Quantification Of Neurite Degeneration With Enhanced Accuracy And Efficiency in An In Vitro Model Of Parkinson’s Disease Part 2

Results

Establishment of a cell culture model of neurite degeneration associated with PD

To develop a method for investigating neurite degeneration in a cell culture model of PD, we used cultured LUHMES cells, a population of conditionally immortalized, human mesencephalic cells. Upon differentiation, LUHMES cells become postmitotic, adopt a gene expression profile characteristic of mature dopaminergic neurons, and develop a neuronal morphology with neurites that extend up to 500 mm in length (Fig. 1A; Lotharius et al., 2005; Scholz et al., 2011; Kraemer et al., 2021). 

Neurodegeneration, a condition in which nerve cells mutate or die, is closely related to memory. In cell culture models, we can control the culture conditions to alleviate the expression of neurodegeneration and thus improve the performance of memory.

Studies have shown that the homeostasis of nerve cells plays an important role in cell survival and system function. We can improve the stability of cells by controlling factors such as nutrients and growth factors in the culture medium. At the same time, pay attention to maintaining the cleanliness and temperature control of the culture environment to prevent cells from being disturbed and damaged by the environment.

In addition, in the cell culture model, we can also inject certain compounds or drugs into the cells to reduce the expression of neuromodulation, thereby improving the survival rate and memory performance of the cells. This technique is also widely used in the study of neurodegenerative diseases.

In short, neurodegeneration has an important impact on cell stability and system function, and cell culture technology is a good model research method, which can conduct in-depth exploration and research on the relationship between neurodegeneration and memory. We can alleviate neuromodulation performance by cultivating environmental control, injecting drugs into cells, etc., to improve cell survival and memory performance and provide a more comprehensive basis for the study of neurodegenerative diseases. It can be seen that we need to improve our memory. Cistanche can significantly improve our memory, because Cistanche can also regulate the balance of neurotransmitters, such as increasing the level of acetylcholine and growth factors, which are very important for memory and learning. In addition, meat can also improve blood flow and promote oxygen delivery, which can ensure that the brain receives sufficient nutrition and energy, thereby improving the vitality and endurance of the brain.

Click know supplements to improve memory

Because of the superior optical qualities of glass compared with plastic, the cultures were established on glass slides containing eight media chambers. Based on previous reports (Lotharius et al., 2005; Scholz et al., 2011; Kurowska et al., 2014), we initially attempted to establish cultures on chamber slides coated with PLO and fibronectin; however, such cultures exhibited adhesion deficits and low viability. As a result, we tested whether coating with additional substrates would improve the viability of LUHMES cell cultures established on glass surfaces. 

LUHMES cell cultures were established on slides that were either uncoated or coated with PLO and fibronectin, PDL and laminin, or a combination of all four substrates. Analysis of viable nuclei by fluorescence microscopy revealed that slide coating significantly affected the viability of the cultures [Friedman x2 (3) = 18, p, 0.0001], with post hoc multiple comparison analyses revealing that slides coated with a combination of all four aforementioned substrates yield a significantly higher rate of viable cells when compared with slides only coated with PLO and fibronectin (p= 0.0437; Fig. 1B, C).

After identifying the optimal coating substrates for culturing LUHMES cells on glass slides, we evaluated the effects of cell density on the suitability of the cultures for neurite degeneration measurements. LUHMES cells were cultured in a differentiation medium with plating densities of 57,500 cells per well, 115,000 cells per well, or 230,000 cells per well, and on their fifth day of differentiation, the cells were fixed with 4% PFA and imaged by phase-contrast microscopy. 

A density of 57,500 cells per well resulted in cultures with poorly networked neurites and irregular neurite morphology, while cultures established with a plating density of 230,000 cells, although viable, contained neurite densities too high for accurate neurite measurement. The intermediate cell density of 115,000 cells per well provided high viability and well-separated axon tracts conducive to accurate neurite degeneration measurements (Fig. 1D).

To model PD-associated neurite degeneration in vitro, we exposed differentiated LUHMES cells to 6-OHDA, a widely used neurotoxin that promotes oxidative stress and degeneration of catecholaminergic neurons (Bové and Perier, 2012). Exposure of differentiated LUHMES cells to 6-OHDA promoted a dose-dependent increase in neurite degeneration, with 5.0 mM 6-OHDA and 7.5 mM 6-OHDA inducing moderate and severe neurite degeneration, respectively (Fig. 2F). To quantify neurite degeneration in cultures of LUHMES cells exposed to 6-OHDA, we used a widely cited method in which phase-contrast images of cultures are binarized, cell bodies in the images are digitally removed, and the particle analyzer algorithm of ImageJ is applied to measure the area of neurite fragments. The fragmented neurite areas are then summed and divided by the total neurite area to determine a DI (Fig. 2A–E; Sasaki et al., 2009; Kraemer et al., 2014; Di Stefano et al., 2015; Hill et al., 2018; Geisler et al., 2019). 

As expected, exposure of differentiated LUHMES cells to 6- OHDA caused a dose-dependent increase in the DI (F(3,24)= 119.8, p = 0.0001, R2 = 0.9374; Fig. 2G). However, during the analyses of cells exposed to 6-OHDA, we observed several issues that appeared to negatively impact the accuracy and efficiency of DI measurements. Thus, we sought to optimize this common method for performing DI measurements, which we henceforth refer to as the traditional DI method, to generate a protocol facilitating more effective measurement of neurite fragmentation.

Reducing artificial fragmentation associated with micrograph binarization

While using the traditional DI method, we frequently observed an issue in which intact neurites in phase-contrast images appeared fragmented in their corresponding binarized image (Fig. 3A). This artificial fragmentation primarily affected neurites with low prominence. Thus, we investigated whether the issue could be reduced by procedures that enhance image contrast. To improve image contrast, we first captured images of neurites using a variety of light intensities to determine the level of image brightness that would produce optimal contrast. 

Images captured with the optimal light intensity were then subjected to further contrast enhancement by using ImageJ to apply a LUT with minimum and maximum intensity values of 90 and 205, respectively. Incorporation of these contrast enhancements provided a moderately more accurate representation of the neurites in binarized images generated from phase-contrast images (Fig. 3B). Nevertheless, quantitative analysis of binarized images of healthy neurites indicated that the contrast enhancements did not significantly decrease artificial fragmentation (p = 0.9999; Fig. 3D). Thus, we next investigated whether use of immunofluorescence imaging would produce images with superior contrast and thereby facilitate production of more accurate binarized neurite images with reduced artificial fragmentation. 

The slides described in Figure 2, featuring fixed LUHMES cells exposed to various concentrations of 6- OHDA, were immunostained for the cytoskeletal protein b III-tubulin, imaged using fluorescence microscopy, and analyzed by a blinded investigator for artificial fragmentation (Fig. 3C). Quantitative analysis of binarized images obtained from fluorescence micrographs and corresponding phase-contrast micrographs revealed that the type of microscopy affected levels of artificial fragmentation [Friedman x2 (3) = 15.44, p, 0.0001]. Specifically, fluorescence micrographs produced binarized images with significantly reduced artificial fragmentation compared with those from phase-contrast micrographs (p = 0.0016; Fig. 3D). 

As a result, images of healthy cultures yielded significantly lower DI scores when fluorescence micrographs were used rather than phase-contrast micrographs (F(1,8)= 131.4, p = 0.0001, R2 = 0.1442; Fig. 3E). Moreover, a notable overall trend is that the fold change between mean DI values associated with healthy (vehicle-treated) cultures and cultures exhibiting neurite degeneration (5.0 or 7.5 mM 6-OHDA-treated) were greater when measured using fluorescence micrographs than when obtained from phase contrast micrographs (Fig. 3E). This occurred despite fluorescence micrographs of cultures with moderate neurite degeneration (induced by exposure to 5 mM 6-OHDA) having significantly lower DI scores than corresponding phase images because of the reduction in artificial fragmentation (p = 0.0001, d = 0.6312).

Optimizing parameters for fragmented neurite and total neurite detection
The traditional method for performing DI measurements involves the use of the Analyze Particles plugin for ImageJ with size parameters of 20
10,000 pixels to detect neurite fragments. However, we observed that a large proportion of neurite fragments were undetected by the particle analyzer when using these widely cited parameters (Shin et al., 2012; Di Stefano et al., 2015; Sasaki et al., 2016; Loreto et al., 2020; Shin and Cho, 2020) to measure neurite degeneration in phase-contrast micrographs of LUHMES cells exposed to 6-OHDA, generated as described in Figure 2. 

Thus, we reanalyzed images from that dataset using different minimum size limits for particle detection to determine the optimal parameters for the identification of neurite fragments. From qualitative assessment of binarized images generated from phase-contrast micrographs, size parameters of 5–10,000 pixels enhanced the detection of neurite fragments, but were also associated with the misidentification of background noise, image artifacts, or other small non-neurite material as neurite fragments. Compared with the widely used parameters of 20–10,000 pixels, however, parameters of 10–10,000 pixels enhanced the detection of neurite fragments without a noticeably large increase in false positives (Fig. 4A, left). 

Similar analyses were also performed with binarized images obtained from fluorescence micrographs of LUHMES cells immunolabeled for b III-tubulin. Particle analyzer size parameters of 5–10,000 pixels facilitated the most sensitive detection of neurite fragments and, since the fluorescence micrographs featured less small debris, such parameters did not noticeably increase the misidentification of non-neurite material as neurite fragments (Fig. 4A, right).

The traditional method for performing DI measurements, by using particle analyzer size parameters of 20– 10,000 pixels, excludes objects smaller than 20 pixels from measurements of neurite fragments. 

However, such small objects are typically erroneously included in the total neurite area measurements, since the traditional DI measurement method involves calculating total neurite area by summing all black pixels in the binarized images. To overcome this issue, we used the Particle Remover plugin for ImageJ, written by ImageJ developer Wayne Rasband, to remove small, non-neurite objects from binarized images. Application of this plugin while performing DI measurements using phase-contrast micrographs of LUHMES cells exposed to 6-OHDA, generated as described in Figure 2, resulted in a significant increase in DI measurements across all concentrations of neurotoxin exposure (F(1,8) = 83.43, p = 0.0001, R2 = 0.0416).
Furthermore, a significant interaction indicated that the increase in DI caused by the use of the particle remover became stronger at higher concentrations (F(1.881,15.05) = 59.97, p = 0.0001, R2 = 0.0094; Fig. 4B, C). We also imaged the same slides of cells by fluorescence microscopy and assessed the effects of small particle removal on DI measurements obtained from fluorescence micrographs. While small particle removal provides DI values that are theoretically more accurate, such values on average did not significantly differ from those obtained without the use of the particle remover plugin when measured using fluorescence micrographs (F(1,8) = 2.613, p = 0.1446, R2 = 0.0013; Fig. 4D).

Automation of DI measurements from fluorescence micrographs

Performance of DI measurements using the traditional method requires tedious and time-consuming image processing, use of the freehand tool to digitally remove individual cell bodies and measurement of fragmented neurite and total neurite areas from each image. Thus, we sought to enhance the efficiency of the method by writing a macro that fully automates DI analyses. To automate soma removal, the macro, titled ANDI, executes the binarization of images of cell nuclei, labeled via the nuclear stain DAPI, followed by modification of the binarized image such that thresholded nuclear regions are enlarged to encompass the entire region of the soma. Such images with binarized and enlarged nuclei are then subtracted from corresponding fluorescence micrographs of neurons immunolabeled for b III-tubulin, generating an image exclusively featuring neurites (Fig. 5A, B). 

To adjust for variable nuclear sizes caused by nuclear shrinkage or fragmentation associated with neurodegeneration, the binarized images of nuclei are enlarged via a combination of dilations and erosions that combine nuclear fragments and expand binarized regions to occupy an area larger than the somas of differentiated LUHMES cells. Images exclusively containing neurites are then binarized and subjected to DI analysis via a series of automated functions (Fig. 5C, D). To evaluate the accuracy of the macro operations associated with soma removal from images, we compared the DI scores obtained following the manual removal of cell bodies from fluorescence micrographs using the freehand tool of ImageJ (as previously depicted in Fig. 4D), to DI scores generated from analyses of identical images following automated cell body removal using the soma removal operations featured in the ANDI Macro. Our analyses revealed that DI scores calculated using automated cell body removal correlated nearly identically with DI measurements obtained from images subjected to manual soma removal (r(34) = 0.980, p, 0.0001; Fig. 5E).

To increase the accuracy of the analyses, we incorporated several of the previously described adjustments that enhance DI measurements into ANDI. The macro features contrast enhancement operations, use of the Particle Remover plugin to remove non-neurite matter from neurite images and non-nuclear matter from images of nuclei, and detection of neurite fragments using size parameters of 5–10,000 for the Analyze Particles plugin. 

We hypothesized that, collectively, the use of fluorescence microscopy, contrast enhancement operations, removal of small objects with the particle remover, and optimized parameters for particle detection would contribute to more accurate measurements of neurite degeneration. To test this hypothesis, the phase-contrast images of LUHMES cells exposed to various concentrations of 6-OHDA, generated as described in Figure 2, were analyzed by blind and subjective scoring using a value range between 0.0 and 1.0, with values representing the apparent proportion of fragmented neurite area. We then evaluated the correlation between the subjective scores and DI measurements obtained via the traditional method, as well as the correlation between the subjective scores and DI measurements obtained using the ANDI Macro. 

While DI measurements obtained using the traditional method correlated with subjective scores (r 2 = 0.7850, p,0.0001; Fig. 5F), the ANDI Macro yielded measurements that correlated much more strongly with subjective ratings (r 2 = 0.978, p, 0.0001; Fig. 5G). Direct statistical comparisons revealed that scores obtained via the ANDI Macro correlated significantly stronger with subjective ratings than do scores obtained via the traditional method (z = 6.626, p, 0.0001; Fig. 5F, G). Further analyses revealed that the macro facilitates the performance of DI measurements with robust and significant improvement in time efficiency. DI analyses performed using the ANDI Macro required an average time of 12.75 s per image analyzed, compared with an average time of 5.27 min per image analyzed via the traditional method [Friedman x2 (1) = 9, p = 0.0027]. Altogether, these results indicate that ANDI yields rapid generation of DI measurements that, compared with scores obtained via the traditional method, more closely approximate the degeneration that investigators perceive from qualitative analysis of neurite images.

After determining the accuracy and efficiency of the ANDI Macro for measuring neurite degeneration in differentiated LUHMES cells exposed to 6-OHDA, we evaluated the suitability of the macro for experiments involving other cell culture models. A pilot study was first conducted to assess the accuracy of ANDI in performing DI measurements using micrographs of primary sympathetic cultures immunolabeled for b III-tubulin and DAPI. Interestingly, however, the cell bodies of sympathetic neurons are larger than those of differentiated LUHMES cells, and thus the operations in ANDI related to automated soma removal did not remove the entirety of the soma (data not shown). 

To overcome this issue, we revised ANDI to feature a dialogue box enabling users to choose the size of the soma removal by controlling the number of times that binarized micrographs of nuclei are dilated. This version of ANDI, version 1.1, was then used to perform DI measurements from micrographs depicting healthy, vehicle-treated sympathetic neurons or degenerating sympathetic neurons exposed to hydrogen peroxide. Images of neurons exposed to hydrogen peroxide yielded significantly higher DI scores compared with images of vehicle-treated neurons (F(1,3) = 71.2, p = 0.0035, R2 = 0.9301; Fig. 6A, C). Moreover, we compared these DI scores that were obtained using ANDI to DI measurements obtained from the same image set following manual removal of cell bodies from fluorescence micrographs using the freehand tool of ImageJ. There was no significant difference in DI scores calculated following manual soma removal compared with scores obtained using ANDI (F(1,3) = 0.1082, p = 0.7639, R2 = 0.0000; Fig. 6A–C). Post hoc tests also confirmed that DI scores obtained using ANDI were not significantly different both in the case of untreated neurons (p = 0.0972) and in the case of neurons exposed to hydrogen peroxide (p = 0.9986). 

Furthermore, DI scores produced from ANDI correlated nearly identically with DI scores obtained following manual removal of cell bodies from corresponding micrographs (r(34) = 0.991, p, 0.0001). These data demonstrate that ANDI is suitable for measuring neurite degeneration in cultures of primary neurons, with soma removal operations that the user can customize to match different neuron types. Additionally, the detection of neurite fragments in cultures exposed to hydrogen peroxide demonstrates the applicability of the macro for use in experiments involving neurite degeneration induced by sources other than 6-OHDA. To further verify the use of ANDI in detecting neurite degeneration associated with various causes, we also used the macro to measure neurite degeneration in micrographs of healthy sympathetic cultures or cultures subjected to NGF withdrawal. As expected, images of cultures subjected to NGF withdrawal yielded significantly higher DI values compared with images of healthy cultures (F(1,2)= 287, p= 0.0035, R2 = 0.958; Fig. 6D, E). Collectively, these findings demonstrate that ANDI can be used to detect neurite degeneration associated with a variety of biological contexts.

Since staining for TH is commonly used to visualize catecholaminergic neurons, we also evaluated the suitability of micrographs depicting TH staining for measuring neurite degeneration using ANDI. Cultures of healthy sympathetic neurons treated with vehicle solution or degenerating sympathetic neurons exposed to hydrogen peroxide were fixed and subjected to immunofluorescence labeling for TH and b III-tubulin, as well as counterstaining for DAPI. Neurite images depicting staining for TH or b III-tubulin, and corresponding images of nuclei featuring DAPI labeling, were then captured at similar fields of view. ANDI was executed, and the images featuring TH staining or b III-tubulin staining were selected as the neurite images. Our analyses revealed that DI scores obtained using images of TH staining correlated very strongly with scores obtained from images of b III-tubulin staining (r(28)= 0.962, p, 0.0001; Fig. 6F, G). These data indicate that accurate DI measurements can be obtained by ANDI using alternative staining procedures that label the entirety of neurites, such as immunolabeling for TH.

To demonstrate the utility of the improved method for performing DI measurements in a scientific experiment, we used the new method to assess the role of JNK in oxidative stress-induced neurite degeneration in human mesencephalic cells. Differentiated LUHMES cells were pretreated for 1 h with the JNK inhibitor SP600125 or vehicle solution. The cells were then treated for 24 h with 6- OHDA to induce oxidative stress or with vehicle solution. Fixed cells were subjected to immunolabeling for b III-tubulin and DAPI staining, and ANDI was used to obtain DI measurements from fluorescence micrographs. While cultures lacking pretreatment with the JNK inhibitor exhibited neurite degeneration following exposure to 6-OHDA, JNK inhibition resulted in a marked and significant reduction in 6- OHDA-induced neurite degeneration (p= 0.0001, d= 2.0209; Fig. 7). These findings support a key role for JNK signaling in neurite degeneration induced by oxidative stress in human mesencephalic cells, and importantly, demonstrate the utility of our new method for generating scientific discoveries by facilitating the rapid and accurate measurement of neurite degeneration. 

Discussion

Neurite degeneration is associated with a variety of neuropathological conditions, yet the molecular mechanisms underlying the degradation of neurites remain incompletely understood. Methods facilitating accurate and efficient analysis of neurite degeneration are essential to the successful identification of novel factors regulating this important cellular event. In the present study, we reveal multiple sources of error associated with a commonly used method for quantifying neurite degeneration. We report experimental data supporting procedural modifications that can be implemented to reduce DI analysis error, and such modifications are incorporated into a new ImageJ macro to provide a tool for rapid, accurate, and objective DI analyses using open-source and free software. Moreover, we demonstrate how the improved method can be applied to measure neurite degeneration in a cell culture model of PD, a disorder in which axonal fragmentation in dopaminergic neurons is an early-stage event that precedes eventual neuronal loss (Tagliaferro and Burke, 2016).

In the present study, we demonstrate the utility of the ANDI Macro in measuring neurite degeneration in a cell culture model of PD consisting of differentiated LUHMES cells exposed to 6-OHDA. 

We report experimental findings supporting the optimal substrates on which the cultures should be established, as well as demonstrate the plating density that is most appropriate for neurite degeneration analyses. The use of this model system to study neurite degeneration has several advantages. The cells develop neurites that are well-networked and typically .500 mm in length, and the cultures are susceptible to neurodegeneration induced by classic, neurotoxin models of PD such as 6-OHDA, 1-methyl-4-phenylpyridinium ion (MPP1), and rotenone (Lotharius et al., 2005; Zhang et al., 2014; Smirnova et al., 2016; Kraemer et al., 2021). Furthermore, the human origin of LUHMES cells increases the translatability of findings associated with the model, and the conditional immortalization facilitates the rapid generation and propagation of cultures (Scholz et al., 2011). Thus, coupled with automated analyses performed using ANDI, LUHMES cells can be used to make novel discoveries related to PD-associated neurite degeneration while facilitating considerably higher throughput compared with nonautomated analyses performed with primary neuronal cultures.

Numerous research groups have performed DI analyses using phase-contrast images to gain insight into factors that regulate neurite degeneration (Press and Milbrandt, 2009; Kraemer et al., 2014; Di Stefano et al., 2015; Hill et al., 2018; Geisler et al., 2019). Here, we report an important drawback associated with this widely used method for measuring neurite degeneration: the binarization of phase-contrast images causes a significant proportion of intact neurites to appear fragmented. While the level of contrast that may be achieved when capturing micrographs varies depending on the microscope system available to investigators, this issue is pervasive among various research groups, as a DI of 0.2 is commonly accepted as a threshold above which cultures are considered to have degenerating axons, and most scientific reports featuring DI measurements from phase-contrast micrographs have reported DI scores for healthy neurons between 0.1 and 0.3 (Sasaki et al., 2009, 2016; Di Stefano et al., 2015; Hill et al., 2018; Loreto et al., 2020). Our DI analyses obtained from fluorescence micrographs, as well as subjective scores from blinded investigators, indicate that 10–20% neurite fragmentation is not common in healthy neuron cultures. Rather, such cultures exhibit only 2.5% neurite fragmentation on average. Thus, our findings underscore the need for methods that improve the accuracy of neurite degeneration analyses by reducing this common and significant source of measurement error.

While the majority of studies involving DI measurements have used phase-contrast images to perform the neurite degeneration analyses, several research groups have recently performed DI measurements using images of neurons immunolabeled for cytoskeletal filaments such as neurofilament medium polypeptide or b -b-tubulin (Wakatsuki et al., 2011; Li et al., 2017; Pease-Raissi et al., 2017; Hernandez et al., 2018; Sundaramoorthy et al., 2020). However, which form of microscopy is best suited for neurite degeneration measurements has remained unclear, as investigations have been needed to understand whether changes in the localization of these cytoskeletal proteins during neurite degeneration appropriately model total changes in neurite morphology, as well as to evaluate the degree to which DI measurements obtained with fluorescence micrographs correlate with similar measurements obtained from phase-contrast micrographs. Here, we demonstrate that fluorescence micrographs depicting b III-tubulin staining not only accurately represent neurite fragmentation, but also facilitate more accurate DI measurements because of their superior contrast and decreased susceptibility to artificial fragmentation on binarization. Thus, our results highlight the utility of using b III-tubulin staining to measure neurite degeneration.

Among the published studies involving DI analyses to measure neurite degeneration, the reported size and circularity parameters used to detect neurite fragments have considerably varied. For example, a recent study involving the assessment of neurite degeneration in cultured hippocampal neurons used size parameters of 4–900 pixels (Li et al., 2017), while a study involving cultured DRG neurons used detection parameters of 0–10,000 pixels (Wakatsuki et al., 2011). One potential reason for such variation is that the efficacy of particle analysis size parameters in detecting neurite fragments varies depending on the resolution of the image since the size parameters for particle detection are associated with pixel units. Unfortunately, most published studies involving DI analyses do not include a description of the image resolution used for the analyses. Thus, the establishment of standard analysis parameters via a protocol that fully discloses important details such as the recommended image resolution has been needed. Here, we report that the most commonly cited parameters for neurite fragment detection, 20– 10,000 pixels (Shin et al., 2012; Di Stefano et al., 2015; Sasaki et al., 2016; Loreto et al., 2020; Shin and Cho, 2020), result in large proportions of neuron fragments being undetected. Reducing the minimum size criterion for neurite fragment detection enhances the sensitivity of fragment detection, while simultaneously increasing the number of image artifacts and non-neurite debris that are falsely detected as neurite fragments. In search for parameters that would best balance detection sensitivity and false-positivity rates, we identified analysis parameters of 10–10,000 as providing the most accurate detection of neurite fragments in binarized images obtained from phase-contrast micrographs, while parameters of 5– 10,000 facilitate the most accurate detection of neurite fragments in binarized images obtained from fluorescence micrographs. While these determinations were made using an image resolution of 1280 1024, such parameters should also enable accurate neurite fragment detection in images captured at other resolutions featuring a vertical size of 1024 pixels, such as the common full-screen resolution of 1024 1024, since neurons depicted in such images would be of similar pixel size.

Numerous research groups have used a minimum size criterion when configuring particle analysis parameters to avoid small, non-neurite debris from being included in neurite fragment measurements (Shin et al., 2012; Cosker et al., 2013; Di Stefano et al., 2015; Sasaki et al., 2016; Li et al., 2017; Loreto et al., 2020; Shin and Cho, 2020; Sundaramoorthy et al., 2020). However, such studies have erroneously included small, non-fragment matter in the measurements of total neurite area, since such measurements are obtained by summing the area of all black pixels in the image. Here, we demonstrate the utility of the Particle Remover plugin to remove small, non-fragment debris from the image before DI analysis. Utilization of the plugin significantly increased the sensitivity of neurite degeneration measurements obtained from phase-contrast images. Thus, our results highlight the value of the Particle Remover plugin for investigators performing DI analyses using phase-contrast micrographs. Interestingly, fluorescence micrographs featured less small, non-neurite debris, and thus, a lower size criterion of five pixels could be used for fragment detection. While particle removal of non-neurite debris less than five pixels in size did not significantly affect DI measurements obtained from fluorescence micrographs, such operations provide values that are theoretically more accurate and therefore were included in ANDI.

Since DI analyses must be performed using images exclusively featuring neurites, application of the analysis method has primarily been limited to use with culture systems amenable to removal of cell bodies, such as explant cultures in which cell bodies can be physically excised (Catenaccio et al., 2017; Shin and Cho, 2020) or cultures in microfluidic devices facilitating segregation of soma and axon compartments (Cosker et al., 2013; Pease-Raissi et al., 2017; Tan et al., 2018). Alternatively performing DI analyses with dissociated cultures can be achieved through tedious and time-consuming labor associated with digital removal of cell bodies from micrographs (Kraemer et al., 2014). In the present study, we reveal that ANDI enables the performance of DI measurements from micrographs featuring LUHMES cells that are dissociated and mosaically-distributed across a 2D culture system. Our findings also demonstrate that ANDI facilitates a 24-fold decrease in time required to perform the analyses, and the operations in ANDI related to soma removal yield DI analysis results that correlate 98% with similar analyses performed following manual soma removal using the freehand tool of ImageJ. Thus, the macro substantially increases the efficiency of the analyses by performing automated soma removal via a process that does not sacrifice accuracy, and such operations expand the suitability of the method for analyzing dissociated cultures with mosaically distributed cell bodies.

In addition to facilitating automated cell body removal from images, ANDI features several revisions to the traditional method for performing DI analysis, including the use of fluorescence micrographs, revised particle analysis parameters, and utilization of the particle remover to remove small, non-neurite debris from images. Our findings indicate that these optimizations collectively yield DI values that more closely approximate values obtained through blinded and subjected scoring. Thus, in addition to substantially reducing the time required to perform the analyses, ANDI facilitates DI measurements that, compared with the traditional method, more accurately reflect the neurite degeneration that investigators reportedly observe from qualitative image analysis.

One drawback of the traditional method for performing DI analyses is that such analyses require training in the use of ImageJ so that the user is familiar with a tedious set of ImageJ operations. Moreover, to perform the analyses efficiently, further training is required for the user to perform batch analyses or to write scripts to semi-automate the execution of particular steps. By fully automating the analysis, ANDI facilitates DI measurements in only a few steps related to downloading and executing the macro. To increase user-friendliness, the macro includes code to provide an interface for selecting directories containing images to be analyzed, as well as to facilitate the output of result images and data tables to a directory of the user’s choosing. Moreover, the script contains numerous features designed to prevent measurement inaccuracies or user mistakes, including automated clearing of previous ImageJ results; removal of image scale information to ensure measurements are accurately performed using pixel units; appropriate configuration of ImageJ color settings, measurement settings, and particle analyzer parameters; error message presentation if users erroneously select directories containing unequal numbers of images depicting b III-tubulin staining and DAPI staining; log descriptions indicating analysis progress; and coding to prevent bugs associated with automated .ini file generation by Windows 10. Altogether, these features make DI analyses more user-friendly, which may foster interest in studies related to neurite degeneration.

To demonstrate its utility in an experiment, we used ANDI to evaluate the contributions of JNK to neurite degeneration in LUHMES cell cultures exposed to 6-OHDA. JNK is a stress-activated kinase that has been reported as key to neurite degeneration in peripheral neurons of nonprimate origin (Shin et al., 2012), but further studies are needed to understand the role of the kinase in neurite degeneration in mesencephalic human cells. Here, we demonstrate that inhibition of JNK significantly protects human mesencephalic cells from neurite degeneration associated with oxidative stress. Such findings not only reveal an important role for JNK in neurite degeneration associated with a model of human disease but also exemplify the utility of ANDI for performing automated and objective image analysis to make new scientific discoveries.

While ANDI was initially designed to facilitate rapid and accurate measurement of neurite degeneration in cultures of differentiated LUHMES cells exposed to 6-OHDA, our findings also indicate that the macro can be used to detect neurite degeneration induced by other causes, such as exposure to hydrogen peroxide or neurotrophin withdrawal. Additionally, version 1.1 of the macro features a dialogue box enabling investigators to customize the soma removal operations for neurons of specific sizes. Our data also supports that staining methods other than immunolabeling for b III-tubulin can be used for accurate measurement of neurite degeneration. Although we recommend the use of pilot studies to verify the accuracy of DI measurements when using ANDI with cell types or staining methods that have not formerly been tested, these findings support that the macro is a generally versatile tool for measuring neurite degeneration in cultured neurons. Since ANDI has not been optimized for measuring neurite degeneration in tissue, we currently only recommend its use for analyzing neurite degeneration associated with cultured cells. However, the code for ANDI is publicly available and thus may also be useful for investigators interested in modifying the script to suit in vivo studies or other applications.

In conclusion, this article presents data revealing common sources of error associated with a widely used method for quantifying neurite degeneration. Experimental approaches were used to improve the method, and such enhancements were incorporated into a free and open-source ImageJ macro that can be used to perform rapid, accurate, and objective analysis of fluorescence micrographs for neurite degeneration. Our findings reveal the efficacy of ANDI for quantifying neurite degeneration in micrographs depicting a cell culture model of PD, as well as provide proof of the principle that the macro can detect neurite degeneration associated with other biological contexts. The user-friendly macro will reduce the time required for investigators to learn to quantify neurite fragmentation while increasing the throughput and accuracy of studies evaluating factors underlying neurite degeneration. The ANDI Macro is available at the following URL: https:// github.com/kraemerb/kraemerlab.

References

1. Aarsland D, Creese B, Politis M, Chaudhuri KR, Ffytche DH, Weintraub D, Ballard C (2017) Cognitive decline in Parkinson's disease. Nat Rev Neurol 13:217–231.

2. Adalbert R, Coleman MP (2013) Review: axon pathology in age-related neurodegenerative disorders. Neuropathol Appl Neurobiol 39:90–108.

3. Bové J, Perier C (2012) Neurotoxin-based models of Parkinson’s disease. Neuroscience 211:51–76.

4. Catenaccio A, Llavero Hurtado M, Diaz P, Lamont DJ, Wishart TM, Court FA (2017) Molecular analysis of axonal-intrinsic and glial-associated co-regulation of axon degeneration. Cell Death Dis 8: e3166.

5. Cosker KE, Pazyra-Murphy MF, Fenstermacher SJ, Segal RA (2013) Target-derived neurotrophins coordinate transcription.

6. Di Stefano M, Nascimento-Ferreira I, Orsomando G, Mori V, Gilley J, Brown R, Janeckova L, Vargas ME, Worrell LA, Loreto A, Tickle J, Patrick J, Webster JRM, Marangoni M, Carpi FM, Pucciarelli S, Rossi F, Meng W, Sagasti A, Ribchester RR, et al. (2015) A rise in NAD precursor nicotinamide mononucleotide (NMN) after injury promotes axon degeneration. Cell Death Differ 22:731–742.

7. Fukui K (2016) Reactive oxygen species induce neurite degeneration before induction of cell death. J Clin Biochem Nutr 59:155–159.

8. Geden MJ, Deshmukh M (2016) Axon degeneration: context defines distinct pathways. Curr Opin Neurobiol 39:108–115.

9. Geisler S, Doan RA, Cheng GC, Cetinkaya-Fisgin A, Huang SX, Höke A, Milbrandt J, DiAntonio A (2019) Vincristine and bortezomib use distinct upstream mechanisms to activate a common SARM1-dependent axon degeneration program. JCI Insight 4:e129920.

10. Hernandez DE, Salvadores NA, Moya-Alvarado G, Catalan RJ, Bronfman FC, Court FA (2018) Axonal degeneration induced by glutamate excitotoxicity is mediated by necroptosis. J Cell Sci 131:jcs214684.

For more information:1950477648nn@gmail.com