additional assessment to determine when a treatment condition was contributing to the recovery or preventing further damage. We determined the percentage of new axons with varicosities and the percentage of recovered axons at each time point (Fig. 9). At Day 3 post-mFPI, all groups had a higher percentage of new damage occurring than recovery, as seen by the higher percentages of axons with new varicosities (solid line, circle) than the percentage of recovered axons (dashed line, square). As expected, new damage was lower in the TBI-45-min treatment group compared to vehicle and TBI-72-hr group at the Day 3 time point. New varicosity development continued to decrease by Day 7 for all groups with the TBI-vehicle-treated group maintaining a higher percentage versus TBI-45-min minocycline. New varicosities continued to develop in TBI-vehicle-treated mice at each remaining time point, suggesting persisting secondary damage 12 . At Day 30, 16.7 ± 33.33% of the axons had newly formed varicosities, demonstrating continued secondary injury in TBI-vehicle mice. In contrast, no new damage developed in tracked axons in either minocycline-treated groups by Day 30 (Fig. 9B,C). Recovery occurred in each minocycline treatment group with a relatively low percentage of axons with varicosities that had resolved by Day 3 and a larger percentage by Day 7. Recovery and new damage were equivalent and unchanged from Day 7 to Day 30 after brain injury in untreated mice. After completion of therapy (Fig. 9B,C, arrowhead), www.nature.com/scientificreports www.nature.com/scientificreports/ the percentage of recovered axons continued to increase, although this was not statistically significant when compared across all time points. These results suggest that minocycline not only attenuates and eventually arrests new secondary damage, but it also promotes axonal recovery.   Table 1. www.nature.com/scientificreports www.nature.com/scientificreports/ Behavioral tests conducted at Day 60 post-injury. At Day 60, three behavioral tests were conducted to demonstrate that neurobehavioral tests can be conducted on the same mice that are used for imaging to assess the neurological effect of minocycline on long-term behavioral outcomes. In both the tail suspension test (Fig. S3) and open field (Fig. S4), no difference was found between treatment groups. However, at Day 60, TBI-45-min minocycline treated mice spent significantly more time exploring the novel object rather than the familiar object during the novel object recognition test (p = 0.04) (Fig. S5A). Both the TBI-vehicle and TBI-72-hr treated mice spent similar amounts of time with the familiar and novel object. However, no significant difference was found between the discrimination index for each of the four treatment groups (Fig. S5B). These data represent group sizes that were powered for imaging outcomes; thus, the behavioral data must be interpreted with caution.

Discussion
In this pilot study, we demonstrated the utility of our novel longitudinal imaging method 1 for preclinical drug testing that allows researchers to observe the timing and magnitude of the effects of drugs at high cellular resolution for a variety of neurodegenerative diseases. This study extends the mechanistic understanding of DAI as axonal pathologies develop and resolve in subcortical white matter. By combining transgenic reporter mice and GRIN lens two-photon imaging, we have advanced the understanding of axonal injury, degeneration, and recovery in the diffuse injured brain, as well as refined the time course of protection of minocycline at a cellular  www.nature.com/scientificreports www.nature.com/scientificreports/ level. At two weeks post-injury, delayed minocycline administration had the same net neuroprotective effects against secondary damage as acute administration, although with seemingly different rates of secondary injury and recovery. Both acute and delayed treatment appeared to arrest development of new damage by 30 DPI, which was in sharp contrast to continued development of new damage in TBI-vehicle treated mice.
To date, research pertaining to the progression of axonal damage and degeneration after TBI has been conducted with in vitro techniques and traditional histological approaches in animal studies 7,8,11,12,16,[46][47][48] . While these methods have offered insights into DAI and secondary injury, individual axons cannot be followed over time to observe the natural and drug-induced recovery of axonal damage that is a confounding factor in comparing the progression of axonal damage over time. Unlike previous studies, we followed individual axons a week before and for over 30 days post-TBI to determine the time course and fate of each injured axon, including the time course of the effect that minocycline had on the reduction of secondary damage and the recovery of axonal damage.
We successfully followed the same axons in intact mouse brain, before and after TBI and confirmed key observations from earlier in vitro and histological studies. For example, we observed that the mechanical forces from the initial injury lead to undulated axons, which quickly develop varicosities in less than one hr. In fact, by 1-hr post-injury there were few undulated axons, and axons had either recovered or developed varicosities in all injured groups. This aligns with prior studies, as axons of cultured cells subjected to stretch injury relaxed or became permeable by 45-min post-stretch 5 and axons in an animal model of FPI developed varicosities in as little as 15 min post-injury 12 . Once axons become damaged, they are susceptible to disconnection, leading to the development of terminal bulbs and ultimately, axon loss 12,46,49 , which we also observed. Furthermore, we also showed that damaged axons do not always develop terminal bulbs but are capable of recovering back to their baseline morphology 50,51 . We further demonstrated that recovery can happen rapidly or slowly, over several days, and that it occurs more frequently with minocycline treatment. Additionally, a recent study using CLARITY demonstrated that varicosity size decreases with survival time, alluding to axonal recovery mechanisms that reinstate transport 52 . Excitingly, our longitudinal study has allowed us to observe the reduction in size of individual axonal swellings over time, directly supporting this finding. The temporal resolution of this technique has not only allowed us to observe spontaneous recovery of axons following injury, but has also demonstrated the efficacy of minocycline, a potential treatment, at the individual axon level.
Our longitudinal imaging approach was further validated by confirming that when minocycline is administered at 45 min after injury, it protects the white matter over time, as others have shown 22,23,53 . In addition, this system was used to determine the effectiveness of expanding the therapeutic window of minocycline treatment to 72 hr after injury. While administration at this late time point did not protect against overall varicosity development, it also protected white matter long term. Furthermore, it appeared to result in a higher percentage of recovered axons. Some differences between treatment groups at a particular time point were often not significant. Our study was powered for repeated measurements over the duration of the study. Prospective preclinical studies would require additional mice per condition to confirm drug effects at specific time points or sex-dependent differences.
Minocycline has long-term protective benefits following TBI when administered up to 4 hr post-injury [22][23][24][25][26][27][28]53,54 . At 24 hr post-injury, minocycline significantly reduced the number of CD11b + cells, a marker for inflammation. At 3 months following experimental closed head injury, minocycline significantly protected the volume of the corpus callosum 27 . These studies suggest, while minocycline does not have immediate protective benefits, it protects white matter long term. However, prior studies did not determine exactly how and when minocycline exhibits its protective benefits.
Minocycline, regardless of the dosing strategy, did not protect against the development of varicosities from initial mechanical damage in the present study, as others have reported 22 . In both treatment groups, a similar percentage of undulated axons at the 1-hr time point developed varicosities, compared to the saline treated mice (Fig. 6). This is not a surprise, as both treatments were administered in a time window after application of mechanical force to the brain that initiates undulations, cytoskeletal damage, and vesicle transport disruption in injured axons. Therefore, post-injury treatment could not prevent this damage, which has been shown to occur within 15 min post-mFPI 12 . Future studies that explore resilience or priming strategies could have an impact on primary mechanical injury to axons. Yet, minocycline treatment had significant therapeutic efficacy in our study with mitigation of secondary injuries, such as varicosities and terminal bulbs, which can otherwise lead to axon loss. One limitation of our study was that the percentage of undulated axons was low with only 2-3 undulated axons in any injured group (Fig. 5B). This could be due to the timing of our first image acquisition, which was one hr after injury. The small breaks resulting from undulations cause a disruption in axonal transport, leading to varicosities in as little as 15 min 12 . For future studies, adding a time window within 15-30 min immediately following injury may reveal a higher number of undulated axons that can be better tracked over time.
Distinct protective profiles emerged between early and delayed treatment in our study. Administration from 45 min through 3 DPI significantly reduced overall damage, whereas delayed administration appeared to establish a more permissive environment for axonal recovery ( Table 2). Possibly the well-known, potent anti-inflammatory mechanism of minocycline is better utilized for recovery when delayed 25,[55][56][57][58] . However, chronic delayed dosing may interfere with microglial activation necessary for repair, as suggested by a clinical study in adults who had a TBI at least 6 months prior to a 12-week treatment with minocycline. Some patients had a biomarker of axonal degeneration, which could have resulted from decreased microglial activation 59 . Perhaps an acute and then slightly delayed, subacute treatment scheme may obviate the need for chronic dosing or enable smaller chronic doses of the drug over time to maintain protective benefits against secondary damage without risking neurodegeneration.
The ability to image the same cells before and following TBI, or development of other types of neurodegeneration, minimizes inter-animal variability, and thus reduces the number of animals required for studies. We had excellent statistical power for most of our end-point comparisons, except for the behavioral tests, which had high variability. Male and female mice were included in each treatment group resulting in low numbers per sex. This may have contributed to the large variance and non-significant differences in neurological function at 2 months post-injury, since female rodents may outperform males in cognitive function and motor coordination tests [60][61][62] . However, all injured animals exhibited acute neurological impairments that resulted in acute sensorimotor deficits, providing an experimental TBI that models clinical symptoms of diffuse brain injury. The injury-induced neurological symptoms were unaffected by the earlier lens surgery or the presence of the head-stage assembly, as we have previously demonstrated 30 .
Findings from this study have broadened our understanding of the effects of DAI and of minocycline on protection and recovery up to 30 days after traumatic brain injury. However, this was a proof-of-concept project intended to demonstrate the capabilities of our imaging method and was not a full-scale preclinical trial or a comprehensive basic research study. As such, some of the results within a particular image day lack sufficient statistical power (most comparisons of cumulative data over 30 days were adequately powered). To confirm trends seen between treatments at a specific time point, future experiments would require more mice per group. Future work would also include evaluation of chemical biomarkers, additional types of behavioral tests, functional imaging, with genetically-encoded fluorescent reporters, such as iGluSnFR to monitor neurochemical signal dynamics. We also anesthetized our mice in this study, which was not a problem for this study of cell-level damage. However, when using fluorescent reporters to record neurochemical signals, anesthesia should not be used. In this case, mice can be head-fixed in a device such as the Mobile Home Cage (Neurotar) or a trackball, both of which facilitate behavioral tests during image acquisition.
In summary, we have developed an in vivo, high-resolution, optical imaging technique to track individual cells deep in the mouse brain to monitor changes in the same cells over several weeks. Using this technique in a brain injury model, we identified the fate of individual axons shortly after brain injury through a time that included the development of secondary damage and we found that delayed dosing of minocycline may enhance recovery. Our technique is applicable to brain disorders involving cell loss and changes in cell morphology, connectivity, dendritic structure, and changes in neurotransmitter levels. This method should be useful for studies ranging from preclinical trials to basic neuroscience research in areas such as brain injury, aging, addiction, and neurodegeneration.