Read-counting for PGT-A

ICSI, source:

This is an introductory article about Pre-Implantation Genetic Testing for Aneuploidy (PGT-A, formerly known as PGS) using sequencing-based read-counting. The procedure is simplified and I am avoiding to talk about specific companies or products here.

Increasing the chances of a successful pregnancy by looking at the embryos chromosomes

During an IVF cycle there are – in most cases – multiple embryos available for transfer to the woman hoping to become pregnant. IVF treatment is expensive and a stressful experience for the patients. The goal of the treatment must therefor be to have success after one or only a few cycles. Factors contributing to success or failure of an IVF cycle are various, many of which are still poorly understood. One of the better-understood factors is the genomic make-up of the embryo(s) transferred: If the cells have a non-standard number of chromosomes, the chances of survival are low for the embryo. 

In order to select an embryo with a high chance of implantation and pregnancy, Pre-Implantation Genetic Testing for Aneuploidy (PGT-A) can be performed. This procedure can assess the genome of an embryo at a high level by identifying the number of chromosomes (or regions of the chromsomes) found in a few cells sampled. Deeper levels of genetic analysis are of course also possible, but the impact of these findings is usually poorly understood and would lead to a selection process that we should stay clear of!

Counting chromosomes by counting sequencing reads

Determining the copy-numbers of every chromosome can be reliably done with sequencing-based approaches nowadays. One way to achieve this is the read-counting method employed by many software solutions and commercial products. The method can be outlined as followed:

  1. Extract and amplify the DNA of a few embryonic cells
  2. Generate a sequencing library 
  3. Perform shallow-depth sequencing
  4. Align the sequencing reads back to the human reference genome
  5. Count how many reads you see in the different regions and chromosomes and
  6. Determine the copy-numbers from this measurement.

This is of course vastly simplified and cannot be discussed at every level of detail here, but we can focus on the last steps a bit further.
The theory we build on is that if the retrieval of the cells, the generation of the sequencing library and the experimental protocol are kept the same, the number of sequencing reads counted in a specific genomic region should only depend on two factors:

A. The characteristics of that genomic region: Some parts of the genome are amplified more rapidly than others and will be overrepresented in the mixture.
B. The amount of input material: If there was an extra copy of the chromosome we should have more DNA going into the sequencing process and also see more reads coming out for this specific region.

Factors underlying point A are mostly based on the characteristics of the amplification reaction mentioned in step 1 above (e.g. which polymerase enzyme is used) and the base-composition of the genome itself. The latter mostly comes into play when we try to place the sequencing reads found back on the reference genome: If the genomic region the read originates from is repeated a few times on different chromosomes, this alignment process cannot reliable determine where to place the read and it has to be discarded. However, as the genome is >99% identical between individual and because the amplification reaction should be performed in a standardised way, this type of bias is relatively constant. Point A above can therefor be analysed beforehand and this amplification and mappability bias can simply be removed.  This leaves us with the differences originating from point B – which are the actual differences that we are trying to assess in the embryo.

In a standard human cell there are 2 copies of each of the autosomes (chromosomes 1 to 22), this becomes our „normal“ level. If we look at the actual numbers of reads counted, we can derive the most likely copy-numbers of the region by comparing them to our normal level:

Theoretical numbers of reads0*50100150200
Assumed copy-number of the region01234

* There will usually be a few reads erroneously found in these regions

Even with very strict protocols there will be variation in read numbers though, making the the assessment more challenging. Additional processing steps can be applied to clean and smooth the data. For this to work it is assumed that most regions are at the “normal” level and that neighbouring measurement point behave similarly in most cases.

Visualising the measured copy-numbers for each region is a key part in the embryo assessment. Autosome read counts outside of the grey shaded area in the copy-number plot above would be considered gained / lost. The sample shown seems to have 3 copies of chromosome 16. Chromosomes X & Y show copy-number 1 as this would be a male sample.

This assessment can now guide the decision which of the embryos might have the highest change of survival bases on its genomic make-up and could be selected for implantation during an IVF cycle.