Combining Laser Capture Microdissection and Microfluidic qPCR to Analyze Transcriptional Profiles of Single Cells: A Systems Biology Approach to Opioid Dependence

Abstract

Profound transcriptional heterogeneity in anatomically adjacent single-cells suggests robust tissue functionality may be achieved by cellular phenotype diversity. Single-cell experiments investigating the network dynamics of biological systems demonstrate cellular and tissue responses to various conditions at biologically meaningful resolution. Herein, we explain our methods for gathering single-cells from anatomically specific locations and accurately measuring a subset of their gene expression profiles. We do this by combining laser capture microdissection (LCM) with microfluidic reverse transcription quantitative polymerase chain reactions (RT-qPCR). We also use this microfluidic RT-qPCR platform to measure the microbial abundance of gut contents.

SUMMARY:

This protocol explains how to collect single neurons, microglia, and astrocytes from the central nucleus of the amygdala with high accuracy and anatomic specificity using laser capture microdissection. Additionally, we explain our use of microfluidic RT-qPCR to measure a subset of the transcriptome of these cells.

INTRODUCTION:

Measuring the gene expression profiles of single cells has demonstrated extensive phenotypic heterogeneity within a tissue. This complexity has complicated our understanding of the biological networks that govern tissue function. Our group and others have explored this phenomenon in many tissues and conditions ^1–6. These experiments not only suggest that regulation of gene expression networks underlie such heterogeneity, but also that a single-cell resolution reveals a complexity in tissue function that tissue-level resolution fails to appreciate. Indeed, merely a small minority of cells may respond to a specific condition or challenge, but the impact of those cells on overall physiology may be substantial. Additionally, a systems biology approach that applies multivariate methods to high dimensional datasets from multiple cell-types and tissues can elucidate system-wide treatment effects.

We combine laser capture microdissection (LCM) and microfluidic reverse transcription quantitative polymerase chain reactions (RT-qPCR) to obtain such datasets. We take this approach here in contrast to gathering single-cells via fluorescence-activated cell sorting (FACS) and using RNA sequencing (RNA-seq) to measure their transcriptome. The advantage of LCM over FACS is that the exact anatomic specificity of single-cells can be documented with LCM—relatively and absolutely. Microfluidic RT-qPCR is less expensive and has a higher sensitivity and specificity as compared to RNA-seq ⁷. However, RNA-seq can measure far more features.

In this representative experiment, we investigated the effects of opioid dependence and naltrexone-precipitated opioid withdrawal on rat neuronal, microglia, and astrocyte gene expression in the central nucleus of the amygdala (CeA) and gut microflora abundance ⁴. Four treatment groups were analyzed: 1) Placebo, 2) Morphine, 3) Naltrexone and 4) Withdrawal (Fig 1). We found that opioid dependence did not substantially alter gene expression, but that opioid withdrawal induced the expression of inflammatory genes—Tnf in particular. Astrocytes were the most effected cell type. The gut microbiome was profoundly impacted by opioid withdrawal as indicated by a decrease in the Firmicutes to Bacteroides ratio: an established marker of gut dysbiosis ^8,9.

Figure 1. Single-cell RT-qPCR workflow and transcriptional heterogeneity.

Modified from O’Sullivan et al. 2019 (A) Experimental protocol (n=4 for each condition) (B) 10-cell pooled sample transcriptome measurement. (C) Bar plot displays median −ΔΔCt expression values. Neurons are purple, microglia are yellow, astrocytes are green. Error bars show standard error. *p<0.05, ***p<0.0003 Tukey’s honest significance test. (D) Heat map shows expression of all samples across 40 assayed genes. Rows are 10-cell pooled samples (930 neuronal samples, 950 microglial samples, 840 astrocyte samples as denoted ); numbers denote sample clusters) and columns are genes.

PROTOCOL:

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

1. Animal model

1.1 Insert 2 75 mg slow-release morphine sulfate pellets or 2 placebo pellets subcutaneously in adult male Sprague-Dawley rats.

1.1.1 Gown and glove appropriately for minor sterile surgery. Shave rate dorsum with clippers if necessary.

1.1.2 Apply vet ointment to eyes of animal. Anesthetize rat with approximately 20 s of isoflurane inhalation. Anesthesia confirmed with loss of consciousness.

1.1.3 Make a midline incision in the rat dorsum with bead-sterilized blunt scissors and separate dermis from body wall with bead-sterilized probe. Insert pellets under dermis with bead-sterilized forceps. Suture the incision closed with sterile needle.

Note: Entire procedure takes about 5 min per rat. Fresh sterile gloves are used for each rat.

1.1.4 Place rat into isolation cage for post-surgical recovery. Ensure heart beat and regular respiratory rhythm. Observe until consciousness is regained. Assess for post-surgical pain.

1.1.5 Assess rats 8 h following surgery and every 12 h following surgery for recovery and infection. Place rats in cage with rest of cohort when fully recovered from surgery—about 24 h.

1.2 Give intraperitoneal naltrexone injection (75 mg/Kg) to Naltrexone and Withdrawal cohort following 6 days of morphine exposure.

Note: There were 4 rat cohorts in this representative experiment (see Figure 1).

2. Sample Harvesting

2.1 Harvest brain 6 days following pellet insertion or 24 h following naltrexone injection.

2.1.1 Place animal in isoflurane chamber for approximately 30 s or until loss of consciousness occurs indicated by lack of motion and decreased respiratory rate.

2.1.2 Use sharp guillotine to rapidly decapitate animal.

2.1.3 Dissect out brain from animal skull.

2.1.4 Use a sharp hand-held razor to make the following gross incisions to the removed brain. Slice off the cerebellum and discard. Separate brainstem from forebrain with a transverse incision. Hemisect forebrain and/or brainstem with a midline sagittal incision.

2.1.5 Place forebrain and brainstem into plastic tissue embedding mold with 3-4 cm of Optimal Cutting Temperature (O.C.T.) already in bottom of mold. Cover rest of sample with O.C.T.

2.1.6 Immediately place plastic tissue embedding mold with sample covered by O.C.T. into bath containing dry ice and methanol. Do not let methanol spill into tissue embedding mold. Keep embedding mold with brain sample in methanol-ice bath until tissue collection is finished (approximately 10-15 min maximum).

2.1.7 Place brain sample into -80° Celsius (C) freezer as soon as possible.

2.2 Harvest gut samples concurrently

2.2.1 Following rapid decapitation, make midline incision in animal abdomen with scalpel.

2.2.2 Find cecum and sever its connection to the ascending colon.

2.2.3 Squeeze cecal contents into 15 mL conical tube.

2.2.4 Immediately place conical tube on dry ice and put into −80° C freezer as soon as possible.

Note: small intestine contents can also be collected by same methods as a negative control.

3. Slicing

3.1 Slice hemisected rat forebrain using a cryostat.

3.1.1 Remove plastic embedding mold with forebrain from −80° C freezer and place into −20° C cryostat.

3.1.2 Remove O.C.T. embedded hemisected forebrain sample from embedding mold. Use razor to slice the corners of the plastic embedding mold vertically if necessary. Mount forebrain for rostral to caudal coronal slicing on cryostat chuck using O.C.T.

Note: Anatomic landmarks to identify the CeA include the optic tract and stria terminalis (Figure 2B). The optic tract branches from the optic chiasm and tracks dorsal-lateral as the brain is sliced rostral to caudal. When the optic track has a morphology similar to what is seen in rat brain atlas bregma −2.12 mm¹⁰, test slices may be viewed under the microscope. Optic tract and stria terminalis morphology can be referenced in a rat brain atlas¹⁰ to determine the bregma and if CeA surrounds the stria terminals.

Figure 2. Laser Capture Microdissection Images.

(A) Four slices of dehydrated hemisected rat forebrain containing CeA on a slide. Slices are were placed on slide exactly 10 μm from previous slice. Left is anterior and right is posterior. Distance from bregma can be estimated using rat brain atlas and landmarks including the optic tract and stria terminalis. (B-D) Sequence of images showing the selection of single-cells in the CeA (C) and their removal from the tissue (D). Multiple LCM caps were used to select these cells. 1 cap is used to pick 10 cells of one cell type.

3.1.3 Slice 10 μm thick coronal sections from hemisected forebrain rostral to caudal until sections containing the central nucleus of the amygdala (CeA) are reached.

Note: Width and height of sections are approximately 200 mm.

3.1.4 Collect 10 μm sections containing the CeA, or preferred brain region, by thaw-mounting 10 μm sections onto plain glass slides. Immediately place glass slides onto metal pan resting on dry ice. Put slides with brain sections into −80° C freezer as soon as possible

Note: Multiple slices may be placed on same slide. If using a different cell type stain for slices on the same slide, leave about 100 mm between slices so hydrophobic pen can be used to separate cell type-specific antibody solutions on the slide. Leave about 20 mm from the edge of the slide on each side of the slice.

4. Immunofluorescence staining

4.1 Stain forebrain sections for brain cell of choice (neuron, microglia, astrocyte, etc.) using immunofluorescence

4.1.1 Remove one or multiple slides with 10 μm sections of CeA from −80° C freezer

Note: Can use hydrophobic pen to separate different slices on the same slide to allow for staining of different cell types on the same slide

4.1.2 Fix slides with 75% ethanol for 30 s. Remove excess liquid.

4.1.3 Block slices for 30 s with 2% bovine serum antigen (BSA) in phosphate buffer saline 1x (PBS). Wash 1x with PBS.

4.1.4 Add primary antibody solution to slide for 2 min. Wash 1x with 2% BSA solution.

Note: Primary antibody solution is composed of 2% primary antibody, 1% RNase Out, and 96% of same BSA PBS solution for blocking step above. Representative experiment used anti-NeuN antibody, anti-Cd11β antibody, or anti-GFAP antibody in the following quantities: 3 μL of primary, 1.88 μL of RNA inhibitor, and 145.12 μL of BSA solution.

4.1.5 Add secondary antibody solution to slide for 3 min. Wash 1x with PBS.

Note: Secondary antibody solution is composed of 1 μL of goat anti-mouse 488 nm fluorescent tag (1:500), 2.5 μL RNA inhibitor, 1.3 μL of DAPI (1:10000), and 196.5 μL of 2% BSA.

5. Ethanol and xylene dehydration series

5.1 Dip slides into 75% ethanol for 30 s. Immediately following this, dip slides in 95% ethanol for 30 s. Immediately following this, dip slides into 100% ethanol for 30 s. Immediately following this, dip slides into a second container containing 100% ethanol for 30 s.

5.2 Following the ethanol dehydration series, dip slides into freshly poured xylene for 1 min. Immediately following this, dip slides into second container of xylene for 4 min.

5.3 Removed slides from xylene bath and let air dry in the dark for 5 min.

5.4 Place slides in desiccator for 5 min to further dry.

6. Laser Capture Microdissection

6.1 If stained, place slide into microscope and find region of interest (CeA) using anatomic landmarks (optic chiasm and stria terminalis).

6.2 Use fluorescence to identify stained cell type and its nucleus in region of interest. Choose one cell or multiple cells if doing single-cell pooled samples. Mark cells of interest using LCM software.

6.3 Place LCM cap on top of slice on region of interest.

6. 4 Use test shots of infrared (IR) laser to adjust IR laser strength, size, and duration so that the LCM cap adhesive melts only over the area of the selected single-cell. This ensures no other cells will be collected other than the cells selected.

Note: In this representative experiment, 10-cell pools of the same cell type were used as a single sample to limit the cell-to-cell variability in gene expression between samples of the same treatment. However, this method can be used for true single-cell experiments.^1,3.

6.5 Select individual cells to be collected for analysis using LCM software tools (Fig. 2C). Cells selected must be in the anatomic area of the CeA (or brain region of choice) based on rat brain atlas and bregma¹⁰. Cells should be at least 3 μm from adjacent stained nuclei.

6.6 Fire IR laser to collect identified single cells.

6.7 Place cap in quality control (QC) station and view it to observe that only desired cells were selected. If other cells were mistakenly selected, ultraviolet laser can be used to destroy these cells while cap remains in QC station.

6.8 Take a photo of the tissue section from where to the cell was collected to document anatomic specificity. Record distance of slice from bregma if appropriate by reference to rat brain atlas¹⁰.

6.9 Remove LCM cap from QC station, attach Sample Extraction Device, and pipette 5.5 μL of lysis buffer onto sample.

6.9.1 Lysis buffer solution: 0.5 μL lysis enhancer, 5 μL Resuspension buffer.

6.10 Fit ExtracSure device onto 0.5 mL microcentrifuge tube and place on hotplate at 75° C for 15 min.

6.11 Spin down sample and lysis buffer for 30 s at low speed (0.01-0.02 x g) and place collected sample into −80° C freezer.

7. Single-cell Microfluidic RT-qPCR

7.1 Pre-amplification of single-cell mRNA for 96.96 Dynamic Array

7.1.1 Combine forward and reverse mRNA qPCR gene primers for all genes being assayed in a primer pool for pre-amplification. Each primer in 500 nM concentration. For example, 1 μL of 80 100 μM primers plus 120 μL of DNA Suspension Buffer.

Note: Primers used for representative experiment can be found in O’Sullivan et al. 2019⁴.

7.1.2 In new 96 by 96 PCR plate, add 1 μL of 5x VILO to each well.

7.1.3 Remove LCM single-cell samples from samples from −80° C freezer, let thaw briefly, centrifuge briefly at low speed (0.01-0.02 x g), and add 5.5 μL of lysed single-cell sample to PCR plate. Each sample is added to its own well.

7.1.4 Place PCR plate with samples and VILO added into thermocycler and heat at 65° C for 1.5 min. Spin plate down for 1 min at 1300 x g at 4° C and place plate on ice.

7.1.5 Add 0.15 μl 10X cDNA synthesis master mix, 0.12 μl T4 Gene 32 Protein and 0.73 μl of DNA Suspension Buffer to each well.

7.1.6 Place PCR plate into thermocycler and run the following protocol: 25 °C – 5 min, 50 °C – 30 min, 55 °C – 25 min, 60 °C – 5 min, 70 °C – 10 min, 4 °C – end.

7.1.7 Add 7.5 μL of taq polymerase master mix to each well.

7.1.8 Add 1.5 μL of primer pool (see above) to each well.

7.1.9 Place PCR plate in thermocycler and run the following pre-amplification protocol: 95 °C – 10 min. 22 cycles of: 96 °C – 5 sec, 60 °C – 4 min.

7.1.10 Add 0.6 μl of Exonuclease I reaction buffer 10X, 1.2 μl Exonuclease I, and 4.2 μl of DNA Suspension Buffer to each well.

7.1.11 Place PCR plate in thermocycler and run the following protocol: 37 °C – 30 min 80 °C – 15 min

7.1.12 Add 54 μl of TE buffer to each well. Spin down PCR plate at 1300 x g at 5 min. Store at 4 °C if immediately continuing to next step. Store at −20 °C if waiting more than 12 h for next step.

7.2 Prepare sample plate for 96.96 Dynamic Array

7.2.1 In new 96 well PCR plate, add 0.45 μl of 20X DNA Binding Dye and 4.55 μl of low rox master mix to each well.

7.2.2 Add 3 μl of pre-amplified sample to each well, spin down PCR plate at 1300 x g, then put plate on ice.

7.3 Prepare assay plate for 96.96 Dynamic Array

7.3.1 In a new 96 well PCR plate, Add 3.75 μl of 2x GE Assay Loading Reagent and 1.25 μl of DNA Suspension Buffer to each well.

7.3.2 Add 2.5 μl of 10 μM qPCR primer to each corresponding well. Spin down PCR plate at 1300 x g for 5 min.

7.4 Load and Run 96.96 Dynamic Array Chip

7.4.1 Prime chip with control line fluid.

7.4.2 Place chip in IFC Controller HX and run the Prime (136X) script.

7.4.3 Add 6 μl from PCR sample plate into corresponding sample wells in the 96.96 Dynamic Array Chip.

7.4.4 Add 6 μl from PCR assay plate into corresponding assay wells in the 96.96 Dynamic Array Chip.

7.4.5 Use needles to pop any air bubbles in the wells of the 96.96 Dynamic Array Chip.

7.4.6 Place 96.96 Dynamic Array Chip into IFC Controller HX and run the Load Mix (136x) script.

7.4.7 Removed chip from IFC Controller HX, peel off protective sticker, and place 96.96 Dynamic Array Chip into microfluidic RT-qPCR platform. Run the GE Fast 96x96 PCR protocol (30 cycles).

Note: RNA quality and result validity is assessed by multiple methods including assay validation via gel electrophoresis, melting temperature curves, sample and assay replicates, and standard dilution series plots. Additionally, transcriptional findings can be validated by independent methods on brain hemisection including Western blot and immunofluorescence assays.

8. Measuring Bacterial Abundance with Microfluidic RT-qPCR

8.1 Extract bacterial DNA following directions of stool DNA extraction kit.

8.2 Estimate bacterial DNA concentration using qPCR.

8.3 Add extracted bacterial DNA to new PCR plate. In this case, 1 μl of extracted bacterial DNA and 9 μl of DNA Suspension buffer.

8.3 Prepare assay plate for 48.48 Dynamic Array (steps 7.2.1 – 7.2.2)

8.4 Prepare sample plate for 48.48 Dynamic Array (steps 7.3.1 – 7.3.2)

8.4.1 In new 96 well PCR plate, add 0.45 μl of 20X DNA binding dye and 4.55 μl of low rox master mix to 48 wells.

8.4.2 Add 3 μl from PCR plate containing bacterial DNA to same 48 wells, spin down this PCR plate at 1300 x g for 5 min. Store at 4° C.

8.5 Load and Run 48.48 Dynamic Array Chip (steps 7.4.1 – 7.4.7)

REPRESENTATIVE RESULTS:

Validation of selection of single-cells is validated both visually and molecularly. Visually, cellular morphology is viewed before cell collection. Cells collected can then be viewed at the QC station and the cellular nuclei stain (DAPI) should overlap with the single-cell selection marker fluorescence. Figure 2A shows representative images of a slide with hemisected rat forebrain containing the CeA. Subsequent images (Fig. 2 B–D) show the selection of single cells and their removal from the tissue for transcriptomic analysis. Molecularly, cell type-specific markers should demonstrate increased expression in that cell type (Fig. 1C). Originally published in O’Sullivan et al. 2019⁴. We looked at neurons, microglia, and astrocytes and measured the expression of NeuN, Maf, and Gfap, respectively.

Further, controls can also be run in the microfluidic platform to validate expression findings. For example, analysis of other areas of the same tissue to demonstrate nucleus specificity. A separate tissue could also be used to demonstrate primer specificity in the tissue of interest. Positive and negative control genes can also be included—genes known to be absent from selected tissue and/or expressed highly. 3 or 4 “house-keeping genes” should also be included for data normalization purposes but also as a measure of experimental quality. These genes should demonstrate the lowest variance in expression across all samples and treatments. In this representative experiment, no alternate brain region was assayed, but house-keeping genes Ldha and Actb were used for normalization. Gapdh was used as an internal control.

Fig. 3 displays some of the multivariate methods our group uses to analyze our data. We find that astrocytes in Withdrawal were the most affected cell-type. We speculate based on these data in the context of other studies that astrocytes play a key role in inflammation in the CeA during opioid withdrawal and that this contributes to the physical and emotional symptoms that drive drug-seeking via negative reinforcement. We also show our gut microflora data (Fig. 4).

Figure 3. Representative Results 1.

Modified from O’Sullivan et al. 2019. (A) Cartoon schematic of cell displaying genes assayed and their location. Gene symbols labeled here are a reference for panel B. (B) Colored squares represent relative gene expression (median −ΔΔCt value) for gene denoted in panel A. Location of squares represents cellular localization or function of the corresponding protein. Panels display relative gene expression represented by color across treatments and cell types. Yellow is high expression, blue is low expression, white is neutral expression.

Figure 4. Representative Results 2.

Modified from O’Sullivan et al. 2019. (A) Gene correlation networks. Pearson correlation was performed on the −ΔΔC_t values within a treatment and cell type. Nodes denote genes and their color signifies relative expression levels (median −ΔΔCt value for each gene). Edges denote expression correlations; thickness signifies strength of expression correlation (ρ). Correlations with a q-value < 0.001 are displayed. Black edges are positive correlations and green edges are negative correlations. (B) Bar plots of select genes demonstrating significant differential gene expression. Statistics were calculated using nested ANOVA (#p<0.1, *p<0.05, **p<0.01, ***p<0.0001 n=4 animals for all treatments).

DISCUSSION:

Single-cell biology has demonstrated the heterogeneity of cellular phenotypes and robustness of tissue function. These findings have provided insight into the organization of biological systems at both macro and micro scales. Here, we describe the combination of two methods, LCM and microfluidic qPCR, to obtain single-cell transcriptome measures that provide anatomic specificity and transcriptional accuracy at a relatively low cost (Fig. 1). Our group takes a systems biology approach and often measures multiple tissues in the same animal. We find these methods to be both flexible and fruitful in determining how biological systems respond to various challenges at the transcriptional level. Additionally, we use these methods in the anatomic mapping of cellular phenotypes in baseline conditions.

We provide data and modified figures from a recent publication exploring how the CeA responds to opioid dependence and withdrawal ⁴. In this example, we used the same microfluidic qPCR platform to measure relative abundance of gut microflora. The methods and workflow are summarized in Figure 1. Originally published in O’Sullivan et al. 2019 ⁴. Major findings from high-throughput microfluidic RT-qPCR can be subsequently be validated by protein measures such as Western blot or immunofluorescence as we have done ⁴.

A major challenge of this systems biology approach is determining specific causal biological mechanisms. Fuzzy-logic is a validated solution that we have employed with success to infer agents in gene regulatory network behavior ². Animal model manipulation may also be employed to provide insight into systemic mechanisms. For example, the same protocol provided herein with the addition of a rat cohort that underwent a gastric vagotomy will yield data providing insight into the flow of information via the vagus nerve on this question.

Figure 5. Relative abundance of gut microflora.

Modified from O’Sullivan et al. 2019. Barplots display relative abundance of bacterial species (−ΔΔC_t values). #p<0.1, *p<0.05, **p<0.008, ***p=0.0009; two-way ANOVA n=4 animals for each treatment.