Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2019 Feb 1.
Published in final edited form as: J Neurosci Methods. 2018 May 18;305:98–104. doi: 10.1016/j.jneumeth.2018.05.008

Semi-automated quantification and neuroanatomical mapping of heterogeneous cell populations

Oscar A Mendez 1,4,#, Colin J Potter 2,#, Michael Valdez 3, Thomas Bello 4, Theodore P Trouard 3,4,5, Anita A Koshy 4,6,7,*
PMCID: PMC6357967  NIHMSID: NIHMS996003  PMID: 29782884

Abstract

Background:

Our group studies the interactions between cells of the brain and the neurotropic parasite Toxoplasma gondii. Using an in vivo system that allows us to permanently mark and identify brain cells injected with Toxoplasma protein, we have identified that Toxoplasma-injected neurons (TINs) are heterogeneously distributed throughout the brain. Unfortunately, standard methods to quantify and map heterogeneous cell populations onto a reference brain atlas are time consuming and prone to user bias.

New method:

We developed a novel MATLAB-based semi-automated quantification and mapping program to allow the rapid and consistent mapping of heterogeneously distributed cells on to the Allen Institute Mouse Brain Atlas. The system uses two-threshold background subtraction to identify and quantify cells of interest.

Results:

We demonstrate that we reliably quantify and neuroanatomically localize TINs with low intra-or inter-observer variability. In a follow up experiment, we show that specific regions of the mouse brain are enriched with TINs.

Comparison with existing methods:

The procedure we use takes advantage of simple immunohistochemistry labeling techniques, use of a standard microscope with a motorized stage, and low cost computing that can be readily obtained at a research institute. To our knowledge there is no other program that uses such readily available techniques and equipment for mapping heterogeneous populations of cells across the whole mouse brain.

Conclusion:

The quantification method described here allows reliable visualization, quantification, and mapping of heterogeneous cell populations in immunolabeled sections across whole mouse brains.

Keywords: Semi-automated quantification, cell-mapping, Toxoplasma gondii, atlas, image processing

1. Introduction

Given the intimate link between neuroanatomic location and function, there has always been an interest in mapping and quantifying cells and disease processes throughout the brain. Until recently, processing these large sets of data has been time consuming and labor intensive as it required manual counting and neuroanatomic localization. To address this issue, various quantification methods have been developed. Stereological methods quantify cells in a precise stereotyped manner (Sterio 1984; Howard et al., 1998; Gundersen et al., 1986; West et al., 1991; Hedreen, 1998). These techniques were developed to count cells in relatively small sample regions of pre-identified neuroanatomic locations and in regions with uniform cell densities (Schmitz et al., 2005). However, in certain scenarios, like infectious models of disease, the cells of interest can be heterogeneously distributed throughout the brain and idiosyncratically vary between mice, thus making proper stereology quantification difficult to do. To address this gap, there has been growing focus on imaging of whole brains or thick serial sections using a variety of imaging techniques such as confocal, 2-photon, or light-sheet microscopy, and micro-optical sectioning tomography. These imaging techniques have been used in combination with complex algorithms and machine learning techniques to localize and quantify cells (Peng et al., 2017; Inglis et al., 2008). The equipment used and methods of quantification are not readily available at many universities. Here, we describe a semi-automated method that can be utilized by most researchers because it relies on standard immunohistochemistry and basic light microscopy coupled to a MATLAB (Mathworks, Inc., Natick, MA)-based program to count and map cells of interest onto the Allen Institute Mouse Brain Atlas (http://www.brain-map.org/).

We developed this methodology to define neurons and CNS regions that are targeted by the protozoan parasite, Toxoplasma gondii. T. gondii is an intracellular parasite that naturally infects a wide range of warm blooded hosts, including humans and rodents (Dubey et al., 1998; Dubey 2008). In most hosts, T. gondii is able to establish a life-long infection in specific tissues. In humans and rodents, the central nervous system is the major organ for T. gondii persistence (Dubey 1998, Remington & Cavanaugh, 1965). This tropism for and persistence in the CNS underlies the devastating symptomatic disease T. gondii causes in those with under-developed immune responses (e.g. fetus, AIDS patients) (Luft & Remington, 1992; McLeod et al., 2009). The persistent form of the parasite, the bradyzoite, is a slow-growing form that establishes intracellular cysts, primarily in neurons (Ferguson and Hutchinson, 1987; Melzer 2010; Cabral et al., 2016). Thus, understanding the T. gondii-neuron interaction is essential to understanding symptomatic toxoplasmosis.

Until recently, the only way to identify infected brain regions was by cyst location. Several studies found that encystment is highest in the neocortex, thalamus, and striatum, along with other forebrain structures (Berenreiterová et al., 2011; Evans et al., 2014). However, cyst location does not necessarily give an accurate representation of which neurons are infected, as cysts can be located in distal processes > 100 microns from the neuron soma (Cabral et al., 2016). The recent development of a system in which permanent host cell GFP-expression is triggered by injection of a parasite protein significantly changed this landscape (Koshy et al., 2012; Cabral et al., 2016). First, GFP expression enables visualization of the whole neuron (both cyst location and soma) (Koshy & Cabral, 2014). Second, as this system is dependent upon injection of parasite protein, not active infection, it revealed that T. gondii injects its effector proteins—parasite proteins that hijack host cell processes and signals—into far more host cells than it productively infects (Koshy et al., 2012). This finding is particularly pronounced in the CNS, where T. gondii-injected neurons (TINs) out-number cysts up to 50-fold (Koshy et al., 2012). Thus, by mapping TINs, we can determine if cysts are commonly found in specific brain regions because these regions are particularly susceptible to T. gondii infection or if neurons display regional differences in the capability to clear intracellular parasites, as has been suggested for West Nile Virus (Cho et al., 2013). To test this hypothesis, we need to reliably quantify and neuroanatomically localize TINs throughout the entire brain.

Unfortunately, TIN location in the brain widely varies by mouse and TINs are non-uniformly dispersed throughout the brain. Manual counting of these large populations of TINs would be time-consuming and inefficient. Thus, the lack of large-scale, efficient methods to quantify and localize idiosyncratic cell distributions throughout a whole mouse brain section was an impediment to moving this work forward. To overcome this barrier, we developed a semi-automated, MATLAB-based method that allows rapid quantification and neuroanatomical localization across whole-brain histological images. While the program has been developed for identifying TINs, it can be applied to count and localize a wide variety of cells or processes within histological images of sagittal mouse brain sections.

2. Methods

2.1. Animal and Parasite Model

The animal model used for these experiments is B6.Cg-Gt(ROSA)26Sortm6(CAG-ZsGreen1)Hze/J on a C57BL/6J background (Madisen et al., 2010). The cells of these Cre reporter mice only express a green fluorescent protein (ZsGreen) after Cre-recombinase mediated genetic recombination. The T.gondii parasite strain used for this study was engineered to inject Cre into host cells concomitantly with parasite effector proteins (parasite proteins used to manipulate host cells) (Koshy et al., 2010; Koshy et al., 2012).

Parasites were maintained via serial passage through human foreskin fibroblasts using DMEM supplemented with 2 mM glutagro, 10% fetal bovine serum, and 100 I.U./ml penicillin/100 μg/ml streptomycin (Cabral, Tuladhar et al., 2016). Mice were infected at 2–3 months of age via intraperitoneal (IP) injection with freshly syringe-released parasites. Mice were inoculated with 10,000 parasites per 200 μl volume in USP grade PBS. At 3 weeks post infection, animals were sedated with a ketamine/xylazine cocktail, intracardially perfused with saline followed by 4% paraformaldehyde, after which brains were harvested.

2.2. Immunohistochemistry

Left and right brain hemispheres were isolated and 40 micron-thick sagittal sections were generated using a freezing sliding microtome (Microm HM 430). Sections were sampled every 200 microns to obtain a set of 20 sections, that would increase the likelihood of matching sections to the Allen Brain Mouse Atlas (ABA). Sections were pre-mounted on slides before immunohistochemical labeling (Fig. 1).

Figure 1. Tissue processing workflow and mapping of sections onto Allen atlas reference.

Figure 1.

Open in a new tab

Three weeks after intraperitoneal infection with parasites, brains are harvested, sectioned, and then sampled every 200um. Slides are premounted onto charged slides, immunloabled for a green fluorescent protein (ZsGreen), counterstained with Cresyl violet and imaged.

To insure adhesion of tissue onto slides, tissue was allowed to air-dry onto slides overnight followed by dehydrated using increasing then decreasing concentrations of 50%, 75%, 95%, and 100% ethanol. Slides were washed with TBS, peroxidases inactivated (3%H2O2/10% methanol), permeabilized (0.6% Triton X-100), blocked (1.5% BSA and 1.5% goat serum), and incubated with Rabbit anti-ZsGreen (Clontech, Cat. No. 632474, 1:10,000) for 15–18hrs. Next, slides were incubated in Goat anti-rabbit polyclonal biotinylated conjugated antibody (Vector Labs, Cat. No. BA-1000, 1:500) for 2 hrs, incubated with avidin-biotin complex kit (ThermoFischer Scientific, Cat. No. 32020) for 2 hrs and visualized with a 3,3’-diaminobenzidine kit (Vectastain, Vector Labs Cat. No. SK-4100). Sections were then counterstained with cresyl violet for Nissl labeling (Dorph-Petersen et al., 2001). Although the Nissl counterstain is not part of the colorimetric thresholds used in cell detection, it increased the contrast between TINs and surrounding tissue, which improved the consistency of automated cell detection. After processing tissue sections were estimated to shrink by over 50%.

2.3. Microscopy

Slides were imaged on a Leica DMI6000 with a motorized stage, using Leica Application Suite X (LAS X). All images were taken with a 10× objective lens. The working distance of the objective used was sufficient to capture the whole depth of TINs. Base background subtraction and white balance was maintained throughout the same experiment to ensure consistent exposure. Image stitching was done automatically through LAS X with a 10% overlap. Images were stored as Leica Image File Format (lif).

2.4. Computer hardware and software specifications

All images were processed on a computer with an Intel 6-core 6800k processor with 32GB of RAM. The operating system used was Windows 10 Professional. All MATLAB code was written and run using MATLAB 2015a.

2.5. Cell detection and counting

Individual histology section images were imported into MATLAB and TINs were automatically identified using colorimetric and size thresholds (Fig. 2A). Immunohistochemistry used to stain TINs brown caused variability in brown background staining. Some regions had a light brown background (high background labeling) while others had no brown background staining. Consequently, TINs in the high background regions consistently stained darker (Fig. 2A” schematic) than those that were on unstained background (Fig. 2A’ schematic). To make identification of the TINs more accurate, the tissue was separated into two regions—light brown background (Fig. 2B) and unstained background (Fig. 2B’). This separation was done by using a color threshold to segment the light brown region of the tissue. The threshold was manually-selected and was tested across different tissue sets. To smooth the edges of the light brown region, a series of erode and dilate functions were used to remove small irregular objects that lay outside the light brown region. The complementary region was used as the unstained tissue background region. Both regions were dilated slightly so they would overlap to account for minor region segmentation errors. This overlapping mask accounts for TINs that may have been located on region border or be located slightly outside the intended region.

Figure 2. Workflow of MATLAB cell detection and quantification includes separation of image into areas of high and no brown background staining.

Figure 2.

Open in a new tab

(A) Original image. Top schematic shows how TINs stain dark brown when in area with high brown background staining (A”). Bottom schematic shows how TINs stain light brown in area with no brown background staining (A’). Experimental image (A) is split into two regions through RGB thresholds: Light brown regions from high brown background labeling (top row, B) and regions with no brown background staining (bottom row, B’). Separate RGB color thresholds are applied to each region to generate a Boolean mask of DAB+ regions (C, C’). Perimeter of shapes is eroded to remove irregular shapes (D,D’), then dilated back to original size (E, E’). The two masks are combined for final quantification (F). Regions are marked with centroids (G), and operators use a dialog box to manually error-check identified cells (H).

In the light brown background regions, where TINs were consistently stained dark brown, ImageJ was used to measure the RGB values of several manually-selected TINs. These RGB values were used to threshold and segment the TINs (Fig. 2C). TINs in the no-brown-background tissue region were light brown and segmented with the same methods (using RGB values defined to be specific for these lighter brown TINs) (Fig. 2C’). Small, non-circular objects (e.g. processes) were removed with a sequence of erode functions (Fig. 2 D/D’). The remaining segmented objects (TINs) were dilated to their original diameter (Fig. 2 E/E’). A size threshold was used to remove objects that were larger than cells. The light and dark brown segmented TINs were merged to eliminate double-counting in the overlapping regions (Supplemental Fig. 1F). The TIN centroids were automatically encircled on the section image (Fig. 2 F,G). The analyzed image was then manually corrected (i.e. unmarked TINs (false-negatives) were marked and false positive TINs were unmarked) using a custom-built graphical user interface (GUI) that allowed for rapid selection and removal of cells. By intent, RGB threshold values were set to over-count TINs (i.e. increased false-positive TIN detection) as it was more efficient to remove incorrectly identified cells than to examine the whole section and identify uncounted TINs (i.e. false-negative TIN detection) if the RGB values were too stringent. A decrease in the RGB threshold range values by 40% led to a decrease in TIN centroids (Supplemental Fig. 2C), while an increase in the threshold value range by 40% increased the number of TIN centroids and false positive cell bodies (Supplemental Fig. 2F). The RGB threshold values used were determined empirically and gave reliable results across different mice, tissue sets, and staining runs.

2.6. Image registration

Mapping of brain section images to the Allen Brain Atlas (ABA) was accomplished in a multi-step process. First, 20 sagittal atlas images were downloaded as vector images (in SVG format). Each anatomical region was outlined with black lines. The outlines were removed using Inkscape software (Albert, M., et al. “Inkscape.” 2015) and the images were saved as pixelated images. A set of 20 corresponding reference infected brain tissue section images (referred to as reference images from here forward) were imported into MATLAB using the Open Microscopy Environment Consortium’s Bio-Formats software library (Linkert et. al., 2010). Pairs of corresponding reference and atlas images were then registered with an affine transformation (translation, rotation, scale, and shear) followed by a piecewise transformation using the MATLAB control point registration tool. Approximately 30 control points were manually placed on identical anatomical landmarks in the reference and atlas images, with emphasis on the tissue borders as well as internal structures. Empirical testing of brain landmarks that gave the best fit led to exclusion of control points on the olfactory bulb and cerebellum. The transformation matrices generated by the control point registration were saved so they could be used in subsequent processing steps. While this procedure required a significant amount of user interaction, it only had to be done once. New, experimentally generated labeled brain section images were registered to the untransformed reference brain section images with an automatic affine transformation. Finally, all the brain section images were warped using the saved transformation matrices so that all experimental sections were aligned to the atlas images. Registration alignment for brain section images were visually inspected by overlaying individual images to the corresponding atlas image (Fig. 3). The number of TINs in 12 major anatomical regions of the brain was determined using the TIN centroids and Boolean masks for each region of the brain identified in the atlas image. In the atlas images, each anatomical region has a defined color (Fig. 3B). Masks for each anatomical region were created by isolating an SVG image and isolating each colored region.

Figure 3. Image registration of brain sections onto atlas image.

Figure 3.

Open in a new tab

(A) Anatomical control points on reference image 14 corresponding to control points on Allen Mouse Brain Atlas (B). (C) Output transformation of (A) onto (B) from cell mapping program.

2.7. Validation

Intra-operator and inter-operator variability were assessed to determine the reliability of our program. Intra-operator analysis was conducted by comparing the semi-automated program results between two independent sessions with the same operator. A single operator used the program to count cells in 3 sets of tissue, each consisting of 20 sections. The same operator counted the same 3 sets of tissue during a second counting session two days later. To assess variability between sessions, the coefficient of variation (CV) was calculated for the total cell counts of each set independently by dividing the standard of deviation of the cell counts by their mean value (SD/Mean). The average CV was calculated between sets to take all sets into account. To assess cell-mapping consistency, counts from Set 1 were broken down into major brain regions to compare the number of cells localized to each region between different sessions.

Assessment of inter-operator error was carried out by comparing the total cell counts between two operators counting the same 3 sets of tissue in independent sessions without a supervisor. The CV was calculated for each set independently and averaged, similar to intra-operator analysis. Set 1 was split into major brain regions to compare cells localized to each region for different operators.

3. Results

3.1. False-positive and false-negative rates

We designed our program to err towards over-counting our manual counts, since we found deleting points easier than manually adding cells. Two separate users sampled tissue sections from one mouse and compared counts to the uncorrected output by the program. Here we demonstrate that our program tends to over-count across most regions of the brain across users with little variability (Fig. 4, Table 1). In two regions, the striatum and the hippocampal formation, the program identified significantly fewer TINs than User 1 (−19 and −26 respectively, Table 1.) After the independent counts, User 1 and 2 reviewed these areas together and agreed that User 1 had incorrectly identified a number of TINs, which is not surprising given that User 1 (undergraduate) has far less experience than User 2 (graduate student). This highlights the point that an automated system is less likely to introduce less user-to-user variability, especially when accounting for differing ranges of human expertise and bias. Overall, these data suggest that the program performs as designed by primarily over-counting within the range of user variability.

Figure 4. Automated program primarily over-identifies TINs compared to manual counts.

Figure 4.

Open in a new tab

TINs quantification from two, independent users compared to uncorrected program counts. The counts are broken down by twelve major brain regions. Blue colored dots represent uncorrected counts generated by the program. Green and red dots represent manual counts from two users. N=1 mouse, 20 sections

Table 1. TINs mapping from two independent users compared to uncorrected counts generated by automated program.

Users sampled sets of tissue that corresponded to specific Allen atlas sections. Final outputed centroids were corrected since by users since output parameters are set to overcount TINs. For user 1 quantification of final counts compared to uncorrected program counts range from −26 to +25. For user 2 final count differences ranged from +2 to +51.

Region User 1 User 2 No correction +/− from User 1 +/− from User 2
Isocortex 448 402 453 5 51
Striatum 136 114 117 −19 3
Pallidum 9 8 10 1 2
Hypothalamus 51 44 46 −5 2
Olfactory Areas 53 38 55 2 17
Cortical Subplate 29 21 26 −3 5
Thalamus 22 17 22 0 5
Midbrain 57 43 53 −4 10
Hippocampal Formation 156 104 130 −26 26
Cerebellum 13 10 20 7 10
Pons 4 2 8 4 6
Medulla 7 7 15 8 8
Tracts 61 65 86 25 21
Open in a new tab

3.2. Intra-operator and inter-operator validation

The average coefficient of variation (CV) across all 3 sets of tissue for intra-operator analysis was 4.97% (Fig. 5A). Comparison of cell localization data by major brain regions between sessions shows similar values in each of the major brain regions, with an average difference of 8 cells in each region (Fig. 5B). Low variation in overall cell quantification and in localization between sessions indicates effectiveness and consistency of the semi-automated quantification method between different sessions with the same operator.

Figure 5. Quantification and validation of semi-automated program.

Figure 5.

Open in a new tab

Three weeks after infection with parasites, a brain was harvested, sectioned, sampled, and stained as previously outlined. TINs were quantified from 3 sets of tissue from this brain. (A) Intra-operator cell counts in 2 independent experiments for 3 different tissue sets. Coefficient of variation was 3.80%, 4.63%, and 6.20% for Sets 1, 2, and 3 respectively, and the average was 4.97%. Low variation indicates that the program is reliable between experiments with the same operator. (B) Intra-operator cell counts for Set 1 broken down into major mouse brain regions. Low variation indicates regional consistency in cell counting, and consistency in semi-automated mapping. (C) Inter-operator cell counts for 3 different tissue sets. Cells were counted in the absence of a supervisor to eliminate unintentional bias. The coefficient of variation was 2.06%, 0.77%, and 2.19% for Sets 1, 2, and 3 respectively, for an average coefficient of variation of 1.68%. (D) Inter-operator cell counts for Set 1 broken down into the major mouse brain regions which showing low regional variability. N = 1 mouse, 20 tissue sections/set, 3 sets total.

To further validate our system, we wanted assurance that our quantification was reliable across operators. An average CV of 1.68% for inter-operator variation was calculated across all 3 quantified sets (Fig. 5C). A comparison of cell counts by major region also shows similar values across these regions, with an average difference of 10 cells (Fig. 5D), indicating an accuracy of cell quantification and localization between different operators.

3.3. Quantification and localization of TINs in the infected mouse brain

To apply this novel semi-automated method, six Cre-reporter mice were infected with T. gondii-Cre parasites for 3 weeks, after which the brain was harvested and tissue was processed and analyzed using the semi-automated method. Sections sampled for these mice ranged from 16–20 sections, a full set of 21 sections was not possible for all animals because of variability from sectioning. Raw cell counts obtained from our program showed that TINs are most commonly found in the isocortex and striatum (Fig. 6A) but that whether TINs are highest in the cortex or striatum can vary widely between mice (Fig. 6B). These data show that this novel semi-automated quantification and localization method can be used to rapidly and accurately analyze large sets of tissue.

Figure 6. TINs are enriched in the isocortex and striatum, though high mouse-to-mouse variability is observed.

Figure 6.

Open in a new tab

Cre-reporter mice were infected with Toxoplasma parasites. Brains were harvested at 3 weeks post infection and sampled as previously described. Total TIN counts were grouped into 12 major regions of the brain, as described by the Allen Institute. Each colored coded point is an individual mouse (A) Total TIN count is highest in the isocortex and striatum, with a mean of 65.17 and 43.5 respectively. (B) Mapping of only 2 mice to demonstrate the mouse-to-mouse variability in counts in specific brain regions. Mouse A (light blue) has a much higher overall TIN count with the absolute highest number of TINs in the isocortex. Mouse B (dark blue) has a relatively low number of TINs overall but has a high absolute number of TINs in the striatum. Bars, mean ±SEM 16–20 sections/mouse, N= 6 mice

3.4. Software efficiency and viability

Compared to manual counting, semi-automated counting is significantly faster. After imaging was completed, quantification and mapping of TINs from 6 mice, each with 16–20 whole-sections, was completed in less than a week, with automated counting and image registration running overnight. This indicated a nearly 3-fold increase in tissue processing capability when compared to manual counting and mapping of experimental tissue. Automatic image counting time ran from 32 seconds on smaller images to 108 seconds on larger images, and image transformation generally ran from 76 – 139 seconds, which was dependent on the image size. GUI-assisted counting was the longest step in the protocol, each section taking on average 6–7 minutes.

4. Discussion

We have developed a unique semi-automated cell quantification and localization protocol. This method allows operators to obtain total cell counts across a whole-brain section throughout the entire brain within a 5 percent variation between trials and operators, effectively eliminating most of the subjectivity associated with manual counting. We implemented this method and were able to effectively quantify and localize TINs in the mouse brain. The low variability in the intra- and inter-operator quantification indicates reliable, reproducible data acquisition between different sessions and different users. While this approach was shown to be effective for identifying TINs, this method can be adapted to quantify and localize other immunolabeled cells or processes (e.g. beta-amyloid plaques in AD mouse models) by changing the parameters the program utilizes to identify cells of interest. In addition, the MATLAB code can be modified to quantify cells in fluorescently labeled tissue, extending the program’s potential usages. Coding in MATLAB allows these parameters to be altered easily with trial-and-error refining of these parameters. Finally, this method could be utilized to quantify large, non-homogeneous cell populations in the spinal cord or any other organ, provided an atlas exists for the organ of interest, as a new reference atlas could be generated. This is especially useful if little or no data exists about localization of a particular cell or aggregate of interest.

Future studies characterizing the localization of TINs are now possible using this semi-automated method. Large-scale quantification of TIN localization could eventually lead to correlational studies, linking the localization of TINs in the brain with specific behavioral abnormalities in infected mice. This method provides accurate data with minimal set up and investment in time. Recently, other methods have been established to accomplish similar goals (Peng et al., 2017; Inglis et al., 2008, Leibman et al, 2016), but these methods require access to high-end equipment, computing resources and dedicated computer science departments that are often unavailable at smaller research institutions. Our method is carried out using a relatively low-cost computer, a standard version of MATLAB, and a light microscope with a motorized stage with image stitching capabilities, all of which are commonly available even at smaller universities. This semi-automated method is a valuable approach to first-steps quantification and localization of immunolabeled objects and overall provides accurate and consistent results using basic laboratory equipment and computing, making it a fast and versatile way to quantify such cells.

Supplementary Material

1
2
3

Highlights:

  • Developed a semi-automated quantification and mapping program.

  • Reliably quantifies and neuroanatomical localizes heterogeneous distributed cells.

  • Using this program, we determined that specific areas of the brain are enriched with Toxoplasma-injected neurons.

Acknowledgements

The authors would like to thank Elizabeth Fernandez for aiding in the cell detection script and Carla Cabral for initial tissue sets used in these experiments. We would also like to thank the additional members of the Koshy and Trouard groups for helpful discussion of the methods and manuscript. Funding was provided by the National Institute of Neurologic Disorders and Stroke R01 NS095994 (AAK), supplemental funding 01S1 (OAM) and the Arizona Biomedical Research Commission ADHS14–082991 (AAK).

Footnotes

Conflict of interest

The authors declare no actual or potential conflict of interest.

References:

  1. Sterio DC. The unbiased estimation of number and sizes of arbitrary particles using the disector. J Microsc 1984;134:127–36. [DOI] [PubMed] [Google Scholar]
  2. Howard CV, Reed MG. Unbiased Stereology: Three-Dimensional Measurement in Microscopy. Springer Verlag, 1998. pp. 1–19, 69,–106, 151–176. [Google Scholar]
  3. Gundersen HJ. Stereology of arbitrary particles: A review of unbiased number and size estimators and the presentation of some new ones, in memory of William R. Thompson. J Microsc 1986;143:3–45 [PubMed] [Google Scholar]
  4. West MJ, Slomianka L, Gundersen HJ. Unbiased stereological estimation of the total number of neurons in the subdivisions of the rat hippocampus using the optical fractionator. Anat Rec 1991;231:482–97. [DOI] [PubMed] [Google Scholar]
  5. Hedreen JC. Lost Caps in Histological Counting Methods. Anat Rec 1998;250:366–372. [DOI] [PubMed] [Google Scholar]
  6. Schmitz C, Hof PR. Design-Based Stereology in Neuroscience. Neuroscience 2005;130:813–31. [DOI] [PubMed] [Google Scholar]
  7. Peng J, Long B, Yuan J, Peng X, Ni H, Li X, Gong H, Luo Q, Li A. A quantitative analysis of the distribution of CRH neurons in whole mouse brain. Frontiers in neuroanatomy 2017;11:63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Inglis A, Cruz L, Roe DL, Stanley HE, Rosene DL, Urbanc B Automated identification of neurons and their locations. Journal of Microscopy 2008;230(3):339–352. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Liebmann T, Renier N, Bettayeb K, Greengard P, Tessier-Lavigne M and Flajolet M, 2016. Three-dimensional study of Alzheimer’s disease hallmarks using the idisco clearing method. Cell reports, 16(4), pp.1138–1152. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Dubey JP, Lindsay DS, Speer CA. Structures of Toxoplasma gondii Tachyzoites, Bradyzoites, and Sporozoites and Biology and Development of Tissue Cysts. Clin Microbio Rev 1998;11(2):267–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Dubey JP. The history of Toxoplasma gondii-the first 100 years. J Eukaryot Microbiol 2008;55: 467–75. [DOI] [PubMed] [Google Scholar]
  12. Remington Jack S., and Cavanaugh Ellen N.. “Isolation of the encysted form of Toxoplasma gondii from human skeletal muscle and brain.” New England Journal of Medicine 273.24 (1965): 1308–1310 [DOI] [PubMed] [Google Scholar]
  13. Luft BJ & Remington JS Toxoplasmic encephalitis in AIDS. Clin. Infect. Dis 15, 211–222 (1992). [DOI] [PubMed] [Google Scholar]
  14. McLeod R, Kieffer F, Sautter M, Hosten T & Pelloux H Why prevent, diagnose and treat congenital toxoplasmosis? Mem. Inst. Oswaldo Cruz 104, 320–344 (2009). [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Ferguson DJ, Hutchinson WM. An ultrastructrual study of the early development and tissue cyst formation of Toxoplasma gondii in the brains of mice. Parasitology Research 1987;73:483–91. [DOI] [PubMed] [Google Scholar]
  16. Melzer TC, Cranston HJ, Weiss LM and Halonen SK, 2010. Host cell preference of Toxoplasma gondii cysts in murine brain: a confocal study. Journal of neuroparasitology, 1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Cabral C, Tuladhar S, Dietrich HK, Nguyen E, MacDonald WR, Trivedi T, Devineni A, Koshy AA. Neurons are the Primary Target Cell for the Brain-Tropic Intracellular Parasite Toxoplasma gondii. PLoS Path 2016;12(2):e1005447. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Berenreiterová M, Flegr J, Kuběna AA, Němec P. The Distribution of Toxoplasma gondii Cysts in the Brain of a Mouse with Latent Toxoplasmosis: Implications for the Behavioral Manipulation Hypothesis. PLoS ONE 2011;6(12);e28925. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Evans AK, Strassmann PS, Lee I-P, Sapolsky RM. Patterns of Toxoplasma gondii cyst distribution in the forebrain associate with individual variation in predator odor avoidance and anxiety-related behavior in male Long-Evans rats. Brain Behav Immun 2014;37:122–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Koshy AA, Fouts AE, Lodoen MB, Alkan O, Blau HM, Boothroyd JC. Toxoplasma secreting Cre recombinase for analysis of host-parasite interactions. Nat Methods 2010;7:307–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Koshy AA, Dietrich HK, Christian DA, Melehani JH, Shastri AJ, Hunter CA, Boothroyd JC. Toxoplasma Co-opts Host Cells It Does Not Invade. PLoS Path 2012;8(7):e1002825. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Koshy Anita A., and Cabral Carla M.. “3-D imaging and analysis of neurons infected in vivo with Toxoplasma gondii.” Journal of visualized experiments: JoVE 94 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Cho Hyelim, et al. “A role for Ifit2 in restricting West Nile virus infection in the brain.” Journal of virology 87.15 (2013): 8363–8371. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Madisen L, Zwingman TA, Sunkin SM, Oh SW, Zariwala HA, Gu H, et al. A robust and high-throughput Cre reporting and characterization system for the whole mouse brain. Nat Neurosci. 2010;13:133–140. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Dorph-Petersen KA, Nyengaard JR, Gundersen HJG. Tissue shrinkage and unbiased stereological estimation of particle number and size. Journal of Microscopy 2001;204(3):232–246. [DOI] [PubMed] [Google Scholar]
  26. Linkert M, Rueden CT, Allan C, Burel JM, Moore W, Patterson A, Loranger B, Moore J, Neves C, MacDonald D and Tarkowska A, 2010. Metadata matters: access to image data in the real world. The Journal of cell biology, 189(5), pp.777–782. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS996003-supplement-1.eps (1.2MB, eps)
2
NIHMS996003-supplement-2.eps (1.1MB, eps)
3
NIHMS996003-supplement-3.docx (145.4KB, docx)

RESOURCES