In this presentation, we will discuss different approaches to prediction of pregnancy complications, including discovery of exosomal biomarkers.
Dr. Louise Laurent, is a Professor and the Vice-Chair for Translational Research in the Department of Obstetrics, Gynecology, and Reproductive Sciences at the University of California, San Diego. She received her medical and doctoral degrees from the University of California, San Francisco, and completed her residency in Obstetrics and Gynecology and her fellowship in Maternal Fetal Medicine at the University of California, San Diego. Her research focuses on using stem cell and genomic approaches to understand early embryo development and normal and complicated pregnancy.
Dr. Laurent has disclosed that she does not have any real or perceived conflicts of interest in making this presentation.
ChiJenny Nguyen: Dr. Louise Laurent is a Professor and the Vice-Chair for Translational Research in the Department of Obstetrics, Gynecology, and Reproductive Sciences at the University of California, San Diego. She received her medical and doctoral degrees from the University of California, San Francisco, and completed her residency in Obstetrics and Gynecology and her fellowship in Maternal Fetal Medicine at the University of California, San Diego. Her research focuses on using stem cell and genomic approaches to understand early embryo development and normal and complicated pregnancy. Dr. Louise Laurent’s presentation is titled, Advances in Prediction of Pregnancy Complications.
Dr. Louise Laurent: Hello, my name is Louise Laurent, and I am a professor in the department of OB/GYN and Reproductive Sciences at the University of California, San Diego. I’d like to thank Lindsey Wimmer and the others at Star Legacy for allowing me to share some of the work from my lab.
Here are my disclosures. A lot of the work of my lab focuses on trying to understand the interactions between mom and baby. In this endeavor, we look at the maternal circulation as a window into placental function. The reason for this is that placental cells come into direct contact with the maternal blood, which allows us to sample the maternal blood and look at the health of the placenta.
Here, what we see is a picture of a placenta here. This is the umbilical that leads to the baby. These structures here represent the placental villi, and these are fetal vessels here. What you can see is that the blood comes in from the mother through these arteries that are indicated here in red. It circulates around, and during that process, the nutrients from the mom including water, sugar, proteins, other nutrients are transferred into the maternal circulation. Then the waste products come out of the villi as well. Things like carbon dioxide and other waste products. Then are carried back into the maternal circulation through these blue veins. Here you can see this really close interaction between mom and baby in this space, so that anything that is being shared by the placenta will end up in the maternal circulation.
The other really attractive thing about using maternal blood as a source of biomarkers that can be used to follow placental health is that it can be safely sampled anytime during pregnancy.
Now we’ve introduced the concept of biomarkers. Biomarkers are molecules that we can sample and test and indicate something about the health or the disease state from a patient. Most biomarkers historically have been things like hormones or proteins. Here, what we’re studying in my group is extracellular RNAs, which are a new kind of biomarker. Extracellular RNAs are really generally small pieces of RNAs that are secreted by cells into their environment.
Here, this represents a cell. This is a nucleus of the cell, and what you can see here, these little green twiggly lines are RNA molecules. What you can see is RNA molecules can be packaged up into little vesicles here that are surrounded by a membrane. There are larger ones called microvesicles that bud directly off of the plasma membrane of the cell, there are also these smaller ones called exosomes, which are formed using this unusual mechanism where they bud into large vesicles and then are released outside.
However, extracellular RNAs can also be carried by little protein complexes called ribonucleoproteins, and they also can be carried by lipoproteins. Combining this knowledge about extracellular RNAs and the concept that the maternal circulation carries signals from the placenta led us to the projects that I’m going to talk about right here in the rest of the talk where we use extracellular RNAs as biomarkers to indicate placental health and disease.
First, I’ll tell you a little bit about some early work that we did that showed that extracellular RNA profiles change over the course of pregnancy. What we did is we collected both placental tissue and maternal serum across pregnancy, and we did some verification on the maternal serum samples to show that they were high quality and showed differences in the levels of different proteins over the course of pregnancy as was previously reported by other groups. Then we did isolation of the extracellular RNAs, and we used a procedure called small RNA sequencing to analyze these small RNAs in the maternal circulation.
Here, what we see is some data from our placental microRNA expression profiling. This is very busy, but it’s basically to show that over the course of pregnancy, there are certain groups of microRNAs that show different patterns. Here, what you see is samples from very early pregnancy, a little farther along, these are both in the first trimester early and late in the first trimester. This is in the second trimester, and then these samples are from the third trimester of pregnancy. What you can see, for example, is that there are some microRNAs that are expressed very highly early on and gradually decrease in expression going from red to blue across the course of pregnancy.
Whereas other ones like this group here start out quite low, increase in the late first trimester, are high in the second trimester, and then their levels fall again in the third trimester in the maternal circulation. Looking at the samples as little ball, so each one of these represents a placenta at a different point in pregnancy, you can see that early placenta samples cluster closer together, and then ones later in pregnancy cluster to closer together showing that the microRNA profiles change over pregnancy in a smooth and systematic way.
Here, these are samples of maternal serum that are taken over the course of pregnancy. These are two very well-known proteins that have been known for a long time to change in the maternal circulation over the course of pregnancy. You can see from all these little dots are maternal serum samples. The days of pregnancy is down this way, and then this is the level of the proteins. This is CSH, this is ADAM 12. What you can see here is that for these two proteins, at least, the levels increase smoothly over the course of pregnancy as expected. This shows that our samples of maternal serum were high quality.
What we did then is we did that small RNA sequencing to look at the profiles of the extracellular RNAs, and here what I’m just showing you is the results from a particular small RNA that had a very interesting pattern. This is its name. In the placenta, the levels dropped very quickly in early pregnancy. Whereas in the maternal serum, they increase slowly over the course of pregnancy. We were really interested in why we should see the opposite patterns in the placenta and in the serum.
Then in collaboration with Mana Parast Lab, we looked at where and when this certain microRNA is being expressed in the placenta. What we see is that early on in pregnancy is being expressed in this very special type of cell, it’s called the primordial syncytium. Whereas in later pregnancy it’s expressed by cells on actually two opposite sides of the placenta, the side of the placenta that’s nearest the baby, as well as nearest the mother in two different areas.
It is actually not expressed at high levels in the middle, that body of the placenta. It turns out that for these placental profiling experiments, we specifically profiled this middle section. This explains why the levels in our placental samples dropped over the course of pregnancy because we’re sampling this area where the levels were dropping, whereas in maternal serum, which samples all of the placenta, the levels actually increased.
Once we did this very early study and demonstrated that we can profile extracellular RNAs in maternal serum, we turned our attention to the discovery of extracellular RNA biomarkers for a very clinical application, and that is the prediction of preeclampsia. Preeclampsia is thought to arise from abnormal placental development and function. It arises in about 5% pregnancies and it is one of the major causes of maternal and fetal illness and death.
It’s characterized by high blood pressure in the mother, as well as protein in her urine. In terms of symptoms, moms typically will develop headaches, blurry vision, and in some cases, abdominal pain. It is also characterized in terms of lab values by dysfunction of the platelets, the kidneys, and the liver. In very severe cases, there can be swelling of the brain and liver, which can be life-threatening to the mom.
The only cure for preeclampsia is delivery. It’s the job of clinicians who take care of moms who develop preeclampsia to try to balance the risk of preeclampsia and prematurity. If the preeclampsia is severe, then sometimes we need to risk some of the complications of prematurity and deliver the babies early. Whereas if the preeclampsia is mild, then sometimes we can delay delivery a little bit to help the babies be less premature at birth. This all points out to the fact that it really would help us clinically if we could predict who was at the highest risk of preeclampsia.
How do we do this now? Well, currently we look at the maternal medical history and her pregnancy history because women who have things like high blood pressure or autoimmune disease or previous pregnancy affected by preeclampsia have a higher risk. We also look at analytes in the serum, including the ones I’ve listed here, some of which are hormones, other of which are proteins. These have been shown to be correlated with increased risk for preeclampsia. We also do an advanced ultrasound method called uterine artery dopplers to predict which women are at highest risk for preeclampsia.
The advantages of the current approaches are that they partially take advantage of the blood test that we do for standard screening for aneuploidies like trisomy 21, otherwise known as Down syndrome. These analytes here are already tested for in our screen for Down syndrome. The other thing is by combining all of these things together, we can detect about 95% of the cases of preeclampsia that occur early in pregnancy and are quite severe.
The disadvantages of the current approach is that it does not detect preeclampsia that arises late in pregnancy, which is actually the majority of cases. Also, uterine artery doppler studies are not standard practice, especially in early pregnancy. Also, history and serum analyzed alone without the early uterine artery doppler studies are not very sensitive. This is why we embarked on a study to try to identify additional biomarkers that might increase our ability to predict women at higher risk of preeclampsia early in pregnancy.
This is the design of our study. We collected paternal serum between 17 and 28 weeks of pregnancy, and we collected them from two different clinical sites. One here at UCSD, and one at multiple sites coordinated by Sera Prognostics, our collaborators. The samples were split into two sets, a training set, which I’ll talk more about later, as well as the verification set. The training set was used first and then the verification set was used later to confirm the results of the training set.
We wanted to first make sure that the makeup of the discovery and verification sets was similar. This is showing that this is the case that in both cohorts, the number of controls was about twice as high as a number of cases. Also looking at the gestational age or point in pregnancy at which the samples were collected, we can see that it’s not an even distribution, and this is expected because we recruited women at ultrasound visits. They peak at two areas here right around 18 weeks of pregnancy and right around 24 weeks of pregnancy. But you see that it’s the same for both the discovery cohort which is in the blue here and the verification in gold. The two cohorts are split quite evenly.
Looking here at the characteristics of the patients in the two sets, we again see that they’re quite similar. There are really no differences between the training and verification sets as indicated in this column with all the numbers being well over the cutoff of 0.05. Here what we see is within both the training and the verification sets that there are a few features that are different between the cases, the women who got preeclampsia and the controls who had healthy pregnancies.
These include the maternal body mass index. In both cohorts, the women who developed preeclampsia had a higher body mass index by about three to five. We also saw that the age of delivery was lower in the cases compared to the controls, which again is expected because women with preeclampsia are generally delivered a little earlier to try to protect the health of the mom and the baby. Also more preterm births in the cases compared to controls in both cohorts, and also the babies were smaller in the cases versus controls. This is partially explained by the earlier gestational age of delivery, but also babies of moms with preeclampsia tend to be smaller. These results were what we expected.
After this, we proceeded with the actual experiments. We isolated the RNA from the maternal blood sample and put them into these plates. We then processed these plates and did the actual small RNA sequencing process which were run on these machines. This is called the Hiseq 4000, which is a big sequencing machine. The data that we got out of this process was then analyzed to identify microRNAs that are differentially present in women who have normal pregnancies versus those who develop preeclampsia later in pregnancy.
We did two things here. One is that we did a procedure called stratification by gestational age, which means that we compared results from women earlier in gestational age in the cases with the ones earlier gestational age in the controls. Then once later in the cases with the ones that are later in gestational age in the controls. This is a very common way of splitting up the data.
The other thing that we did was a little bit more unique, and that is, we looked at what are called univariate predictors, which are single microRNAs one at a time. For example, we took microRNA A, and we looked in cases and controls and said, “Are these different or the same?” Then we did the same thing with microRNA B, did the same thing. That’s very standard.
We also did something called by bivariate analysis, which is a little bit more novel. What that means when you do bivariate analysis is that you look at the microRNAs two at a time in pairs, and you actually look at the relative expression level of those two microRNAs together as your biomarker. To try to explain how this method might be useful, I’ve drawn these little cartoons here.
Let’s say that we’re looking at a group of people and we want to identify the people who have long legs proportionally. How would you do this? Well, if you did a univariate analysis, what you would do is you would look at the length of legs and you’d say, “How long is this person’s legs?” You would say, “Well, maybe this is an inch.” And then, “Let’s look at this person’s legs.” “Well, this person’s legs are maybe one and a half to two inches.” And then you would look at this person, you would say, “How long is this person’s legs?” and say, “Oh, maybe that’s about two inches.” In that case, what you would say is that this person had the shortest, this one had intermediate, this one had the longest legs.
However, if you stand back and you look at these two people, these two people have the same proportionate length of legs. This person’s just a bigger person. In this way, looking at the univariate measurement, the single measurement is actually not going to identify well who has short legs and who has long legs compared to their overall size.
What you do in a bivariate analysis is, for example, you might take the ratio of the leg length to the arm length. In this case, this person has about the same length of legs and arms. That ratio is 1:1. In this case, this person is proportionally just bigger than this person here, and so this ratio is 1:1 as well. In this case, these two people have the same length of legs proportion to their overall size. This person, though, as you can tell, has much longer legs, and that ratio is about 2:1. Using this ratio, we can identify the individual with the longest legs based proportionate to their size pretty easily.
I’ve also shown in these three little people here that they have different colored eyes. This is just to show that sometimes you can make ratios that are meaningful like leg to arm ratio. If you did a bivariate analysis looking at leg length versus eye color, that would be pretty meaningless. Part of the data analysis process is to identify which ratios are useful and which ones are not useful.
Here are our results here. What you can see is we have univariate analysis here, bivariate analysis. Remember, these are the pairs of microRNAs in bivariate analysis. What we did is we looked at how many of our biomarkers passed our filters in our training set, or a discovery set, and then in a second round in our verification set. You can see here in the univariate analysis that we overall had very few microRNAs that passed our filters in training, and many, many fewer even that passed verification.
There are four lines because, again, we stratified our cases based on how far along they were in pregnancy, and we looked at actually four different windows. We looked at people across the whole window, 17 to 28 weeks. We looked at people early in that window. People whose gestational age were in the middle of that window when they had their blood drawn, and then people who were late in that overall window. You can see the results are not that different here across the different gestational age windows.
In the bivariate analysis, we had a lot of pairs of microRNAs that passed our filters in training. We just picked the top 50 for further analysis. Then we looked at how they performed in verification. Here you can see that there was a difference in performance and- [sound cut]
This presentation was part of the Stillbirth Summit 2021. This individual lecture will be awarded .5 hours of continuing education credit to include viewing the lecture and completing evaluation and post-test. Once received a certificate will be emailed to the address you provide in the post-test. If you did not register for the Summit WITH Continuing Education, you can purchase the continuing education by clicking here. This purchase will provide you access to all Stillbirth Summit 2021 lectures including continuing education credit. There is no charge for viewing the presentation.
Please feel free to ask questions of the presenter. We will obtain their answers/comments and provide them here as received.