Investigating Drivers of Antireward in Addiction Behavior with Anatomically Specific Single-Cell Gene Expression Methods

Summary

The combination of laser capture microdissection and microfluidic RT-qPCR provides anatomic and biotechnical specificity in measuring the transcriptome in single neurons and glia. Applying creative methods with a system's biology approach to psychiatric disease may lead to breakthroughs in understanding and treatment such as the neuroinflammation antireward hypothesis in addiction.

Abstract

Increasing rates of addiction behavior have motivated mental health researchers and clinicians alike to understand antireward and recovery. This shift away from reward and commencement necessitates novel perspectives, paradigms, and hypotheses along with an expansion of the methods applied to investigate addiction. Here, we provide an example: A systems biology approach to investigate antireward that combines laser capture microdissection (LCM) and high-throughput microfluidic reverse transcription quantitative polymerase chain reactions (RT-qPCR). Gene expression network dynamics were measured and a key driver of neurovisceral dysregulation in alcohol and opioid withdrawal, neuroinflammation, was identified. This combination of technologies provides anatomic and phenotypic specificity at single-cell resolution with high-throughput sensitivity and specific gene expression measures yielding both hypothesis-generating datasets and mechanistic possibilities that generate opportunities for novel insights and treatments.

Introduction

Addiction remains a growing challenge in the developed world^1,2. Despite major scientific and clinical advances, rates of addiction continue to increase while the efficacy of established treatments remains stable at best^3,4,5. However, advances in biotechnology and scientific approaches have led to novel methods and hypotheses to further investigate the pathophysiology of substance dependence^6,7,8. Indeed, recent developments suggest that novel concepts and treatment paradigms may lead to breakthroughs with social, economic, and political consequences^9,10,11,12.

We investigated antireward in the withdrawal of alcohol and opioid dependence^13,14,15,16. Methods are central to this paradigm^17,18. Laser capture microdissection (LCM) can select single cells with high anatomic specificity. This functionality is integral to the neuroinflammation antireward hypothesis as both glia and neurons can be collected and analyzed from the same neuronal subnucleus in the same animal^{13,14,15,16,19}. A relevant portion of the transcriptome of selected cells can then be measured with high-throughput microfluidic reverse transcription quantitative polymerase chain reactions (RT-qPCR) providing high-dimensional datasets for computational analysis yielding insights into functional networks^20,21.

Measuring a subset of the transcriptome in neurons and glia in a specific brain nucleus generates a dataset that is robust in both sample number and genes measured and is sensitive and specific. These tools are optimal for a system's neuroscience approach to psychiatric disease because glia, mainly astrocytes and microglia, have demonstrated a central role in neurological and psychiatric disease over the past decade^22,23. Our approach can measure the expressive response of glia and neurons concomitantly across numerous receptors and ligands involved in local paracrine signaling. Indeed, signaling can be inferred from these datasets using various quantitative methods such as fuzzy logic²⁴. Further, the identification of cellular subphenotypes in neurons or glia and their function can provide insight into how brain cells in specific nuclei organize, respond to, and dysregulate at the single-cell level. The dynamics of this functional system can also be modeled with time series experiments¹⁶. Lastly, animal models can be perturbed anatomically or pharmacologically to lend a mechanistic condition to this system's approach.

Representative experiment:
Below, we provide an example of the application of these methods. This study investigated rat neuronal and microglia gene expression in the solitary nucleus (NTS) in response to alcohol dependence and subsequent withdrawal¹⁶. Rat cohorts comprised 1) Control, 2) Ethanol-dependent (EtOH), 3) 8 h withdrawal (Wd), 4) 32 h Wd, and 5) 176 h Wd (Figure 1A). Following rapid decapitation, brainstems were separated from the forebrain and cryosectioned, and slices were stained for tyrosine hydroxylase-positive (TH +) neurons and microglia (Figure 1B). LCM was used to collect both TH+ and TH- neurons and microglia. All the cells were from the NTS and analyzed as samples of 10-cell pools. Four 96 x 96 microfluidic RT-qPCR dynamic arrays were run on the RT-qPCR platform measuring 65 genes (Figure 1B-C). Data were normalized using a -ΔΔC_t method and analyzed using R, and single-cell selection was validated with molecular markers (Figure 1D-E). Technical validation was further verified by technical replicates analyzed within a single batch and across batches (Figure 2 and Figure 3). TH+ and TH- neurons organized into different sub phenotypes with similar inflammatory gene clusters but differing γ-aminobutyric acid (GABA) receptor (R) clusters (Figure 4 and Figure 5). Sub phenotypes that had elevated expression of inflammatory gene clusters were over-represented at 32 h Wd while GABA-receptor (GABAR) expression remained low in protracted alcohol withdrawal (176 h Wd). This work contributes to the antireward hypothesis of alcohol and opioid dependence which conjectures that interceptive feedback from the viscera in withdrawal contributes to the dysregulation of visceral-emotional neuronal nuclei (i.e., NTS and amygdala) resulting in more severe autonomic and emotional sequelae, which contribute to substance dependence (Figure 6).

Protocol

This study was carried out in accordance with the recommendations of Animal Care and Use Committee (IACUC) of Thomas Jefferson University. The protocol was approved by Thomas Jefferson University IACUC.

1. Animal model

1. House male Sprague Dawley (>120 g, Harlan, Indianapolis, IN, USA) rat triplets individually with free access to ethanol-chow (2 rats) or control-chow mixture (1 rat).
   NOTE: This representative experiment employed the Lieber-DeCarli protocol to study the neurobiology of alcohol withdrawal^25,26. Rat triplets consist of a three-rat cohort to be fed the same number of calories but end up in different arms of the study at sacrifice. The three arms for this study are: 1) Control, 2) Ethanol-dependent (EtOH), and 3) Withdrawal (Wd) (Figure 1A). There are five total conditions in this study as there are three alcohol withdrawal time points (Figure 1A).
   1. Every other day, measure the ethanol-chow mixture consumed by the Wd rat and replace the amount consumed with the same amount of ethanol mixture to the EtOH rat feeder. Add the equivalent caloric amount of control mixture to the feeder of the Control rat.
      NOTE: The average daily ethanol consumption is around 12-16 g/kg after 3 weeks²⁶.
   2. Following stable long-term ethanol consumption and dependence (>5 weeks), induce acute alcohol withdrawal in Wd rat by emptying his or her food container and filling it with control mixture. Perform this in such a way that all three rats are sacrificed at the same circadian time point²¹.
   3. At the pre-chosen time point, sacrifice all three rats in the triplet at the same time point (Control, EtOH, and Wd).

2. Sample harvesting

1. Harvest brain at the same circadian time point for each rat triplet at proper Wd time point (8 h, 32 h, 176 h).
   1. Prepare a methanol and dry ice bath for rapid cooling of fresh tissue to preserve RNA integrity. Put to the side.
   2. Put the rat in isoflurane tank (5% in oxygen) for ~30 s or till loss of consciousness occurs, as indicated by the reduced respiratory rate and absence of motor activity. Put the rat head into a properly sharpened guillotine to rapidly decapitate.
   3. Open the animal's skull to dissect out the brain using forceps. Remove the cerebellum from the fresh brain with a hand-held razor by gross slicing and discard it. By transverse incision, slice the brainstem from the forebrain.
      NOTE: A hand-held razor can be used further to hemi-sect the forebrain or brainstem into left and right hemispheres according to the experimental design. For example, to validate findings from one hemisphere with different methods, investigate left-right divergence, or increase the sample number.
   4. Add optimal cutting temperature (O.C.T.) medium to approximately 3-4 cm depth in the tissue embedding mold so that the tissue sample can be fully immersed. Place the tissue, forebrain and/or brainstem, in the tissue embedding mold and add more O.C.T. to fully cover the tissue sample.
   5. Add the plastic tissue embedding mold containing the tissue sample (rat forebrain) into the cooling bath of dry ice and methanol to immediately freeze the tissue sample. Allow the tissue sample in the embedding mold to remain in the cooling bath until tissue collection is complete but not longer than 15 min.
      NOTE: Be careful to prevent methanol from spilling into the tissue container.
   6. Rapidly store tissue samples at -80 °C.
      ​NOTE: Microfluidic RT-qPCR measures gene expression by amplifying and measuring mRNA transcripts. These transcripts are relatively unstable, so many steps are taken in this process to keep the sample as cold as possible to prevent mRNA degradation.

3. Cryosectioning

NOTE: A rat neuronal nuclei is approximately 10 µm. Thus, 10 µm is the optimal slice thickness for this animal model. Slice thickness is adjusted according to the animal model for the study.

1. From the -80 °C freezer, take out the tissue embedding mold that contains frozen brainstem and thaw the sample at -20 °C in cryostat for 5-10 min. Perform this freeze-thaw to -20 °C only once for mRNA preservation.
2. With a hand-held razor, vertically cut the corners of the plastic embedding mold. Remove the brainstem embedded in O.C.T. from the plastic tissue embedding mold. On the chuck of the cryostat set at -20 °C, use room temperature liquid O.C.T. as a glue to mount the brainstem in the rostral to caudal direction for coronal cryosectioning.
3. Cut 10 µm coronal cryosections in the rostral to caudal direction from the rat brainstem until sections containing the region of interest (NTS) are reached. The height and width of these cryosections are ~200 mm based on the dimensions of the plastic embedding mold.
4. Collect the 10 µm cryosections of brainstem tissue that include the NTS, or other region of interest based on the study, by thaw-mounting onto room temperature plain glass slides. Quickly, place these slides with sections into a cooling metal pan that is situated on top of dry ice. As soon as possible, place glass slides containing cryosections into -80 °C freezer for storage.
   ​NOTE: A single glass slide can fit multiple 10 µm cryosections as the width and height is ~200 mm. Thus, the same slide can contain cryosections that are stained for distinct cell types.
   1. When several sections are on the slide, ensure that they are separated by 100 mm of empty space. To create a border between the cryosections, use a hydrophobic pen. This allows for different antibody solutions to stain for different cell types on the same glass slide. Maintain 20 mm of space for the cryosection from the edge of the glass slide.

4. Immunofluorescence staining of single cells

1. Use rapid staining immunofluorescence protocol on brain cryosections to label desired brain cell types (neurons, microglia, astrocyte, etc.) for collection using LCM.
   NOTE: The immunohistochemistry staining protocol used for these experiments was designed to be rapid to prevent excess mRNA degradation during this process.
   1. From the -80 °C freezer, take out glass slides containing 10 µm cryosections of rat brainstem containing NTS. Dip slides for 30 s into 75% ethanol bath to fix cryosectioned tissue. Remove excess liquid.
   2. To the fixed cryosectioned brainstem tissue, apply 2% Bovine Serum Albumin (BSA) in phosphate buffer saline (PBS) for 30 s for blocking untargeted binding of antibodies. Afterwards, wash with PBS.
   3. Apply enough amount of primary antibody solution to cover the cryosectioned tissue (now fixed and blocked). Incubate the tissue section in primary antibody solution for 2 min. Wash the tissue with 2% BSA solution once. The primary antibody solution contains 96% of the BSA-PBS solution for blocking, 3% primary antibody, and 1% RNase inhibitor. Primary antibody was diluted in a 1:25 ratio.
      NOTE: In this representative experiment, primary antibodies consist of anti-NeuN antibody and anti-Cd11β antibody. Neurons were further subdivided into TH+ and TH- subgroups, so primary antibody solutions for neuronal staining contained 93% of the BSA PBS solution, 3% anti-NeuN antibody, 3% anti-tyrosine hydroxylase antibody, and 1% RNase Out.
   4. Apply enough amount of secondary antibody (1:200) solution to cover the brainstem tissue. Let the secondary antibody solution bathe the tissue. After 3 min, wash the tissue with 2% BSA solution once.
      ​NOTE: The solution containing the secondary antibody was composed of 196.5 µL of 2% BSA, 1 µL of goat anti-mouse 555 nm fluorescent tag for cell type, 1 µL donkey anti-rabbit Alexa 488 nm for TH staining, 2.5 µL RNase inhibitor, and 1.3 µL of DAPI (1:10000).

5. Standard ethanol and xylene tissue dehydration series

1. Place glass slides upon which the stained cryosectioned tissue rests in a 75% ethanol bath for 30 s, followed by placing in 95% ethanol bath for 30 s, a 100% ethanol bath for 30 s, and finally a 100% ethanol both for 30 s (in a second container).
2. After completing the ethanol dehydration series, pour two fresh 100% xylene baths. Then, place the glass slides in the first xylene bath for 1 min. Rapidly remove and place the slides into the other fresh xylene bath for 4 min.
3. Take glass slides containing stained and dehydrated tissue cryosections out of xylene and place in a dark but ventilated container for 5 min to air dry. Following air drying, place glass slides in a desiccator for 5 min for final drying.

6. Select single cells using laser capture microdissection (LCM)

1. Insert glass slide containing stained, fixed, and dehydrated tissue cryosections into the LCM microscope sample holder. Use anatomic landmarks to locate the region of interest from which cell collection will take place (NTS in this representative example).
2. Turn on microscope fluorescence and find stained cell type(s) of interest. Ensure the nucleus of the desired cell is in the region of interest and is isolated from the nuclei of other cells by a distance >3 mm. Identify and mark one cell (if experiment is true single-cell) or multiple cells (if experiment calls for pooled single-cell samples [i.e., single-cell scale] as demonstrated in this representative experiment) to collect using the LCM software.
   NOTE: Samples comprised 10-cell pools were used in this representative experiment. This is done to decrease the gene expression variability between samples and increase the number of single-cells analyzed, although this still remains a single-cell scale experiment. True single cell experiments can be completed with this method^20,27. Additionally, larger tissue sections can also be selected²¹.
3. After the desired cells are selected, use the LCM robot arm, place the LCM cap on the region of interest on the marked tissue cryosection.
4. Calibrate the intensity and targeting of the infrared (IR) laser using test shots.
   NOTE: This is done for each sample. Calibration should be performed on a section of the cap that is not directly over the tissue so as not to mistakenly select non-experimental tissue.
   1. To calibrate the intensity, adjust the duration, strength, and size of the IR laser shot so that a shot only melts the cap adhesive enough to select a single cellular nucleus and is also strong enough to melt the cap to the cryosection on the slide. These values will be different for each cap.
   2. Calibrate targeting by localizing the laser crosshairs precisely where the laser melted the cap.
5. Gather the cells identified by firing the LCM infrared laser to melt the cap onto the cryosectioned tissue over those cells. Ensure that only selected cells were lifted from the tissue slice by using the robot arm to locate the cap to the quality control (QC) area of the LCM microscope. To remove any excess cells on the cap, use ultraviolet (UV) laser to obliterate the genetic material of the extra cells.
6. Record anatomic specificity with the LCM software camera. Snap a picture of the cryosectioned tissue from which the cell(s) was removed.
7. Use an anatomic atlas (an atlas of the rat forebrain in this example) to determine the distance of this tissue slice from bregma and record this information as the Z distance²⁸. Determine the location in the X and/or Y plane from the photo to fully localize the selected cell.
8. With gloved hands, remove the LCM cap from the QC area. Next, fasten the sample extraction device to the LCM cap. Use a pipette to apply 5.5 µL of lysis buffer to the selected cell(s). This is now referred to as the sample.
   NOTE: Lysis buffer solution is composed of 5 µL resuspension buffer along with 0.5 µL of lysis enhancer.
9. Fit a 0.5 mL microcentrifuge tube onto the sample extraction device that is attached to the LCM cap. Place this arrangement onto a 75 °C hotplate with the sample and lysis buffer closer to the plate. Let the sample heat for 15 min.
10. Use a low-speed centrifuge at 0.01-0.02 x g to spin the sample and lysis buffer down into the bottom of the 0.5 mL microcentrifuge tube. Store the sample at -80 °C until microfluidic qPCR.

7. Run qPCR chip on a microfluidic RT-qPCR on platform

1. To measure gene expression for single-cell samples, perform mRNA pre-amplification as described below.
   NOTE: The protocol does not involve mRNA extraction as the samples are single cells containing a very low amount of RNA (10 pg), which will be lost during the mRNA isolation step. The single cells are lysed and directly proceeded for reverse transcription to minimize the sample loss.
   1. Create a primer pool by adding the mRNA qPCR gene primers (forward and reverse) for every gene being assayed into a 1.5 mL microcentrifuge tube. Ensure that the final concentration of each primer is 500 nM. Primers in this example experiment are listed in Supplementary Table 1¹⁶.
   2. Pipette 1 µL of 5x cDNA reaction mix into every well of a 96-well PCR plate.
   3. Let single-cell samples thaw briefly (2-3 min) by removing them from the -80 °C freezer and letting them sit at room temperature. Use a low-speed centrifuge at 0.01-0.02 x g to spin down the samples for 20-30 s. Pipette 5.5 µL of each single-cell sample to a different well in the 96-well PCR plate.
      NOTE: The sample number and type to be put into each well is determined before beginning this step.
   4. The 96-well PCR plate now has the cDNA reaction mix and a single-cell sample added to each well. Place this plate into a thermocycler for 1.5 min at 65 °C to activate the cDNA reaction mix. Spin down the contents using high-speed centrifugation at 1,300 x g for 1 min at 4 °C and put the PCR plate on wet ice.
   5. Into each well in the PCR plate, pipette 0.12 µL of T4 Gene 32 Protein, 0.73 µL of DNA suspension buffer, and 0.15 µL of 10x cDNA synthesis master mix. Run the PCR plate through the following protocol in the thermocycler: 25 °C for 5 min, 50 °C for 30 min, 55 °C for 25 min, 60 °C for 5 min, 70 °C for 10 min, and a final hold at 4 °C.
      NOTE: T4 Gene 32 protein (a single-stranded DNA (ssDNA) binding protein) is required for bacteriophage T4 replication and repair. The T4 Gene 32 protein was used in the reverse transcription step of the protocol to improve the yield and efficiency of reverse transcription and to increase the PCR product yield.
   6. Into each well in the PCR plate, pipette 7.5 µL of Taq polymerase master mix. Following this, in each well, pipette 1.5 µL of the primer pool created in step 7.1.1. Run the PCR plate through the following linear preamplification step in the thermocycler: 95 °C for 10 min; 22 cycles of: 96 °C for 5 s and 60 °C for 4 min.
   7. Into each well in the PCR plate, pipette 1.2 µL of exonuclease I, 4.2 µL of DNA suspension buffer and 0.6 µL of 10x exonuclease I reaction buffer. Run the PCR plate through the following protocol in the thermocycler: 37 °C for 30 min, 80 °C for 15 min.
      NOTE: Exonuclease catalyzes the removal of nucleotides from linear single-stranded DNA in the 3' to 5' direction. We use exonuclease for sample cleanup for the removal of unincorporated primers and any single-stranded cDNA that may be present following pre-amplification.
   8. Into each well in the PCR plate, pipette 54 µL of TE buffer. Use a high-speed centrifuge at 1,300 x g for 5 min at 4 °C to spin down contents of the PCR plate.
   9. If planning on continuing with the next phase of the protocol, refrigerate the PCR plate at 4 °C. If more than 12 h are there between completion of the pre-amplification phase and commencement of microfluidic qPCR, cover the qPCR plate and place it into a -20 °C freezer.
2. Make the sample plate (new 96-well PCR plate) for qPCR chip as described below.
   1. If the pre-amplification PCR plate from step 7.1 was placed in a -20 °C freezer, remove and let thaw at room temperature for 10 min.
   2. Into a new 96-well PCR plate, pipette into each well 4.55 µL of PCR supermix low rox and 0.45 µL of 20x DNA binding dye. Then, pipette 3 µL of pre-amplified sample from the PCR plate made in step 7.1. Use high-speed centrifugation at 1,300 x g for 5 min at 4 °C to spin down the PCR plate and place the PCR sample plate on ice.
3. Make an assay plate (a new 96-well PCR plate) for qPCR chip as described below.
   1. Into a new 96-well PCR plate, pipette into each well 1.25 µL of DNA suspension buffer and 3.75 µL of 2x assay loading reagent. Then, pipette the corresponding qPCR primer at 2.5 µL of 10 µM primer. Use high-speed centrifuge at 1,300 x g for 5 min at 4 °C to spin down the PCR plate and place the PCR assay plate on ice.
4. Prepare the qPCR chip for loading into the microfluidic RT-qPCR platform as described below.
   NOTE: The qPCR chip is a real-time qPCR platform that works on microfluidic principles. The qPCR chip is essentially a matrix of 96 sample channels and 96 RNA qPCR primer channels (i.e., assays) which intersect in 9,216 (96 x 96) chambers. A sample and a specific assay are combined in each chamber of the array in which a real-time qPCR reaction occurs. The readout is generated as threshold cycle (Ct) values and amplification plots for the primers used.
   1. Inject control line fluid into the qPCR chip for priming. Insert the qPCR chip into the microfluidic mixing device. Select the prime (136x) script and run this program.
   2. When the program is complete after ~45 min, remove the primed qPCR chip. Into the primed qPCR chip, pipette 6 µL of the reaction from the PCR sample plate into the corresponding sample well in the qPCR chip.
   3. Into the primed qPCR chip, pipette 6 µL of the reaction from the PCR assay plate into the corresponding assay well in the qPCR chip.
      NOTE: Air bubbles may form at the bottom of the sample and assay wells in the qPCR chip, which can prevent the solutions entering the microfluidic conduits. A sharp needle can be used to pop or remove these air bubbles before inserting the qPCR chip into the microfluidic mixing device.
   4. Insert the qPCR chip into the microfluidic mixing device. Select the load mix (136x) script and run this program.
5. Load qPCR chip into the microfluidic RT-qPCR platform.
   1. Turn on the microfluidic RT-qPCR platform and warm up the bulb (~20 min). Remove the qPCR chip from the microfluidic mixing device. Peel the protective sticker from the bottom of the qPCR chip.
   2. Open the microfluidic RT-qPCR platform and load the qPCR chip into the microfluidic RT-qPCR platform. Run the fast 96 x 96 PCR protocol (30 cycles) on the microfluidic RT-qPCR platform.
   3. Launch the Data collection software-click on Start a New Run.
   4. Verify the chip barcode and chip type-click on Next. Click on the Chip Run file and browse the file location for the data collection storage.
   5. Click on Application Type and select Gene Expression; select ROX for passive reference, select Single Probe and select EvaGreen for probe type.
   6. Click to select the thermal cycling program and select Biomark HD: GE Fast 96x96 PCR+Melt v2.pcl file.

8. Data analysis

1. Upload data from the qPCR chip run to the analysis software for QC as described below.
   1. Download the data analysis software. This can be found on the following website: https://www.fluidigm.com/products-services/software#. Launch the software to analyze the microfluidic RT-qPCR experiment.
   2. Within the software, open Chip Run. Then, open the ChipRun.bml file that was created by the experiment. A window will appear with the experimental details, including the passive reference, probe, and the PCR thermal program.
      NOTE: Quality control (QC) and data analysis can be done on the computer attached to the microfluidic RT-qPCR platform or by transferring the .bml file to a different computer via a flash drive or other means.
2. Define the sample setup as described below.
   NOTE: This step creates a labeled template of samples and assays (qPCR primers) for the quality control (QC) and data analysis procedures.
   1. Select Chip Explorer > Sample Plate Setup and create a new sample plate. Select the appropriate container type and container format (SBS96 for a 96-well plate used in this representative experiment).
   2. Paste the sample labels into the software spreadsheet as per the experimental design and select Map from the Task menu. Select SBS96-Left.dsp. The sample setup is now mapped. Select Detail View > Analyze to update these changes in the file.
      NOTE: Any changes made in this software will not be saved until the Analyze button is clicked.
3. Define the control samples in the software. Select the cells for the control samples by left click and hold while dragging through the cells for selection. Individual cells can be selected by pressing and holding the Ctrl key and clicking on individual cells. For the RNA standard samples (dilution series), enter the sample name and the RNA concentration used. Click on Analyze to save.
4. Define the detector set up similar to step 8.2 above but with RT-qPCR assay names, i.e., the names of the genes assayed. Select Detector Plate Setup and create a new assay plate. Select the appropriate container type and format (SBS96 for a 96-well plate). Paste the assay names as per the experimental design and click on Analyze to save.
5. Begin the user quality control (QC) process as described below.
   NOTE: This representative experiment used a cycle time (Ct) thresholding method. That is, samples that did not reach a threshold signal defined by the software (Auto Global) or manually (User Global) will be deemed failed reactions and not included in the dataset.
   1. Select Analysis View > Task Frame. In the analysis settings pane, change the settings to Auto Global (automatically calculates a threshold that is applied to the entire chip) or User Global. Set the baseline correction to linear derivative. If User Global is used, the failure threshold must be determined manually as in the representative experiment.
   2. This representative experiment used a QC rule as follows: if an individual assay or sample had a failure rate of 70% or greater, then that entire row or column in the dataset was failed.
   3. Manually review each reaction in the 96 x 96 chip. Visualize the amplification curves and melt curves to deem if each reaction followed the expected qPCR pattern. If the amplification or melt curve does not match with what is expected, fail that reaction.
6. Following QC, export the data by selecting File > Export and save the dataset as a .csv file.
7. Prune the dataset as the exported .csv file has both a Pass-Fail matrix and a matrix with raw Ct values. Use the Pass-Fail matrix to replace any failed cells in the data set with NA.
8. Download the most updated version of open-source R software, and then download the R-Studio application. Upload the normalized dataset into R. Analyze data as fit for the study.
9. Normalize the dataset using the -ΔΔC_t method as described below.
   NOTE: Median centering or housekeeping gene normalization can be used across a sample row to generate a -ΔC_t value. In the representative experiment, median centering was used and validated by housekeeping gene normalization. This can be done in .csv format or by uploading to a data analysis software.
   1. For median centering, calculate the median Ct value calculated from all the Ct values for an individual sample (10-celled pool). Then, subtract all individual Ct values from this median value. This yields a -ΔC_t value.
   2. For housekeeping gene normalization, calculate the average expression of the housekeeping genes (Actb, Gapdh, and Ldha in the representative experiment) for each sample and use this value as the one from which to subtract the individual Ct values.
   3. To generate a -ΔΔC_t value, take the median of each assay column, which now contains the -ΔC_t value. Subtract the assay median from each -ΔC_t value to produce a -ΔΔC_t value.
10. Calculate z-scores for each -ΔΔC_t value using the scale function in R. Use the R heat map function or a separate software to generate a heat map.
11. Use the Pearson correlation function to calculate Pearson correlations between each gene. Use melt function in R to organize the dataset for other functionality. Export this data and upload it into a gene correlation network software.

Representative Results

Validation of single-cell collection is performed visually during LCM procedures. Cell nuclei are assessed at the QC station. The cell type can be determined by emission of tagged fluorophore for that cell type and its general morphology. If non-desired cells have been selected on the cap, their genetic material can be destroyed with a UV laser at the QC station. Further validation by molecular analysis is also necessary. In this representative example¹⁶, two types of neurons were selected-tyrosine hydroxylase (Th) positive (+) and Th negative (-), in addition to microglia. Figure 1D demonstrates that samples meant to be neurons demonstrated statistically significant elevated expression of the neuronal marker, NeuN. Concurrently, microglial samples had significant elevation of the microglial markers, CD34 and Cx3xr1, suggesting that sample selection was performed with high fidelity. Additionally, Th+ neuronal samples demonstrated significantly elevated expression of Th compared to Th- neuronal samples and microglia samples while Th- neurons demonstrated significantly elevated expression of GCG (a preproglucagon coding gene) suggesting that these Th- neuronal samples are enriched with neurons that use GLP-1 as a neurotransmitter (Figure 1D). A more global computational measure known as linear discriminate analysis showed that these three cell types had differing gene expression profiles across all 65 genes measured with neurons and microglia separating along the x-axis and Th+ and Th- neuronal samples dividing along the y-axis (Figure 1E).

The quality of the data was also validated with intra and inter-batch technical replicates (Figure 2 and Figure 3). Intra-batch replicates control for the quality of data from that specific batch. If intra-batch replicates do not align, another qPCR chip experiment can be performed from the same sample and assay plates that were made to run the first batch. Inter-batch replicates ensure that data from across batches can be compared. This is crucial for experiments in which large numbers of samples are assayed requiring multiple batches such as this representative example. There was one outlier in this experiment as sample 40 in batch 4 (Figure 3). However, sample 40 in batches 1, 2, and 3 correctly aligned, and the other replicates from batch 4 aligned with batches 1, 2, and 3 suggesting that this was an isolated bad reaction, which occurred in this one replicate. Comparing across batches also requires proper data normalization techniques. This experiment used median centering, though housekeeping genes (Actb, Gapdh, Ldha), were also assayed in this experiment and served as a control for the median-centering method. That is, both methods yielded highly analogous datasets further suggesting high data quality.

Figure 4 displays a heat map of GLP-1 enriched neuronal samples that is organized by cellular subphenotype. This figure not only demonstrates the importance of cellular subphenotypes in neurobiology, but also demonstrates how ratios of subphenotypes change through alcohol withdrawal. The heat map shows that subphenotype A, which highly expresses the inflammatory gene cluster 1, increases in ratio at the 8 h Wd time point and reaches a maximum at 32 h Wd. By 176 h Wd, this inflammatory subphenotype is normalizing its prevalence back to control. Subphenotype B, which highly expresses GABAR gene cluster 2, demonstrates an overall suppression in the expression of this gene cluster by the 176 h Wd condition. These findings demonstrate how this GLP-1 neuronal cell type becomes hyperexcitable by increasing its inflammatory subphenoype and concurrently decreasing the level of expression of GABARs in its GABA subphenotype during alcohol withdrawal. Figure 5 combines the gene expression of individual samples into averages so that the expression of clusters of genes and the location of the protein of that gene transcript can be visualized throughout the time series.

Figure 1: Experimental design and single-cell selection. (A) Rat triplets were randomly assigned to one of the five treatment conditions. (B) Single-cell transcriptome data generation. (C) Cartoon representation of genes assayed and their function. Genes in green were not assayed. Official gene symbol is used. (D) Gene expression of cell type markers. Error bars show standard error. Neurons compared to microglia, p-values = 0.0273, 3.94 x 10^-10, 7.73 x 10^-12, respectively. Th+ neurons showed elevated Th expression compared to Th- neurons (p = 4.56 x 10^-11) and microglia (p = 2.95678 x 10^-15). Th- neurons showed elevated expression of GCG compared to Th+ neurons (p = 0.0106) and microglia (p = 0.0435) indicating they are GLP-1+ neurons. *p < 0.05, ***p < 4 x 10^-10. (E) Linear discriminate analysis of all samples displays the difference across all genes measured between the three cell types collected in a two-dimension space. Centroid distance between NE neurons and GLP-1 neurons = 3.30, NE neurons and microglia = 1.57, GLP-1 Neurons and microglia = 2.92. This figure has been modified from O'Sullivan et al. 2021¹⁶. Please click here to view a larger version of this figure.

Figure 2: Technical replicate plots of raw C_t values within a batch. Each graph displays raw C_t values for intrachip technical replicates plotted against each other demonstrating technical experimental integrity. This figure has been modified from O'Sullivan et al. 2021¹⁶. Please click here to view a larger version of this figure.

Figure 3: Technical replicate plots of raw C_t values between batches. Each graph displays raw C_t values for interchip technical replicates plotted against each other demonstrating that all batches are comparable to each other. This figure has been modified from O'Sullivan et al. 2021¹⁶. Please click here to view a larger version of this figure.

Figure 4: Heat map of GLP-1 enriched neuronal samples. Heat map displays cellular subphenotypes of GLP-1 enriched neuron samples through an alcohol withdrawal time series. Rows represent 10-cell pooled samples with cellular subphenotype clusters labeled with uppercase letters. Columns represent the z-score of -ΔΔC_t gene expression values on a -1 to +1 color scale for that gene in that sample. Gene clusters are labeled by number. This figure has been modified from O'Sullivan et al. 2021¹⁶. Please click here to view a larger version of this figure.

Figure 5: Suphenotype gene expression in GLP-1 enriched neuronal samples. Cellular diagrams display boxes representing relative gene expression (average z-score of -ΔΔC_t values) of subphenotypes shown in the heatmap in Figure 4. The diagrams were constructed using Cytoscape version 3.8.0 and z-scores were calculated using the scale function in R version 3.5.2. -ΔΔC_t labels on the right show which boxes correspond to which gene and the color represents expression (blue is low expression and yellow is high expression). The location of the box represents the localization or function of the protein product from that gene transcript. Green numbers indicate subgroups within subphenotypes. This figure has been modified from O'Sullivan et al. 2021¹⁶. Please click here to view a larger version of this figure.

Figure 6: Interoceptive vagal circuit, visceral-emotional neuraxis, and schematic of opponent-process model of addiction. (A) Interoceptive vagal afferents relay the state of the gut, which is highly influenced by gut microflora, and other peripheral organs to the nucleus tractus solitarius (NTS). This information is subsequently relayed to the amygdala and influences emotional states. (B) A simplified cartoon representation showing the integrative roles of the nucleus tractus solitarius (NTS) and the central nucleus of the amygdala (CeA) in emotion, stress, and autonomic regulation. Two neuronal subtypes, GLP-1 and NE neurons are highlighted. Many anatomical and functional connections are omitted for clarity. Abbreviations: NE = norepinephrine; GLP-1 = glucagon-like peptide 1; GABA = γ-aminobutyric acid; HPA axis = hypothalamic-pituitary-adrenal axis; CRF = corticotropin releasing hormone. (C) Alcohol and/or opioid exposure has two actions: stimulate reward, via the mesolimbic dopamine pathway, and inhibit antireward. These actions motivate substance use via positive reinforcement. (D) Alcohol and/or opioid withdrawal has two actions: inhibit reward, by inhibiting the mesolimbic dopamine pathway (not shown) and stimulate antireward. This study proposes that visceral-emotional neuroinflammation is an endpoint in antireward stimulation, though this hypothesis warrants further testing. These actions, whatever the mechanism, motivate substance dependence via negative reinforcement. This figure has been modified from O'Sullivan et al. 2021¹⁶. Please click here to view a larger version of this figure.

Supplementary Table 1: Primers for microfluidic RT-qPCR. All primer pairs for mRNA amplification are listed along with the amplicon length formed by each primer pair. This table has been modified from O'Sullivan et al. 2021¹⁶. Please click here to download this File.

Discussion

Alcohol use disorder remains a challenging disease to treat. Our group has approached this disorder by investigating antireward processes with a systems neuroscience perspective. We measured gene expression changes in single NTS neurons and microglia in an alcohol withdrawal time series¹⁶. The NTS was chosen for its prominent role in the autonomic dysregulation that occurs in alcohol withdrawal syndrome. We combine LCM with single-cell microfluidic RT-qPCR allowing for robust numbers of samples and genes to be measured at low cost with anatomic and molecular sensitivity and specificity (Figure 1B-C). Single-cells were identified and selected by IHC staining, and these cellular subphenotype groupings were validated with molecular expression (Figure 1D-E). However, some GLP-1 enriched neuronal samples (TH- neurons) moderately expressed Th (Figure 4). The establishment of cell phenotype was made at cell selection based on fluorescent visualization, and this criterion was used for the study as molecular grouping based on the expression of a single gene, Th, failed to define expression patterns suggestive of a cellular phenotype¹⁶. Thus, the results characterize major NTS microglial and neuronal subphenotypes and their dynamics in alcohol withdrawal.

This manuscript describes a method for single-cell transcriptomics, and it is imperative that all the steps in the protocol are followed as described. Some modification may be required when working with a different organ tissues or cell types. One of the challenges of this protocol is maintaining RNA quality and integrity during LCM. We have established protocols for IHC staining, LCM, and microfluidic RT-qPCR to address the challenges as detailed above. Briefly, these processes include the addition of RNase inhibitor to all staining solutions to prevent RNA degradation, a fast-staining protocol to prevent possible RNA degradation, processing slides for LCM immediately after IHC staining and dehydration and maintaining proper cold conditions whenever possible during sample processing or transfer.

The transcriptomics assay is performed on lysed single cells without RNA isolation, followed by reverse transcription, pre-amplification, and qPCR. The pre-amplification step is to amplify cDNA molecules to detectable levels for qPCR selectively. There are several limitations to the pre-amplification step. These include gene transcripts that fail to be amplified, which will result in no detection in microfluidic RT-qPCR. There is also the possibility of primer dimer formation as the PCR reaction contains all primers for the experimental assay. That is, the formation of dimers with a forward and reverse sequence of a single primer pair and with all the forward and reverse sequences in the experiment is possible. Hence, primer testing using positive experimental control like RNA standards is advisable.

The RT-qPCR chip design is another crucial aspect of this protocol for quality control and batch effects. Positive controls consist of standard RNA from the animal and organ being assayed (rat brain in this representative example). An RNA dilution series can then be loaded into each chip and should demonstrate the appropriate quantitative signal for each gene assayed. Water can be used as a negative control. These controls are critical for proper normalization that controls for batch effects which may occur when combining data from different chips. This is discussed in detail in step 8.

In the representative experiment shown here, these methods were applied for hypothesis-generating and providing insight into possible mechanism. The study was performed in the context of another work and provides important information for the interceptive antireward hypothesis¹³. In brief, this model conjectures that antireward is stimulated in substance withdrawal by interoceptive signaling from the vagus via the NTS to the amygdala and that the initial substrate of this antireward is neuroinflammation (Figure 6). This work supports this hypothesis by establishing the inflammatory changes in gene expression in neurons and microglia upon alcohol withdrawal.

A major weakness of this study is its lack of mechanistic claims. However, these methods can be used to determine a mechanism after a baseline experiment such as this one is performed. For example, an animal model intervention or perturbations, such as a subdiaphragmatic vagotomy or gene knockout, can demonstrate anatomic or genetic mechanisms of inflammatory regulators. Future studies that take this approach can contribute to the development of the interoceptive antireward hypothesis, which aims to implement these insights into clinical practice for addiction treatment.

Materials

Name	Company	Catalog Number	Comments
20X DNA Binding Dye	Fluidigm	100-7609	NA
2x GE Assay Loading Reagent	Fluidigm	85000802-R	NA
96.96 Dynamic Array IFC for Gene Expression (referred to as qPCR chip in text)	Fluidigm	BMK-M-96.96	NA
Anti-Cd11β Antibody	Genway Biotech	CCEC48	Microglia Stain
Anti-NeuN Antibody, clone A60	EMD Millipore	MAB377	Neuronal Stain
Anti-tyrosine hydroxylase antibody	abcam	ab112	Stain for TH+ neurons
ArcturusXT Laser Capture Microdissection System	Arcturus	NA	NA
Biomark HD	Fluidigm	NA	RT-qPCR platform
Bovine Serum Antigen	Sigma-Aldrich	B4287
CapSure Macro LCM Caps	ThermoFisher Scientific	LCM0211	NA
CellDirect One-Step qRT-PCR Kit	ThermoFisher Scientific	11753500	Lysis buffer solution components
CellsDirect Resuspension & Lysis Buffer Kit	ThermoFisher Scientific	11739010	Invitrogen
DAPI	ThermoFisher Scientific	62248	Nucleus Stain
DNA Suspension Buffer	TEKnova	T0221
Donkey anti-Rabbit IgG (H+L) ReadyProbe Secondary Antibody, Donkey anti-Rabbit IgG (H+L) ReadyProbe Secondary Antibody, Alexa Fluor 488	ThermoFisher Scientific	R37118	Seconadry Antibody
Exonuclease I	New Englnad BioLabs, Inc.	M0293S	NA
ExtracSure Sample Extraction Device	ThermoFisher Scientific	LCM0208	NA
FisherbrandTM Superfrost Plus Microscope Slides	ThermoFisher Scientific	22-037-246	Plain glass slides
GeneAmp Thin-Walled Reaction Tube	ThermoFisher Scientific	N8010611
Goat anti-Mouse IgG (H+L), Superclona Recombinant Secondary Antibody, Alexa Fluor 555	ThermoFisher Scientific	A28180	Seconadry Antibody
IFC Controller	Fluidigm	NA	NA
RNaseOut	ThermoFisher Scientific	10777019
SsoFast EvaGreen Supermix with Low Rox	Bio-Rad	PN 172-5211	NA
SuperScript VILO cDNA Synthesis Kit	ThermoFisher Scientific	11754250	Contains VILO and SuperScript
T4 Gene 32 Protein	New Englnad BioLabs, Inc.	M0300S	NA
TaqMan PreAmp Master Mix	ThermoFisher Scientific	4391128	NA
TE Buffer	TEKnova	T0225	NA
TempPlate Semi-Skirted 96-Well PCR Plate, 0.2 mL	USA Scientific	1402-9700	NA