Comparative study of pre- and post-mortem perfusion of fixative for the quality of neuronal tissue preparation

Abstract

Ante-mortem transcardiac perfusion of a fixative agent is generally recommended for quality preparations for cerebral histology, ensuring rapid and deep penetration in the tissue to preserve the most fragile brain structures. Despite being performed under anesthesia and with proper analgesia, this procedure is cumbersome for the experimenter and raises ethical questions. Recently, alternative protocols have been proposed based on prior animal euthanasia followed by an injection of a fixative agent into the circulation. These so-called post-mortem perfusion protocols should in theory ensure an equivalent quality of tissue fixation, without exposing live animals to a procedure. Before adopting this new method, it is necessary to validate that sample quality is equivalent, ensuring the validity of scientific results. Here we performed a parallel comparison of several protocols of tissue fixation by ante-mortem or post-mortem transcardiac perfusion and measured the impact on the maintenance of axonal structures, dendritic spines and mitochondrial morphology. Our results showed that histological parameters show variable sensitivity to perfusion conditions and fixatives used. For instance, axon fragmentation and altered mitochondrial morphology were observed in post-mortem transcardiac perfusion groups. We furthermore determined that the fixation condition had a variable effect on immunostaining, impacting the detected expression level or pattern. Our results serve as a guide to orient the experimenter in selecting the best condition for optimal tissue fixation, which minimizes animal suffering while guaranteeing the integrity of the biological results obtained.

This study assesses various methods and timings for perfusion of fixatives to enhance brain histology, addressing ethical dilemmas associated with ante-mortem transcardiac perfusion and the lack of data on tissue quality using different techniques.

Main

The use of live animals in life sciences is a major societal challenge and source of debate within the scientific community. Research based on animals has led to major advances in knowledge in fundamental and biomedical sciences and is currently essential, despite the rapid development of alternative methods. Humane animal research adheres to the principles of the ‘3Rs’—reduce, replace and refine—first defined in 1959. This principle is embedded in the European legislation on the protection of animals used for scientific purposes (Directive 2010/63/EU). High standards of animal care and experimentation benefit not only the animals but also the researchers and staff who work with them, helping to reduce compassion fatigue. Furthermore, it is now well acknowledged that better animal welfare ultimately contributes to the quality and reliability of scientific results, as summarized in the expression ‘happy animals make good science’¹.

Refinement is defined to minimize pain, distress and harm from procedures on animals when these procedures cannot be avoided. This is often the case in neuroscience research, which involved nearly 849,000 animals—primarily mice and rats—in the European Union in 2020. These species are widely used because the complexity, organization and regulation of their neural networks closely resemble those of the human nervous system, enabling researchers to investigate fundamental brain processes that may be disrupted in neurodevelopmental, neuropsychiatric and neurodegenerative disorders. Because much of the function of the nervous system relies on the study of intercellular connectivity, histological preparations are a common practice. Tissue fixation is a critical step to prevent autolysis and degradation of tissues and their components, which occur rapidly after the death of the animal². Immersion in a solution containing a fixative agent is an efficient method for fixation of small elements such as the sciatic nerve³. However, this method is generally considered not appropriate for the fixation of larger tissues such as the whole brain of a rodent, because the penetration of fixative agents in the depth of tissue is too slow to prevent hypoxia and cellular changes in deeper brain structures. Transcardiac perfusion of a fixative agent is generally performed to clear blood and preserve deep cellular structures^4,5. Transcardiac perfusion is usually performed in deeply anesthetized animals as a terminal procedure during which the last heartbeats help to propel fixative into the general circulation. However, recent reports demonstrated that the choice of anesthetic and fixative agent can have a major impact on some cellular structures, such as mitochondria, with consequences for data analysis⁶. Furthermore, ante-mortem transcardiac perfusion has been recently challenged, and it is now proposed that the procedure could be performed post-mortem, ensuring similar tissue quality with less impact on the animal.

In this study, we compared the impact of different protocols for tissue fixation by ante-mortem or post-mortem transcardiac perfusion on several parameters that are routinely analyzed in neuroscience studies in the mouse brain. Following in utero cortical electroporations (IUCEs) of plasmids⁷, pups were randomized into four groups to compare two distinct protocols of ante-mortem transcardiac perfusion in anesthetized animals (with volatile and injectable anesthetic, respectively), one protocol of post-mortem transcardiac perfusion and one condition of direct immersion in the fixative agent without perfusion. Tissues were prepared in parallel, and we quantified tissue quality, axon structure, dendritic spine density and morphology, and dendritic mitochondria integrity. Furthermore, we tested distinct immunostainings. Overall, the distinct parameters in the study were unequally affected by fixation conditions. While some parameters were essentially similar regardless of the fixation condition (for example, axon branching density and dendritic spine density), other parameters were strongly variable from one group to the other (for example, fragmentation of axons, dendritic mitochondria integrity and phospho-CREB immunoreactivity). Overall, our results demonstrate that the choice of fixation condition can have a major impact on the quality of results and must be balanced with the impact to the animal to ensure robust results.

Results

Study design and fixation of tissues

To compare the influence of fixation condition on various parameters relevant for neurosciences, we performed IUCE with plasmids encoding the red fluorescent protein mScarlet-I, as well as the outer mitochondrial membrane-tagged green fluorescent protein (OMM-GFP) as a mitochondrial marker (Fig. 1a). At birth, we selected pups based on transcranial fluorescence observation, to keep only electroporated pups for the rest of the procedure. At postnatal day (P)21, the pups were randomly divided into four groups corresponding to the experimental conditions. Groups were balanced for sex, and pups from the same litters were distributed across different groups to minimize potential batch effects. All mice received an injection of buprenorphine between 30 min and 2 h before starting the procedure.

Fig. 1. Experimental design and overall quality of tissues.

a, Experimental design. Mice were split into four groups: group 1 (antemortem, isoflurane), group 2 (antemortem, ketamine/xylazine), group 3 (post-mortem) and group 4 (no perfusion). b, Duration of the procedure and perfusion time for each group. Data are the average of nine animals per group ± 95% CI. c, Detail (representative images) of confocal images acquired from coronal sections of brains for each group. Images in the green (anti-GFP) and red (mScarlet-I) channels. All images are taken with the same confocal settings (laser power and gain). Images were taken from the contralateral side (opposite to the electroporation side). d,e, Background signal intensity was quantified in the cortex in green (d) and in red (e) by drawing ROIs and measuring average signal intensity (arbitrary units (a.u.)). ROIs were drawn in regions devoid of specific signal (electroporated neurons). Bar graphs represent average ± s.e.m. Each circle on the graph represents the average of five ROIs taken from one image; one to four images were quantified per animal. In total, we quantified N = 10 (group 4) to 12 (groups 1, 2 and 3) images per condition. N (animals) = 5 for groups 1, 2 and 3, and N (animals) = 4 for group 4. Sex ratio was equilibrated in each group (male/female ratio 3:2 or 2:3; see details in Methods). Statistical analysis: one-way ANOVA with Dunnett’s multiple comparison. *P < 0.05. Leftmost schematics in a adapted from ref. ⁷ under a Creative Commons license CC BY-NC-ND 4.0. Elements in right schematics in a created in BioRender. Courchet, J. (2025) https://BioRender.com/xjf5hle.

Two groups (group 1 and group 2; Fig. 1a) corresponded to an ante-mortem transcardiac perfusion of fixative in anesthetized mice (either with volatile anesthesia for group 1 or with injectable anesthesia for group 2). Once a deep sedation state was constated (no reflexes to pinching, slow respiration), a thoracotomy exposed the heart in which first a volume of phosphate-buffered saline (PBS), then the fixative agent paraformaldehyde (PFA) were injected in the left ventricle. The last heartbeats immediately followed the fixative injection.

In comparison, group 3 corresponded to a post-mortem procedure in mice that were previously euthanized (Fig. 1a). We first performed a light sedation, followed by intravenous (IV) injection of heparin and injection of a euthanizing agent. Transcardiac perfusion was performed post-mortem, which we defined as a respiratory and cardiac arrest. Here, we used a peristaltic pump to force the PBS and then the fixative agent into the general circulation despite the lack of heartbeats.

In group 4, after an intraperitoneal (IP) injection of xylazine, mice were euthanized by pentobarbital and the brain was dipped into a fixative solution, without prior perfusion (Fig. 1a). This condition was designed as a reference to demonstrate the necessity of injecting the fixative agent into the general circulation to achieve good tissue quality. Therefore, groups 1–3 were expected to perform as well as or better than group 4 across all analyzed parameters.

We performed two rounds of IUCE to ensure the robustness of our results and to rule out any batch effects in the data. In total, five animals (groups 1–3) or four animals (group 4) were analyzed after PFA fixation.

Implementation of the protocol and general tissue quality

Volatile anesthesia was the fastest way to reach deep sedation (Fig. 1b). Injection-based methods took longer, during which mice were placed in a quiet box to allow sedation to take effect. When considering only the perfusion, the procedure was overall faster in groups 1 and 2, where heartbeats helped the delivery of the fixative agent. Although all groups were subject to at least one injection (buprenorphine before the procedure for proper analgesia), groups required different levels of handling of the animals, because each injection required some contention. The most handling and contention was required for animals from group 3, who in total received four injections.

For perfused animals (groups 1–3), rigor of the limbs and tail was indicative of a good fixation. The discoloration of the brain caused by blood removal was a further sign that perfusion was efficient (Supplementary Fig. 1), although we noticed some variability in the quantity of residual blood in group 3. In comparison, blood was still present in the brain of non-perfused animals.

Following the procedure, brains were dissected and post-fixed with 4% PFA overnight (Fig. 1a). We then performed coronal sections, immunostainings and confocal imaging. For each animal, we analyzed one to four brain sections, which we selected in the same brain regions. We first reasoned that perfusion quality might change background fluorescence of the tissue. Overall, background fluorescence was more important in the green channel, which is typically more prone to autofluorescence. Measurement of average fluorescence level in the cortex revealed, as expected, that non-perfused brains (group 4) had more background fluorescence than other conditions (Fig. 1c,d). Groups 1, 2 and 3 had similar levels of background fluorescence, although there was some variability between animals revealing potential heterogeneity in perfusion quality. On average, background signal in the red channel was much lower than in the green channel, so individual variations in background had a smaller overall impact (Fig. 1e).

Sub-optimal perfusion conditions are associated with axon fragmentation and can affect the quantification of axonal projections

We first sought to assess the impact of fixation conditions on the more fragile neuronal structures. Axons are long, thin structures prone to rapid degradation upon cellular stress. Neurons from the superficial layer of the somatosensory cortex typically project axons to cortical layers II/III and V of the ipsilateral and contralateral cortex (Fig. 2a). Because all animals came from the same electroporation procedure and were allotted randomly, our prediction was that any difference measured on average on axon branching patterns should be the result of the fixation condition. On ipsilateral layer V, the dense ramification of axons can be assessed by measuring the optical density in the red channel, corresponding to the fluorescence of mScarlet-I by electroporated neurons (Supplementary Fig. 2). Indeed, there were no statistically significant differences between groups in the axon density measured in the ipsilateral cortex after background fluorescence subtraction (one-way analysis of variance (ANOVA) test P = 0.385 for an effect of group), indicating that the signal-to-noise ratio was correct in the ipsilateral cortex (Fig. 2b). By contrast, we noted differences between groups in the contralateral hemisphere. The best results were observed with group 1, where there was a marked density of axon projections in the S1/S2 boundary in the somatosensory cortex, with axons that were well marked (Fig. 2c, left). In comparison, terminal axons density was reduced in groups 2 and 3 and strongly reduced in group 4. Interestingly, we noted that axons appeared more fragmented in groups 3 and 4, which may in part explain the apparent decrease of axon projection density (Fig. 2d). A two-way ANOVA test confirmed an overall effect of genotype (F(3,44) = 5.190, P = 0.0037) on axon density profiles. Multiple comparison assays revealed a decrease in the density of axons on cortical layers II/III and V for groups 2 and 3 as compared with group 1, despite a similar density of afferent axons in the white matter (Fig. 2e). The most pronounced reduction was observed in the non-perfused group (group 4).

Fig. 2. Effect of perfusion condition on axons.

a, Representation of a coronal brain section showing the soma of electroporated neurons (left side, ‘ipsilateral’) and axon projection and pattern of branching on the ipsilateral and contralateral cortex. The method of quantification of axon branching varies between the ipsilateral and contralateral side to adapt to the density of projections. Normalization to the density of white matter (WM) axons in the corpus callosum (CC) serves to compensate for potential differences in the number of electroporated neurons. b, Quantification of axon density in layer V of the ipsilateral cortex. Data represent the average signal value in five ROIs (a.u.) and normalized to signal intensity measured in the CC from the same image. Each circle on the graph represents the average value for one brain slice; two slices were analyzed per animal. N = 8 (group 4), 12 (group 3) or 14 (groups 1 and 2). Bar graphs represent the average ± s.e.m. Statistical analysis: one-way ANOVA (F(3,44) = 1.039, P = 0.385). c, Representative images of terminal axon branching on the contralateral side. I, II, III, IV, V and VI: cortical layer identity. d, Magnification (×4) shows axon fragmentation in the case of group 3, not group 1. e, Quantification of mScarlet-I intensity along an axis from white matter to layer I in the different groups. Data represent the average value for N = 6 (group 4), 13 (group 3) and 16 (groups 1 and 2) brain slices from four to five animals. N (animals) = 5 for groups 1, 2 and 3, and N (animals) = 4 for group 4. Sex ratio was equilibrated in each group (male/female ratio 3:2 or 2:2; see details in Methods). The curve represents the average fluorescence signal (a.u.) and s.e.m., for each percentile across a line from the white matter to the pial surface of the cortex. Statistical analysis: two-way ANOVA.

Fixation conditions had a moderate impact on dendritic spines density and morphology

Synapses are the focus of numerous studies owing to their central importance in brain function. In cortical pyramidal neurons, glutamatergic synapses are formed at the extremities of dendritic spines, whose density and shape dynamically changes, reflecting neuronal activity. We reasoned that sub-optimal tissue fixation could alter the detection of dendritic spines, making it harder to count or classify them. Moreover, dendritic spines are dynamic structures that could be rapidly altered post-mortem if fixation is not optimal. Thus, we imaged isolated basal dendritic segments from electroporated cortical neurons and manually analyzed dendritic spines (Fig. 3a,b). We first reasoned that fixation condition may affect our capacity to detect dendritic spines, or that slow fixation of synaptic structures could lead to dendritic spine shrinkage. Thus, we first compared dendritic spines density, measured from at least four basal dendritic segments from four individual neurons for each animal (Fig. 3c). After quantification, there was no overall effect of the group on the density of dendritic spines (one-way ANOVA F(4,72) = 0.8222, P = 0.4861) (Fig. 3d). It has been proposed that glutaraldehyde in combination with PFA contributes to better prepare tissues for synapses and mitochondria⁶. Thus, we compared results using an alternative 2% PFA–0.075% glutaraldehyde solution (PFA-GA) instead of 4% PFA (Fig. 3e). As before, groups were formed following two rounds of IUCE and then equilibrated (five animals in groups 1–3 and four animals in group 4). As for PFA, the density of dendritic spines was similar in all four groups upon fixation with PFA-GA (one-way ANOVA F(4,111) = 0.5654, P = 0.639) (Fig. 3f). Pairwise comparisons confirmed that no datasets were statistically different from one another. Interestingly, we noted a trend toward a global decrease of dendritic spines density in all conditions after fixation with PFA-GA, which probably comes from the overall increased autofluorescence of the tissue caused by glutaraldehyde⁸.

Fig. 3. Effect of perfusion condition on basal dendritic spines.

a,b, Schematic representation of the experimental design (a) and example of an optically isolated neuron, from which segments of basal dendrites could be isolated and analyzed (b). Spines were classified as immature or mature based on their morphology. c, Typical examples of dendritic segments from the different groups upon fixation with 4% PFA. d, Quantification of dendritic spines density in each group upon fixation with PFA. e, Typical examples of dendritic segments from the different groups upon fixation with PFA-GA. f, Quantification of dendritic spines density in each group upon fixation with PFA or PFA-GA. g,h, Quantification of mature dendritic spines expressed as a percentage of total spines upon fixation with PFA (g) or PFA-GA (h). All quantifications were performed blind to condition. N (animals) = 5 for groups 1, 2 and 3, and N (animals) = 4 for group 4. Sex ratio was equilibrated in each group (male/female ratio 3:2 or 2:3; see details in Methods). For each animal, we imaged and analyzed dendritic segments of four neurons. Bar graphs represent the average ± s.e.m. Each dot represents a neuron. **P < 0.01, ***P < 0.001. Elements in a created in BioRender. Courchet, J. (2025) https://BioRender.com/xjf5hle.

In parallel, we reasoned that the speed of tissued fixation might have a bigger impact on spine morphology, because plasticity happens on a faster time scale than spine addition or elimination^9,10. Dendritic spines were classified as mature (larger head, ‘mushroom’-shaped) or immature (filopodial or long-necked) and compared between groups (Fig. 3a,b). Although as expected from animals at this age, we observed on average a majority of mature spines in all conditions, a statistical analysis demonstrated a group effect for both the PFA (F(4,106) = 12.03, P < 0.0001) and PFA-GA (F(4,107) = 7.785, P < 0.0001) condition (Fig. 3g,h). Pairwise comparison revealed a trend toward more mature spines in group 2, which could be due to the dissociative effect of ketamine and its compound effect on NMDA receptors in excitatory and inhibitory cortical neurons^11,12. Conversely, mature spines were reduced in group 3, which could be caused by the overall longer anesthesia time before tissue fixation. Overall, our data reveal that tissue fixation condition may have a small, yet significant, effect on dendritic spines.

Post-mortem fixation is associated to altered morphology of mitochondria

Mitochondria undergo rapid cycles of fusion and fission in response to the cellular environment. Mitochondrial length is thought to correlate with their metabolic activity and furthermore, mitochondrial morphology is associated with pathological states¹³. Therefore, fixation conditions should preserve mitochondria morphology for accurate research on neuronal energy metabolism. Previous studies demonstrated that peri-mortem hypoxia has a major impact on the quality of mitochondria preparation⁶. Here, we measured the impact of fixation on dendritic mitochondria morphology. As a whole, mitochondria appeared more elongated in groups 1 and 2, whereas mitochondria were overall shorter in group 3, and especially in group 4 (Fig. 4a,b). Furthermore, we observed frequent swollen mitochondria, suggesting they were actively being fragmented at the time of fixation, in neurons from group 3 and 4 (Fig. 4c). Quantification of average mitochondria length per dendritic segment confirmed an overall effect of the experimental condition on mitochondria length in PFA-fixed animals (one-way ANOVA F(3,103) = 3.561, P = 0.017) (Fig. 4d). Pairwise comparison revealed a decrease of the average mitochondria length in group 3 and 4 neurons compared with group 1 or group 2. A similar result was obtained when considering the cumulative length of mitochondria regardless of the neuron or animal or origin (Fig. 4e). Interestingly, the overall same group effect was observed in the PFA-GA-fixed neurons (Fig. 4f,g). Compared with PFA-fixed conditions, the PFA-GA samples showed less data variability, although mitochondria were generally smaller in size—contrary to recent reports suggesting that glutaraldehyde fixation protects mitochondria from fragmentation⁶. Collectively, our data demonstrate that perfusion and tissue fixation have a major effect on the morphology of dendritic mitochondria.

Fig. 4. Effect of perfusion condition on dendritic mitochondria.

a,b, Representative images of dendritic mitochondria observed by confocal microscopy in each group, following perfusion/fixation with 4% PFA (a) and with PFA-GA (b). c, Examples of elongated, swollen and fragmented mitochondria observed in dendritic segments. d, Measurement of the average length of mitochondria per dendritic segment upon fixation with PFA. Data represent the average length of mitochondria in one dendritic segment. Bar graphs represent the average ± s.e.m. Each circle on the graph represents one dendritic segment. N (animals) = 5 for groups 1, 2 and 3, and N (animals) = 4 for group 4, in total 22–34 segments per group. Statistical analysis: one-way ANOVA (F(3,103) = 3.561, P = 0.017). e, Cumulative frequency distribution of mitochondria size per group, and average mitochondria length upon fixation with PFA. Inset: the average length of all mitochondria (average ± s.e.m.). N = 354–525 mitochondria. Statistical analysis: one-way ANOVA (F(3, 1665) = 6.984, P < 0.0001). f, Measurement of the average length of mitochondria per dendritic segment upon fixation with PFA-GA. Data represent the average length of mitochondria in one dendritic segment. Bar graphs represent the average ± s.e.m. Each circle on the graph represents one dendritic segment. N (animals) = 5 for groups 1, 2 and 3, and N (animals) = 4 for group 4, in total 18–33 segments per group. Statistical analysis: one-way ANOVA (F(3,97) = 3.552, P = 0.017). g, Cumulative frequency distribution of mitochondria size per group, and average mitochondria length upon fixation with PFA-GA. Inset: the average length of all mitochondria (average ± s.e.m.). N = 240–562 mitochondria. Statistical analysis: one-way ANOVA (F(3,1828) = 24.8, P < 0.001). *P < 0.05; **P < 0.01.

Immunochemistry is known to depend upon fixation quality

Finally, we wanted to see the impact of the fixation condition on histological preparations. For this, we used various classic antibodies that, based on our experience, may vary in sensitivity to tissue fixation conditions. We first tested whether distinct fixation condition could impact cell integrity, which could alter the quality of tissues and lead to artifacts. We then evaluated apoptosis, an intrinsic program leading to cell death, which occurs during brain development and in pathological contexts (neurodegenerative diseases)¹⁴. For this, we performed cleaved caspase-3 immunostaining (Fig. 5a and Supplementary Fig. 3a). We observed a few apoptotic cells in the histological sections, especially in the hippocampus, demonstrating that the cleaved caspase-3 staining was effective (Supplementary Fig. 4). However, no notable differences were observed between experimental groups, ruling out that the initiation of a stress response led to apoptotic cell death—probably because the fixation time was relatively short.

Fig. 5. Effect of perfusion condition on axons on immunostaining.

a, Cell apoptosis observed by cleaved caspase-3 staining (red) in each group, following perfusion/fixation with 4% PFA. Magnification images of the hippocampus, showing positive (apoptotic) cells provided in Supplementary Fig. 4. b, Ratio between pCREB (Ser133) and non-phosphorylated CREB in each group. Images are colorized on a rainbow Look Up Table (LUT; purple: low ratio, red: high ratio). Magnification of the cortex and hippocampus. Raw images for CREB and pCREB are provided in Supplementary Fig. 5.

We next examined Ca/cAMP-responsive element binding protein (CREB), a transcription factor involved in processes underlying neuronal plasticity leading to learning and memory. CREB is a so-called immediate early gene, whose rapid phosphorylation on serine 133 and activation occurs upon elevation of cAMP or Ca²⁺ in the cytoplasm, and this is usually correlated with the long-term memory or after a psychophysical stress^15–17. For this staining, we observed marked differences between groups. Groups 1 and 2 showed some phosphorylated CREB (pCREB) immunoreactivity, corresponding to expected patterns in the mouse brain, with nuclear signal observed in several brain regions—strong in the cerebral cortex and hippocampus, and strongest in the dentate gyrus. (Fig. 5b and Supplementary Fig. 3b). pCREB signal was overall stronger in group 2 than in group 1, and the pCREB/CREB ratio was slightly different in some brain regions between the two groups (see, for example, the cingulate cortex or the CA3 region of the hippocampus), which may correspond to the effect of distinct anesthetic agents. By contrast, the pCREB signal was strongly reduced in group 3, which presented the lowest signal overall. In group 4, some pCREB signal—comparable to group 2—was observed in the superficial part of the cortex, but the signal diminished rapidly in deeper brain regions, most likely due to the slow penetration of fixative into the tissue. Observations deeper in the tissue in areas of the brain known to be involved in memory such as the hippocampus (especially the dentate gyrus) demonstrated a profound loss of pCREB staining, which was weak or even non-existent in groups 3 and 4 compared with groups 1 and 2 (Fig. 5b). Although we cannot determine from these data whether the fixation conditions caused artefactual loss of CREB activity, whether the immunostaining was affected by sample preparation, or both, we conclude that the fixation protocols used in groups 3 and 4 were not suitable for the study of CREB staining.

Discussion

In this study, we compared three methods for brain sample preparation. Methods 1 and 2 rely on the ante-mortem injection of a fixative agent in the general circulation by transcardiac perfusion in an anesthetized animal. Method 3 relies on the post-mortem delivery of a fixative agent using a peristaltic pump to distribute the fixative in the circulation of an animal that has been put to sleep. A fourth group, where brains were fixed by simple immersion in the fixative solution, served as a reference to measure objective benefits of methods 1–3. In groups 1–3, the overall quality of perfusion was deemed comparable, although we observed some inter-animal variability in the post-mortem group. Fixative distribution and tissue penetration are believed to be enhanced by ongoing heartbeats, which justifies the classical use of ante-mortem perfusion procedures. Thus, we can presume that the lack of heartbeats could contribute to the variability in perfusion efficiency in group 3. However, we cannot rule out that this difference stems from the use of a peristaltic pump in group 3 and not in groups 1 and 2, although we adopted comparable flow rates and injection pressure.

Brains were subsequently processed, and we analyzed parameters that would be typical for such an experiment, such as axon projections, dendritic spine density and morphology, mitochondrial morphology and immunostaining. Importantly, our study demonstrates that the method of tissue fixation has a major impact on experimental parameters, warranting caution when comparing results that were not obtained with the same method. Furthermore, there was not clearly one condition that gave the best results for each and every analysis parameter. Rather, our results suggest that tissue fixation and sample preparation must be adapted to the needs of the experiment that will be performed afterward. As a general rule, the more dynamic a biological process is, the more sensitive it is to a fixation protocol. This highlights the importance of predefined analysis parameters, as well as pilot studies, to ensure that the most valid scientific information be obtained from each experiment.

It is assumed that brain structures deteriorate within minutes following the animal’s death. Accordingly, our results demonstrate that the method of fixation influences the axonal structure and is critical for preserving mitochondrial morphology. Although the causes of these changes are multiple, one possible explanation could be the spreading depolarization waves and excitotoxicity due to neurotransmitter release¹⁸, caused by the hypoxia consecutive to cardiac arrest. In fact, isoflurane and ketamine—but not pentobarbital—have been suggested to offer protection against spreading depolarization^19,20, which may explain some of the differences observed between groups 1 and 2 and groups 3 and 4. Moreover, mitochondria respond rapidly to sudden changes in cerebral oxygenation, and even brief periods of post-mortem hypoxia lead to mitochondrial fragmentation, as demonstrated in a recent publication⁶. These peri-mortem alterations drastically change the baseline of the measured parameter, and although the implication for data collection might be minimal when comparing a control and experimental condition prepared in the same way, these changes are important to consider when switching from one method of tissue fixation to another. This would, for example, complicate the interpretation of longitudinal studies or comparisons with past experiments and may warrant redoing some previous experiments to generate results in comparable situations. Furthermore, axon fragmentation and smaller-sized mitochondria in the control condition may make it more difficult to detect subtle phenotypes in the experimental condition, leading to false negative results.

On the contrary, our results show no impact on dendritic spine maturity. One limitation of our study is that it was carried out on in one strain of mouse, looking at one specific brain region and one specific age (P21). We do not rule out the possibility that different results could be observed at time points when synaptic plasticity is more prominent, or that other brain regions might also be affected. However, tissue quality can (1) create difficulties during confocal microscopy acquisition and (2) interfere with analysis, for example, when excessive background noise is present. Note that our study was carried out with wild-type animals in which dendrites are normally formed. Alterations in the morphology and/or dynamics of dendritic are often associated with neurodegenerative or mental retardation diseases such as fragile X syndrome²¹. In such pathological models, more variability between animals could have a higher impact and increase bias when interpreting results.

It is often assumed that blood removal and perfusion serve the purpose of removing unwanted autofluorescence. Accordingly, we observed increased background fluorescence in group 4, although not to an extent that rendered the imaging uninterpretable. Importantly, although brains from group 3 animals were generally correctly perfused, there were some regions in the brain where small blood vessels remained visible, indicating that fixative agent distribution using a peristaltic pump may introduce some variability from one tissue to the other. Although this was generally not consequential, poorly perfused regions sometimes interfered with the analysis, such as for the visualization of CREB activity (Fig. 5). Here, we chose to electroporate neurons with a plasmid encoding the bright red fluorescent protein mScarlet-I, which is among the brightest fluorescent proteins available. mScarlet-I has several advantages: it provides clear visualization of cellular structures, remains bright after fixation and does not require immunoamplification, which can increase tissue autofluorescence. Furthermore, the red channel is usually less impacted by autofluorescence than the green channel. Finally, we imaged dendritic spines with a sensitive Nikon A1 confocal microscope and GaASP detectors. In these optimal imaging conditions, the signal-to-noise ratio is optimal and potential differences in autofluorescence may have a lesser impact. Experimenters need to adapt their choice of fluorescent protein and microscope setup and verify that, under their specific conditions, the signal can be reliably distinguished from autofluorescence before selecting a tissue fixation method.

Handling time varied substantially between groups, with differences in how many times the animal had to be manipulated, which can be a source of stress. Also, the quality of sedation was variable between the four experimental groups. In our experience, more variability was observed with injectable anesthesia in group 2. Moreover, the delay between the beginning of the procedure and the arrival of the fixing agent in the brain could be very important (especially in group 3). This had a direct impact on the duration of the perfusion per animal and therefore on its well-being, but it also affects the experimenter. As a typical experiment usually involves the preparation of a whole litter of animals, it is not uncommon to prepare 10–12 animals at the same time. The increased manipulation and overall preparation time impose a stress on the experimenter that should not be disregarded.

The justification for using live animals in experimental neuroscience is that the advancement of knowledge (that is, the benefit of the experiment) outweighs the ethical costs involved. Aside from the 3R principle from Russel and Burch, it is equally important to generate robust data. Complementary ‘R’ words, including ‘reproducibility’ or ‘robustness’ have been proposed in a broader ‘6R’ principle for animal experimentation²². Our study falls in line with this movement, by guiding the community toward choosing the best fixation condition. Although we do not claim that one method is inherently better than the others, we demonstrate that the choice of fixation method must consider the specific parameters to be analyzed, the well-being of the experimenter, and the minimization of animal suffering. Thus, our results strongly argue against a one-size-fit-all method that could be uniformly applied to all experiments. Future studies may address other important aspects of this procedure, for example, by defining early (and easy to implement) criteria for detecting death, which will make it possible to consider earlier post-mortem perfusions.

Methods

Animals

Experiments were carried in accordance with the French legislation regarding animal experimentation. Experimenters have FELASA (Federation of European Laboratory Animal Science Associations) level C (‘conceptor’) certification and validated the mandatory continuous training in laboratory animal science in accordance with the EU Directive 2010/63. They also passed a certifying training in surgery. The experimental protocol was approved by the ACCeS Ethics Committee and authorized by the French Ministry of Higher Education and Research under the reference APAFIS #42003-2023012012569139. Four pregnant Swiss females were purchased from Janvier Labs. They were transported to the animal facility at embryonic day (E)7.5 and were allowed ad libitum access to food and water and maintained on a 12-h light–dark cycle. IUCE was performed at E15.5.

DNA and plasmids

Endotoxin-free plasmid DNA was obtained using the Macherey Nagel midi-prep kit, according to the manufacturer’s instructions. The pCAG-mScarlet-I was described previously²³. The pCAG-OMM-GFP, in which GFP is fused to a fragment of the ActA protein to target the outer mitochondrial membrane, was kindly provided by Tommy L. Lewis (Oklahoma Medical Research Foundation, USA).

In utero cortical plasmid electroporation

In utero cortical plasmid electroporation were performed at E15.5 as described in ref. ⁷. A mix containing 1 µg µl⁻¹ of pCAG-mScarlet-I (to visualize neurons), 0.25 µg µl⁻¹ pCAG-OMM-GFP (to visualize mitochondria) and 0.5% Fast Green (Sigma; 1/20 ratio) was injected into one lateral hemisphere of embryos using microcapillaries. Electroporation was performed using an ECM 830 electroporator (BTX) with 3 × 5 mm GenePad electrodes, applying four pulses of 45 V for 50 ms each, with 500ms interval, to target cortical progenitors. Analgesia was achieved by pre-operative injection of Buprenorphine (0.1 mg kg⁻¹ of body weight). Post-surgical pain was monitored for 2 days using a standardized scale approved by the ethical committee, and analgesia was provided accordingly using Buprenorphine (0.1 mg kg⁻¹ of body weight) for 2 days. Delivery typically occurred at gestational age E19. At P2, we verified the efficiency of electroporation with a Xite Fluorescence flashlight system (Nightsea, distributed by Electron Microscopy Sciences). Non-electroporated pups were culled to ensure that no animal was subjected to the procedure unnecessarily.

Group composition and sample size

We performed two rounds of in utero electroporation. On the first round, the electroporation was performed on two pregnant dams. Sixteen pups were conserved and allocated to 4 experimental groups in order to mix litter and male/female ratio in each group. On the second round of electroporation, the electroporation was performed on two additional pregnant dams. Twenty-two pups were conserved and allocated at P21 to groups 1–3 (6 animals per group) or group 4 (4 animals). At P21, animals were euthanized using the appropriate procedure for each group, with either PFA or PFA-GA as a fixative. In total, experimental groups were composed of ten animals (five with PFA and five with PFA-GA) for groups 1–3, and eight animals (four with PFA and four with PFA-GA) for group 4.

Group composition and sex ratio:

Group 1-PFA: two males, three females, Group 2-PFA: two males, three females, Group 3-PFA: three males, two females, Group 4-PFA: two males, two females.

Group 1-PFA-GA: two males, three females, Group 2-PFA-GA: two males, three females, Group 3-PFA-GA: two males, three females, Group 4-PFA-GA: two males, two females.

Tissue fixation method

All perfusions (groups 1, 2 and 3) were performed by the same experimenter. All mice received premedication/analgesia: Buprenorphine (Buprenex) was diluted to 0.03 mg ml⁻¹ and administered by subcutaneous (SC) injection at a final dose of 0.1 mg kg⁻¹, given at least 30 min to 2 h before euthanasia.

In this study, we compared four tissue fixation methods:

- Groups 1 and 2: ante-mortem transcardiac perfusion

The mice were anesthetized by volatile anesthesia (isoflurane 5%, O₂ 1.5 l min⁻¹) (group 1) or by fixed anesthesia by an IP injection of ketamine (150 mg kg⁻¹)/xylazine (15 mg kg⁻¹) (group 2). Once deep sedation was achieved (assessed by the loss of reflex to toe pinching and slowing of the respiration rate), mice were placed on dorsal decubitus. For group 1, mice were maintained under anesthesia under a mask (isoflurane 5%, O₂ 1.5 l min⁻¹). A large incision of the skin revealed the ribcage, which was incised using scissors to expose the heart. The right atrium was incised, then mice were euthanized by intracardiac perfusion of 1× PBS (total volume 5 ml), followed by injection of 4% PFA/1× PBS (total volume 10 ml) (PFA). The flow rate was not controlled and depended on the surgeon’s experience; however, we estimate that a 5 ml intracardiac injection takes between 30 s and 1 min, corresponding to a flow rate of 5–10 ml min⁻¹. We used a 26G needle for the injection.

- Group 3: post-mortem transcardiac perfusion

The mice were anesthetized by IP injection of xylazine (15 mg kg⁻¹) to induce sedation. Five minutes later, we performed an IV injection of heparin (500 UI kg⁻¹, 10 µl) in the caudal vein. Then, the mice were euthanized by IP injection of a veterinary approved euthanizing agent (pentobarbital, 140 mg kg⁻¹). Death was defined by the absence of respiratory movement and absence of reflex to toe pinching. As quickly as possible after death was confirmed (5–10 min after injection), mice were placed in the dorsal decubitus position. A large incision of the skin revealed the ribcage, which was incised using scissors to expose the heart. The right atrium was incised, then 1× PBS was injected in the heart using a peristaltic pump (speed was estimated to 10 ml min⁻¹) for 1 min. Immediately afterward, a PFA solution was injected using the peristaltic pump at an estimated speed of 10 ml min⁻¹ for approximately 3 min. We used a 26G needle for the injection.

- Group 4: no perfusion

This group is a negative control to verify that the fixative injections of groups 1–3 provide better-quality tissue.

The mice were euthanized by an IP injection of xylazine (15 mg kg⁻¹) to induce sedation, and 5 min later an IP injection of veterinary euthanizing agent (pentobarbital, 140 mg kg⁻¹). Death was characterized by the absence of respiratory movement and absence of reflex to toe pinching. As quickly as possible after death was confirmed (within 5–10 min), the mice brains were extracted and immersed in a fixative solution (PFA).

For all groups, a second series of animals underwent the exact same procedure, but with the PFA solution replaced by 2% PFA-GA.

All whole brains are post-fixed overnight in PFA or in PFA-GA at 4 °C, protected from light.

Brain slicing and immunostaining

Following fixation and 1× PBS washing (three times for 10 min), brains were embedded in 3% low-melt agarose in 1× PBS. Sections measuring 80 µm in thickness were prepared using a Leica VT1000S vibratome, with two slices focusing on the corpus callosum and two on the hippocampus per animal. Slices were then incubated overnight at 4 °C under agitation with primary antibody polyclonal chicken anti-GFP (Rockland) diluted at 1/2,000 in permeabilization buffer (PB: 5% BSA, 0.1% Triton X-100 and 1× PBS), protecting from light. The next day, slices were washed three times for 10 min in 1× PBS under agitation and then incubated with secondary antibodies (polyclonal goat anti-chicken, Alexa 488, 1/2,000 (Life Technology). Nuclear DNA was stained using Hoechst 33258 1/5,000 (Pierce) 1 h at room temperature. After three washes in 1× PBS, for 10 min at room temperature under agitation, slices were mounted on slides with Fluoromount G (Invitrogen) and kept at 4 °C.

Image acquisition and analyses

Axonal projections

Confocal images were acquired in 1,024 × 1,024 mode with a Nikon Ti-I microscope equipped with a C2 or an A1 laser scanning confocal microscope.

Microscope control and image analysis was performed using the Nikon software NIS-Elements (Nikon). We used the following objective lenses (Nikon): 10× PlanApo; numerical aperture (NA) 0.45, 20× PlanApo VC; NA 0.75. The parameters (laser power and detector gain) were set during the first series of images to avoid signal saturation and were kept constant within that series to ensure comparable signal intensity across slices.

To quantify axon density on the ipsilateral layer V, five regions of interest (ROIs) were placed in cortical layer V (mScarlet-I signal), in the corpus callosum (for normalization) and in the cortex (background fluorescence). Importantly, all ROIs were placed on the same image to account for potential variation in imaging parameter from one image to the other. Signal was defined as (average (layer V) – average (background))/(average (corpus callosum) – average (background)).

On the contralateral side (terminal axon branching), we measured the mScarlet-I fluorescence intensity of axonal projections across the six cortical layers and normalized these values to the mScarlet-I intensity in the white matter axons, which correlates with electroporation efficiency. We used an Excel spreadsheet to perform an automated background subtraction and binning of data along the ventricular zone to pial axis (from 0% to 100%), as described previously²³.

Mitochondria and dendritic spines

Images were acquired with Nikon Ti-I microscope equipped with a C2 or a A1 laser scanning confocal microscope. High-resolution images of dendritic spines and mitochondria from pCAG-mScarlet-I/pCAG-GFP-OMM mix electroporated neurons were acquired using the 100× (NA 1.45, zoom 1.5, resolution 2,048 × 2,048 pixels). Z stacks (z-step 0.15 μm) of digital images defined from top to bottom were captured using the Nikon software NIS-Elements (Nikon). For the figures, Z stacks were collapsed into a single image using the maximum intensity projection function in Fiji (ImageJ) software. Analysis of mitochondrial length and occupancy as well as dendritic spine density and morphology was performed using Fiji. Spines were categorized on morphological bases according to categories classically defined in the literature. Filopodial-like and thin spines (with a long, thin neck and lacking a bulbous head) were considered immature, whereas mushroom-shaped spines (characterized by a well-defined bulbous head and a narrow neck) were considered mature^24,25. Based on these morphological criteria, the number of dendritic spine subtypes was quantified on secondary segments of basal dendrites. Dendritic spine density is displayed as a ratio normalized to 10 µm of dendritic length.

Quantifications and statistical analyses

Statistical analyses were performed using GraphPad Prism (version 10.1.1). The parameters measured (axonal fragmentation, atrophy of synapses and fragmentation of mitochondria) were compared by the statistical assay indicated in the figure legends. As a rule, we used a Student’s t-test to compare two experimental groups, an ANOVA for three or more groups, and non-parametric tests when the assumptions of normality and equal variance were not met.

Quantifications were performed blind to experimental condition. Brains were identified using a letter–number code by the experimenter at the time of the procedure, and this information was not available to the person performing quantifications.

Statistical significance is indicated in all figures by the following annotations: *P < 0.05, **P < 0.01, ***P < 0.001.

Reporting summary

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

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at 10.1038/s41684-025-01633-1.

Author contributions

G.M.-D. and J.C. conceived of the study and performed IUCEs and tissue preparation. G.M.-D. prepared tissues and histological samples and quantified and interpreted experiments for axonal preparations and immunohistochemistry. S.E. quantified and interpreted experiments for mitochondria morphology and dendritic spine density and morphology. O.R. provided critical insight on study design, and participated in immunohistochemistry analysis with G.M.-D. J.C. supervised the work. J.C. and G.M.-D. drafted the paper, and all other authors edited and approved the paper.

Peer review

Peer review information

Lab Animal thanks Christoph Harms and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41684-025-01633-1.