Cotyledon-Specific Flow Evaluation of Rhesus Macaque Placental Injury Using Ferumoxytol Dynamic Contrast-Enhanced MRI

Abstract

Background:

Recently, dynamic contrast enhanced (DCE) MRI with ferumoxytol as contrast agent has been introduced for the non-invasive assessment of placental structure and function throughout. However, it has not been demonstrated under pathological conditions.

Purpose:

To measure cotyledon-specific rhesus macaque maternal placental blood flow using ferumoxytol DCE MRI in a novel animal model for local placental injury.

Study type:

Prospective animal model

Subjects:

Placental injections of Tisseel (three with 0.5 ml and two with 1.5 ml), monocyte chemoattractant protein 1 (three with 100ug), and three with saline as controls were performed in a total of 11 rhesus macaque pregnancies at approximate gestation day (GD 101). DCE MRI scans were performed prior (GD 100) and after (GD 115 and GD 145) after the injection (term = GD 165).

Field strength/sequence:

3T, T1-weighted spoiled gradient echo sequence (product sequence, DISCO)

Assessment:

Source images were inspected for motion artefacts from the mother or fetus. Placenta segmentation and DCE processing was performed for the dynamic image series to measure cotyledon specific volume, flow, and normalized flow. Overall placental histopathology was conducted for controls, Tisseel, and MCP-1 animals and regions of tissue infarctions and necrosis were documented. Visual inspections for potential necrotic tissue were conducted for the two Tisseelx3 animals.

Statistical tests:

Wilcoxon rank sum test, significance level p<0.05

Results:

No motion artefacts were observed. For the group treated with 1.5 ml of Tisseel, significantly lower cotyledon volume, flow, and normalized flow per cotyledon were observed for the third gestational time point of imaging (day ~145), with mean normalized flow of 0.53 min⁻¹. Preliminary histopathological analysis shows areas of tissue necrosis from a selected cotyledon in one Tisseel-treated (single dose) animal and both Tisseelx3 (triple dose) animals.

Data Conclusion:

This study demonstrates the feasibility of cotyledon-specific functional analysis at multiple gestational time points and injury detection in a placental rhesus macaque model through ferumoxytol-enhanced DCE MRI.

INTRODUCTION

Maternal placental vasculature plays a crucial role in facilitating the exchange of oxygen, carbon dioxide, nutrients, and waste between the mother and the fetus during pregnancy. Abnormal development of the placental vasculature, which is associated with conditions such as local malperfusion, can lead to gestational complications, such as preeclampsia, small for gestation age infants, and pre-term birth(1-3). Therefore, early assessment of placental perfusion and blood flow could be valuable for predicting pregnancy trajectories and possibly allow early intervention if problematic suboptimal growth is identified.

Ferumoxytol is an FDA-approved iron nanoparticle-based supplement for the treatment of anemia, including during pregnancy. Ferumoxytol is frequently used as an off-label MR contrast agent, including clinical applications in pediatric and pregnant patient populations(4, 5). Studies show that ferumoxytol does not cross the maternal-fetal barrier, thus decreasing the risks to the fetus associated with contrast infusion(2, 6, 7). Additionally, ferumoxytol was determined to have little to no risk to the mother or the fetus through detailed pathological analysis in a non-human primate model(6). It has been found to be a safe and viable off-label alternative to gadolinium-enhanced magnetic resonance angiography for assessment of pulmonary embolism (PE) in pregnancy in 94 patients(8, 9), and has been proposed as a new approach to characterize internal placental structure in placenta accreta spectrum(10, 11).

Dynamic contrast enhanced (DCE) MRI is a promising tool to non-invasively assess placental perfusion and blood flow in vivo throughout gestation. Recent studies have shown promising results in animal studies using DCE MRI to identify functional units in the placenta, or cotyledons, which facilitate nutrient and oxygen exchange between the mother and the fetus, and quantify their number, volumes, and flow(2, 7, 12). However, few cotyledon-specific flow values in the presence of placental abnormalities have been reported(7). Here, we report initial results for DCE analysis of two novel rhesus macaque models designed to induce placental injuries(13). We hypothesized that placental Tisseel injections would create local infarcts that could be detected as regions of reduced flow with DCE MRI and that placental MCP-1 injections would provoke an inflammatory response with subsequent flow changes detectable with DCE MRI.

METHODS

Subjects:

This study was approved by the Research and Graduate Education Institutional Animal Care and Use Committee. Female rhesus macaques were housed with compatible males and monitored for breeding and menses. Date of conception was determined (+/− 2 days) based on initiation of the menstrual cycle, observation of copulation and presence of ejaculate, and ultrasound measurements of fetus and gestational sac(7). Full term in Rhesus Macaques is approximately 165 days(14). Three out of the eleven animals went through multiple pregnancies for this study (two went through two pregnancies and one went through three pregnancies), each with a different treatment. Macaques were cared for as outlined in the Animal Welfare Act and the Guide for the Care and Use of Laboratory Animals.

Imaging:

Eleven female rhesus macaques were imaged at up to three gestational days (GDs) of ~100, ~115, and ~145 days (Table 1). One day after the first MRI session, monkeys were injected with 0.5 ml Tisseel (n=3), 100 ug MCP-1 (n=3), or saline as controls (n=3). All injections were conducted by an experienced obstetrical clinician under ultrasound guidance and targeted to occur in the intervillous space of the anterior lobe of the placenta. Additionally, another n=2 monkeys received 1.5ml Tisseel injections, with the Tisseel being instilled as the needle was being withdrawn from the placenta. The fetuses were delivered via cesarean section at around gestational day 155, and the placentas were obtained for histopathological analysis by a veterinary pathologist.

Table 1. Rhesus macaque treatment, imaging, and delivery date, in gestation day.

The number of * in the first column corresponds to the same animal who went through multiple pregnancies for this study. Scans not included in this study due to image quality and missing data is marked ‘-’ in place of gestation day.

Rhesus & Treatment	MRI #1	MRI #2	MRI #3	Delivery & Dissection
Control #1	95	112	145	155
*Control #2	99	-	148	156
***Control #3	-	107	137	155
Tisseel #1	100	-	147	155
**Tisseel #2	-	117	146	154
***Tisseel #3	90	117	146	154
Tisseelx3 #1	-	121	149	156
*Tisseelx3 #2	-	114	139	155
*MCP1 #1	100	115	142	155
MCP1 #2	100	114	144	155
**MCP1 #3	-	114	142	156

^*,**,***:Animals that went through multiple pregnancies in this study.

Tisseelx3: Treatment group that received 3 doses of Tisseel.

All scans were performed on a 3.0 T clinical system (Discovery MR750, GE Healthcare) with a 32-channel phased array coil. The animals were sedated by inhalation of 1.5% isoflurane mixed with oxygen and imaged in right-lateral position. Respiratory bellows were used to track respiratory motion during imaging.

A series of time-resolved 4D DCE data sets were acquired during ferumoxytol (Feraheme, AMAG Pharmaceuticals; 4mg/kg, diluted 5:1 with saline), using a respiratory-gated, T1-weighted spoiled gradient echo sequence (DISCO, TR=4.8ms, TE=1.8ms, spatial resolution=0.86×0.86×1.00mm³, number of timeframes = 40). Ferumoxytol and a subsequent flush with 20 ml of saline was administered at 0.5 ml/s with a power injector (MEDRAD^® MRXperion MR Injection System, Beyer Healthcare) at the time the scan was started. Equidistant temporal resolution was 4.5s for the control, MCP-1, and Tisseel groups, and 7.7s for the two Tisseelx3 animals to capture potential longer contrast arrival times.

Processing:

Semi-manual segmentation (ITK-SNAP 4.0)(15) of the placenta was performed by a graduate student with 5-year experience in placental MRI analysis using the 3D volume from the last timepoint of the DCE imaging sequence. Subsequently, the temporal signal intensity of each voxel in the placenta was fitted to a sigmoid arrival model, and the inflection point of signal enhancement was used to form a contrast arrival time map(7).

Different from prior work, a 3D watershed algorithm (3D suite, ImageJ(16)) was used to identify individual perfusion domains based on detected boundaries in the 3D arrival time map. This algorithm first uses distance transforms to obtain the distance maps, which shows the relative distance between each point and the edge of the object boundaries. The local maxima of the distance map (which correspond to the center of each cotyledon) is then set as the seed point of each domain. The region grows out at a speed depending on the value of the distance map: the higher the distance, the larger the cotyledon and the faster the domain grows. The region stops growing when it reaches the minima of the distance map, which also represents the boundary of each region. Blood flow to each domain (ml/min) was calculated as the largest slope in the total volume of enhanced voxels over time. This is accomplished by calculating the zero point of the secondary derivative of the fitted sigmoid function. Additionally, a visual cotyledon matching was performed between color photos of the dissected placenta after delivery at term and the DCE MRI-derived 3D in vivo renderings of the individual domains. Figure 1 shows an illustration of this workflow which was adopted from previous studies(2, 7, 12) and modified by replacing the previously used 2D watershed with a 3D watershed algorithm.

Figure 1.

Dynamic contrast enhanced MRI processing workflow. A) The DCE MRI generates a dynamic series of 40 3D volumes, shown here for 4 timeframes during the arrival of the Ferumoxytol bolus. B) The corresponding signal enhancement over time is used to determine the contrast arrival time highlighted in square. C) 3D contrast arrival time mapped for the entire placenta indicating arrival zones (brown) and peripheral zones (green) of the perfusion domains. D) The perfusion domain map determined using the watershed algorithm, where each perfusion domain is shown in a different color.

Statistical Analysis:

The two animals who received the 1.5ml dose of Tisseel are considered as a separate treatment group (Tisseelx3) from the single dose. Each of the four (Control, MCP-1, Tisseel, Tisseelx3) treatment groups were compared for the distribution of cotyledon volume and volumetric flow for each cotyledon, which was defined as the volume of maternal blood entering a cotyledon per unit time. Additionally, we recorded normalized flow for each cotyledon, which was calculated as flow over volume for each cotyledon, to visualize changes in flow without impact from the cotyledon volume. Box-and-whisker plots were used to visualize such distributions, and the Wilcoxon rank sum test (MATLAB 2022b, MathWorks, Natick, USA) was used to determine the statistical difference (p<0.05) between each treatment group. Finally, total blood flow, calculated as the sum of each cotyledon’s flow (not normalized), was calculated. In order to track placental growth through time, longitudinal plots across three gestational time points were generated for total placental volume, placental surface area, and total placental flow.

Detailed pathological analysis on all animals was conducted by a veterinary pathologist with over 15 years of experience. Methods include morphological inspection and histological analysis of selected hematoxylin and eosin-stained tissue slices that are central-cut through the entire placenta. Qualitative necropsy report was generated for each animal detailing the ischemic changes throughout the placenta.

RESULTS

Image acquisition and analysis was successfully conducted for all imaging exams. Unexpectedly, a number of the injury model placentas only developed one placental disc, which deviates from the normal rhesus two-disc pattern (2/3 Tisseel, 2/3 MCP1, 0/3 controls, 0/2 Tisseelx3). Out of the eleven rhesus macaques analyzed, one Tisseel-treated placenta (0.5ml dose) was noted upon histological analysis (Figure 2) to have tissue ischemia that multifocally affects large portions of several cotyledons. In addition, both Tisseelx3 animals were shown via histopathology to have visible areas of multifocal ischemia with coagulative necrosis across the entire placental tissue. Qualitative assessment of written pathology reports from the other animals shows very moderate pathology likely due to underlying factors not associated with Tisseel injection.

Figure 2.

Histopathological central slice of placenta treated with Tisseel injection. Green arrows are pointing to areas with diffused ischemia and coagulative necrosis characterized by loss of secondary and tertiary villi; dark blue arrow is pointing to healthy villus tissue; brown arrow is pointing to villus tissue likely disrupted by Tisseel.

Table 2 shows mean and standard deviation of cotyledon volume, flow per cotyledon, normalized flow per cotyledon (flow over cotyledon volume), and total flow for each animal at the third imaging timepoint (GD ~145). For reference, the total number of cotyledons (n) for each treatment group at that timepoint is n=31 for controls, n=28 for MCP-1, n=29 for Tisseel, and n=36 for Tisseelx3. The mean cotyledon volumes (5.5 ml, 5.7 ml), flow per cotyledon (2.9 ml/min, 3.3 ml/min) and total maternal placental flow (57.4 ml, 52.9 ml) for the two Tisseelx3 animals are lower than the same parameters for the control animals.

Table 2. Cotyledon volume, flow per cotyledon, normalized flow per cotyledon, and total flow for each animal at the third imaging timepoint (GD ~145).

Rhesus & Treatment	Cotyledon Volume (ml)	Flow per Cotyledon (ml/min)	Normalized Flow per Cotyledon (min−1)	Total Placental Flow (ml/min)
Control #1	7.8±4.5	5.2±3.0	0.70±0.18	67.0
Control #2	9.2±5.3	7.0±3.9	0.79±0.17	69.7
Control #3	10.8±6.1	9.1±3.8	0.91±0.32	72.5
Tisseel #1	12.6±3.8	10.9±2.5	0.84±0.10	65.5
Tisseel #2	10.2±3.5	7.5±2.3	0.75±0.17	60.0
Tisseel #3	7.6±2.4	6.9±1.9	0.95±0.31	102.9
Tisseelx3 #1	5.5±2.9	2.9±1.5	0.53±0.08	57.4
Tisseelx3 #2	5.7±3.3	3.3±2.9	0.53±0.15	52.9
MCP1 #1	7.5±4.2	5.9±3.1	0.82±0.17	89.0
MCP1 #2	18.8±9.0	11.8±6.4	0.61±0.05	59.0
MCP1 #3	11.7±8.7	10.0±5.6	1.03±0.33	79.9

Longitudinal analysis (Figure 3) confirms increases in placental volume, placental surface area, and total maternal placental flow with gestational age, except for the two Tisseelx3 animals, which show slight decrease in total flow from the second to the third gestational timepoints confirming the utility of Tisseel in inducing placental injury.

Figure 3.

Longitudinal plots of placental surface area, placental volume, and total maternal flow across three imaging timepoints for all eleven animals. All variables increase slightly with gestation, except for the total flow, where the animal who received three Tisseel injections decrease in flow from the second to the third gestation time points.

Significantly lower flow, cotyledon volumes, and normalized flow (flow divided by volume) for the Tisseelx3 treatment groups at GD 145 (approximately 45 days after the Tisseel treatments) were observed and are presented in Figure 4. Interestingly, this is not observed at the second gestational time point, which is approximately two weeks after treatment. There is no such change in blood flow between the other two treatment groups (MCP-1 and Tisseel) for either gestational time points.

Figure 4.

Box-and-whisker plots of flow, volume, and normalized flow of each cotyledon for each gestational timepoint (GTP; left to right). The p-value of the Wilcoxon rank sum test for each treatment group comparing to the control animals are shown in the legends on the bottom of each plot.

The fast-arrival regions (center of cotyledon, red color) on the contrast arrival time maps (Figure 5) decrease in volume with gestation, likely due to the injection of Tisseel. The number of cotyledons decreases with gestation for the Tisseel animal. The similar decrease in the fast-arrival regions is also observed with the Tisseelx3 animal on the bottom of Figure 5. Both animals have good agreements between arrival time maps and the placenta dissection photo. A comparison between the arrival time histograms between the Tisseel-treated animal and a control animal show decreased mode arrival time with gestation for the Tisseel animal. This could further corroborate the negative influence of Tisseel injection on placental intervillous blood flow. A comparison between the Tisseelx3 animal and the control animal shows a longer mode arrival time in the Tisseelx3 animal. Although the mode arrival time does not change for the two gestational timepoints, the arrival time is more heavily skewed to the right for the third gestational timepoint in the Tisseelx3 animal’s histogram.

Figure 5.

Contrast arrival time maps for each gestational timepoint (A), a photo of the placenta at dissection (B), and arrival time histograms (C) at available gestational time points (GTP) for a Tisseel-treated and a control animal. Apparent decrease in cotyledon number is observed throughout gestation; additionally, the central region of each cotyledon (arrival time < 20s) decreases slightly relative to the cotyledon size. The arrival time map at GD 146 matches well anatomically with the placenta at dissection. Bottom: arrival time maps (D) at two gestational timepoints (~1 day and ~15 days after treatment), dissection photo (E), and arrival time histograms (F) of the two timepoints (GTPs).

DISCUSSION

This study introduces 2 new models for placental injuries in the rhesus macaque using MCP-1 and Tisseel injections and demonstrates successful measurement of cotyledon-specific flow parameters for these models. To date, no established model for local placental injuries exists. The Tisseel injections, particularly the triple dose, caused placental injury that was identified in histopathology and MRI, making it a promising approach that warrants further pursuit into optimal intervention timing and dose to become a robust animal model. Meanwhile, the MCP-1 injections did not evoke trackable changes in histopathology or MRI, thereby this animal model with its currently protocol is not useful.

The analysis workflow in this study is adapted from previous studies ((17), (12)), with changes including imaging parameters, watershed algorithm, and flow fitting methods. This could contribute to a lower average total placental flow of the three control animals (69.7 ml/min) compared to previously reported volumetric flow (403 ml/min) ((12)). However, the total flow in our study is comparable to that of a previous study using similar analysis workflow (63.5 ml/min) (7). Additionally, significantly lower cotyledon volume was observed in the Tisseelx3 animals at around gestation day 145, which agrees with a previous study using pathological models induced by ZIKA virus(7). Longitudinal analysis confirms slight increase in placental volume, placental surface area, and total maternal placental flow with gestation age as previously observed(7).

Significantly lower flow parameters were observed for the Tisseelx3 treatment groups at GD ~145 (approximately 45 days after the Tisseel treatments), which agrees with the preliminary histopathology analysis that indicates seemingly more infarcted and fibrotic tissue. Tissue fibrosis could have hemodynamic impact that compromises placental transport of nutrients to the fetus(18). Interestingly, the decrease in flow parameters is not observed at GD ~115 (15 days after Tisseel treatments), even though a previous study has shown that Tisseel induces almost immediate (within 4 to 10 minutes) hemostasis(19). This could be due to the combination of progressive placental insufficiency with time, as well as the increasing demand from the fetus as gestation progresses(20). There was no such change in blood flow between the other two treatment groups (MCP-1 and Tisseel) for either gestational time point, comparing to the controls. This is possibly due to insufficient injury in the first place or placental plasticity that can compensate for certain degrees of injury; a previous study has shown that the rhesus macaque placenta can compensate for up to a 40% reduction in functional capacity(21). This also agrees with our preliminary pathological findings in that the two treatments only resulted in modest placental injuries, except for one animal treated with Tisseel that was histologically shown to have ischemia in multiple areas of the placenta.

The decrease of the central region of the contrast arrival time maps with gestation could further corroborate the negative influence of Tisseel injection on placental intervillous blood flow, since this suggests larger regions of the placenta that take longer for the contrast agent to reach. Interestingly, for the Tisseel animal and the two Tisseelx3 animals, although the injections occurred locally (and likely distributed through multiple cotyledons for the Tisseelx3 group), all cotyledons show various amount of decreased center of flow. Depending on the significance of the vascular structure at the injection site, the introduction of Tisseel could impact one or multiple portions of the maternal villous tree, causing a global effect on placental flow(22).

This study applied several improvements to existing methodology(7), which was adapted from a study by Schabel et al as previously mentioned(12). We made some changes in the algorithms that could potentially increase the accuracy of cotyledon blood flow measurements. First, this study uses a 3D watershed algorithm (3D suite, ImageJ(16)) instead of a 2D-based algorithm to automatically segment the individual cotyledons. This new algorithm calculates the distance map and performs region growth in 3D, rather than in 2D with a pixel connectivity index, which gave us improved robustness. In addition, the calculation of individual cotyledon flow in this study is based on the highest slope in the enhanced volume vs. time curve. This was accomplished by calculating the first derivative numerically and choosing the arrival time with the highest value. This method could result in greater robustness for flow calculation than the previous implementation which used a linear fit of the cumulative total volume of enhanced voxels over time, which heavily depends on accurately determining the start and end time of signal enhancement(7). In addition, this study reports normalized blood flow, calculated as blood flow per volume of each cotyledon. Previous studies have reported the positive association of placental growth and flow changes with time(23, 24). In our study, the flow is calculated as the volume of maternal blood entering the cotyledon per unit time, which could be largely impacted by the cotyledon size, since larger cotyledon usually contains more maternal blood. Therefore, normalizing the flow could potentially remove the confounding factor cotyledon size has on the flow.

Limitations

Our study has a modest sample size: 2–3 placentas per treatment group, 6–15 cotyledons per disc of placenta. A considerable number of animals only developed a single placental disc, which was unexpected given the reported low percentage (9.5%) of monodiscoid placenta in rhesus macaques(25). This was not caused by the injury models as the single disc placentas had already developed at the time of injury. Single disc placentas could potentially bias the flow per cotyledon due to the different distribution of blood to a single placenta as opposed to two discs. The small sample size also limits the available quantitative statistical approaches available for this study. Potential movements of the fetus during imaging, especially for a time-dependent imaging series, could also result in inaccuracies in flow measurements.

Each pregnancy in this study is treated as an independent experiment due to the limited number of individual animals, although three of our animals went through multiple pregnancies. This could impact the independence of individual data points.

Lastly, although this study offers an overall impression of the pathological findings, detailed quantitative per-cotyledon analysis of pathology is lacking, therefore limiting the cotyledon-to-cotyledon comparison between flow and pathology.

Conclusion

This study examined the effects of MCP-1 and Tisseel injection into the placental intervillous space in rhesus macaques. Cotyledon-specific maternal placental blood flow was successfully calculated, and significantly lower cotyledon volume, flow, and normalized flow was shown for the animals who received the higher dose Tisseel injection. In addition, the injections of the lower dose of Tisseel or MCP-1 were shown to have similar pathology findings to the controls and only modest effects on placental cotyledons and flow and are unlikely to induce pregnancy complications. Directions of future studies include injection of Tisseel at an earlier gestation stage to track placental development through a longer gestational window, as well as comparing placental flow with a more quantitative pathological description of each cotyledon.