Discovering new cell types one at a time

From a single zygote, the human body grows into 37 trillion cells. Rough estimates have put the number of different cell types in the human body at around 200, a number which seems low given the amount of diversity and specialization in our body. Many cell types are likely to be indistinguishable by simple morphology. What if you wanted to figure out all of the cell types within a population of cells? To do that, you'd need to be able to look at the characteristics of many cells individually.

Recent work from Steven McCarroll's Lab at Harvard Medical School and the Broad Institute has addressed this problem in a fascinating way. The researchers have developed a technique called drop-seq that sorts out single cells and examines their mRNA to determine the cell's gene expression profile. The mRNA present in each cell gives a snapshot of the genes that are expressed and important to that cell. It is a good way to characterize cells; different types need to express different genes and so accumulate different mRNAs. It's far easier to measure a cell's mRNA than, say, its protein levels, because we are very skilled at analyzing at amino-acid sequences.

To accomplish this difficult task, the team used a microfluidic device that takes in 3 different liquids. The first is a water-based buffer carrying cells collected from cell culture or tissues and dissociated until they are in a single-cell suspension. The second liquid is an oil. Because oil and aqueous solutions can't mix, the oil forms separate small droplets out of the buffer containing the cells. The final liquid is a water-based cell lysis buffer than contains very special microbeads with tails of amino-acid chains coming off them. The chains can bind to mRNA, and each one has a barcode spelled out with As, Ts, Cs and Gs to identify the microbead it came from. The liquids are all pumped in at the same time in the hope that one cell and one microbead will end up together in the same droplet created by the oil stream.

The researchers pump the bead and cell solution into each other and push this through the oil, getting one cell and one bead together in each droplet. This is displayed very well in the figure to the right. The cells are then lysed by the lysis buffer, and the mRNA from those cells quickly bind to the beads in that droplet. Next, the droplets are broken and all of the beads are pooled. From here, the researchers use reverse transcriptase to make cDNA from the mRNAs attached to the microbead barcodes. This cDNA will have the sequence of the gene the mRNA came from and the barcode from the specific microbead. Following this, they used complex algorithms to analyze all of the cDNA sequences, determine which cell the sequence came from, and create gene profiles for each cell. Altogether, this informs us about the precise gene expression in each of thousands of cells.

##&&##

#A&A&A

The team carried out a relatively simple experiment to prove this technique would work on a large scale. They mixed human fibroblasts and mouse fibroblasts in culture and put the mixture through drop-seq. They found that nearly every cell they profiled expressed only human genes or only mouse genes, as illustrated in the figure to the left. This is an important proof-of-principle experiment, because it shows that they can use their technique to correctly characterize a heterogeneous collection of cells . Some drops (one in the figure) express both human and mouse genes because two cells were caught in the droplet with the microbead. This leads to an even mixture of mouse and human genetic expression. Even the best single cell analysis will occasionally be contaminated with doublets. However, the rate of doublets in drop-seq is extremely low, as you can see, making this technique very strong.

Confident that they could distinguish different populations, the team moved on to complex tissues. They chose the mouse retina because it has been well studied and many of the cells have been characterized. They looked at almost 45 thousand cells and successfully classified them into 39 distinct populations based on gene expression. In the figure on the below , the 39 different populations are clustered based upon their similarities in gene expression. The 39 populations included several expected classes, like photoreceptors and retinal ganglion cells. In addition, they found sub-populations among the large clusters. One cell-type with sub-populations was the amacrine cells, a diverse population of mostly inhibitory interneurons. At least three types of amacrine cells have already been identified based on the neurotransmitters they release; with drop-seq, the authors parsed out 21 different sub-populations of amacrine cells based on their gene expression. The gene expression data enabled them to identify markers, or unique genes expressed by one cell type, to label each sub-population.

Drop-seq enables us to characterize and profile individual cells from a large population by looking at each cell's gene expression. While several techniques exist to sort out and analyze hundreds of cells, sorting and profiling of this magnitude hasn't been done before. Advancements like this are important, since discovering and characterizing new cell types helps us understand the body better. This technique could help explain apparent discrepancies in homogenous populations and sharpen our focus when we study diseased cells or tissues. It also allows for the study of homogenous cell populations by looking at discrete differences in genetic expression during specific cellular states, such as at different points in the cell cycle. With techniques like this, we can make some serious headway into creating a map of all 37 trillion cells in our body, one at a time.

References:

http://sciencenetlinks.com/student-teacher-sheets/cells-your-body/

Bianconi, E., et al. An estimation of the number of cells in the human body. Annals of Human Biology, 40, 463-471 (2013)

Macosko, E.Z., et al. Highly Parallel Genome-wide Expression Profiling of Individual Cells Using Nanoliter Droplets. Cell, 161, 1202-1214 (2015).

Image credits:

All images are augmented from figures in the Macosko et al. paper referenced above.