Abstract

Materials and Methods

Results

Immunostaining for CK-7 was used to identify implantation sites containing EVT invasion in first trimester decidua. These were further investigated by serial immunostaining for CD31 and α-SMA to identify decidual arteries and assess remodeling status. Arteries, lymphatic vessels, and veins (Figure 1, A, B, D–F) were detected in all decidual samples. Leukocytes were detected in abundance in decidua parietalis (Figure 1C). Arteries displayed a thick, multilayered vascular wall (Figure 1A) unlike that seen in veins, which were excluded on the basis of morphological attributes, such as a dilated lumen, a thin layer of VSMC and intact endothelium (Figure 1, D and E). CD45-positive leukocytes were detected clustering around venous walls (Figure 1F). The identity of vEVTs was confirmed by immunohistochemistry for HLA-G to prevent confusion of remodeled vessels with regressing decidual epithelial glands (Figure 1, G and H). Staining with D2–40 (lymphatic endothelial cell marker) was used to exclude lymphatic vessels from our analysis (Figure 1, I and J).

Open in a separate window

Figure 1

Immunohistochemical identification of remodeling vessels in decidua basalis. A: Unremodeled artery stained for α-smooth muscle actin (SMA). B: Lymphatic vessel stained with D2–40. C: Decidua parietalis stained with CD45 demonstrating the abundance of leukocytes. D–F: Vein exhibiting: (D) a thin layer of VSMC (α-SMA immunostaining), (E) intact endothelium (CD31 immunostaining) and (F) leukocytes clustering around the venous wall (CD45 immunostaining). G: Cytokeratin (CK)−7 immunostaining was present in the glandular epithelium (EG) and extravillous trophoblasts (EVT), while (H) HLA-G immunostaining identified only the EVT. I: CD31 immunostaining stained all endothelial cells and (J) absence of staining for D2–40 on a serial section revealed that this vessel was not lymphatic. K–M: Remodeling vessel exhibiting: (K) disrupted vascular smooth muscle cells (VSMC), (L) loss of endothelium, and (M) absence of EVT in the vessel wall. Inset: Representative negative controls for: (K) α-SMA, (L) CD31, and (M) HLA-G immunohistochemistry. N–P: More extensively remodeled vessel exhibiting: (N) complete loss of VSMC, (O) further loss of endothelium, and (P) EVT presence within the stroma, vessel lumen, and relining the vessel wall. Scale bars: 50 μm (D–F) and 100 μm (A–C and G–P).

Careful examination of serial sections identified a range of stages of vascular remodeling, within each tissue specimen. Unremodeled spiral arteries were detected, characterized by multiple layers of intact circumferential VSMC (Figure 1A) and intact, continuous endothelium. The majority of vessels exhibited features consistent with remodeling, including varying degrees of disruption and loss of thickness and integrity of VSMC layers and endothelium (Figure 1, K and L). A further subset of arteries displayed substantial or complete loss of VSMC and endothelium, with vEVTs present lining the vessel; these were classified as remodeled (Figure 1, N–P).

Importantly, some vessels that showed substantial changes (eg, VSMC disruption and disorganization) contained few vEVTs (Figure 1, K–M) or lacked them altogether, suggesting that such alterations can occur independently of vEVT in the immediate vicinity. This prompted detailed investigation to delineate stages of vascular remodeling and the timing of trophoblast involvement.

Stages of Vascular Remodeling

Four discrete stages of vascular remodeling were defined based on the extent of VSMC disruption and loss (Table 3). Stage I vessels displayed intact VSMC layers and endothelium, with no detectable vEVTs present (Figure 2, A–C). Stage II vessels were characterized by disruption and partial loss of VSMC, extensive disorganization and separation of layers of VSMC indicating loss of cohesion, and breaks in the endothelial layer (Figure 2, E, F, I and J). vEVTs were absent although iEVTs were observed in the nearby decidua in some, but not all, of these vessels (Figure 2, G and K). Stage III vessels exhibited substantial loss of VSMC and endothelium, with vEVTs present in the lumen, some of which were adherent to the vascular wall (Figure 2, M–O). Stage IV vessels exhibited complete loss of VSMC and endothelium, which was replaced by an often continuous layer of vEVTs. Fibrinoid was apparent in most vascular walls in stage IV (Figure 2, Q–S).

Open in a separate window

Figure 2

Stages of vascular remodeling and leukocyte involvement. Remodeling vessels were identified by immunostaining of serial sections with α-smooth muscle actin (SMA), CD31, Cytokeratin (CK)−7/HLA-G and CD45 (leukocyte common antigen). A–D: Stage I: Unremodeled vessels displayed: (A) intact and organized vascular smooth muscle cells (VSMC), (B) intact endothelium, (C) absence of extravillous trophoblast (EVT) from the vessel wall, and (D) leukocytes absent from the vascular wall. E–L: Stage II: Vessels displayed: (E and I) dramatically disrupted and disorganized VSMC, (F and J) loss of endothelium, (G and K) absence of EVT from the vessel wall and lumen and (H and L) leukocyte infiltration into the vascular wall and VSMC layers (arrowheads). M–P: Stage III: Vessels displayed: (M) substantial loss of VSMC, (N) and endothelium, (O) presence of EVT within the vessel lumen and (P) leukocytes were largely absent from the vascular wall. Q–T: Stage IV: Remodeled vessels displayed: (Q) complete loss of VSMC, (R) further (but not complete) loss of endothelium from the vascular wall, (S) EVT relining the vessel and (T) leukocyte absence from the vascular wall. Asterisk denotes fibrinoid. Scale bars; 50 μm (A–L) and 100 μm (M–T).

Table 3

Summary of the Stages of Vascular Remodeling

Stage	I	II	III	IV
Vascular smooth muscle cells (VSMC)	Intact	Disruption, disorganization and partial loss	Substantial loss	Little/none remaining
Endothelium	Intact	Swelling and some breaks	Substantial loss	Little/none remaining
Endovascular trophoblasts (vEVTs)	Absent	Absent	Present in lumen	Present in lumen and relining vessel
Fibrinoid	Absent	Absent	Detected in vessel wall	Detected in vessel wall

Open in a separate window

Analysis of the remodeling status of all vessels within each sample revealed no significant association between gestational age and the progression of vascular remodeling (Figure 3A). However, within an individual tissue sample, there was a tendency for vessels to be either classified as in the early stages or the late stages of remodeling rather than a spread of all stages.

Open in a separate window

Figure 3

Quantification of leukocyte infiltration of remodeling vessels. A: Proportions of vessels within each stage of remodeling with respect to gestational age. No significant difference was found. B: Proportions of leukocytes that infiltrate the wall of vessels per 1000 μm² in the four stages of remodeling. Significantly higher numbers of leukocytes infiltrate vessels in stage II of remodeling than in stage I (P < 0.01), stage III (P < 0.05) and stage IV (P < 0.001). C: Proportions of uNK cells that infiltrate the wall of vessels per 1000 μm² in the stages of remodeling. Significantly higher numbers of uNK cells infiltrate vessels in stage II than in stage I (P < 0.001), stage III (P < 0.05), and stage IV (P < 0.01). D: Proportions of macrophages that infiltrate vessel walls per 1000 μm² in the four stages. Significantly higher numbers of macrophages infiltrate vessels in stage II than in stage I (P < 0.01) and stage IV (P < 0.001).

Leukocyte Infiltration of Remodeling Spiral Arteries

The involvement of leukocytes in vascular remodeling was investigated using immunostaining for CD45 on serial sections. Leukocytes were abundantly distributed throughout the decidual stroma, and particularly clustering around and infiltrating the VSMC layers of vessels undergoing remodeling. Leukocyte infiltration was characterized in relation to the stages of vascular remodeling. Leukocytes were absent from the VSMC layers of intact, unremodeled vessels in stage I (Figure 2D), but were detected infiltrating the vessel wall of actively remodeling stage II vessels (Figure 2, H and L). Fewer leukocytes were detected infiltrating spiral arteries in stage III (Figure 2P) and leukocytes were absent from the vessel wall of fully remodeled stage IV vessels (Figure 2T). On quantification of this phenomenon, we found that a significantly higher number of leukocytes infiltrated vessel walls during stage II of remodeling compared with stage I (P < 0.01), stage III (P < 0.05) and stage IV (P < 0.001) (Figure 3B). Leukocytes infiltrating remodeling vessels (stage II) were identified as CD56+ uNK cells and CD68+ macrophages (Figure 4, A–F). On quantification we found that significantly more uNK cells infiltrated vessel walls during stage II of remodeling compared with stage I (P < 0.001), stage III (P < 0.05), and stage IV (P < 0.01) (Figure 3C). There were also significantly more macrophages infiltrating vessel walls in stage II compared with stage I (P < 0.01) and stage IV (P < 0.001), but not stage III (Figure 3D).

Open in a separate window

Figure 4

Identification of leukocyte subtypes in stage II remodeling vessels. Immunostaining for (A and D) CD45, (B and E) CD56 (uNK cells), and (C and F) CD68 (macrophages) revealed that leukocytes infiltrating remodeling vessels are uNK cells and macrophages (arrowheads). C: Inset– representative negative control. Serial immunostaining (A) for (G–I) matrix metalloproteinases (MMPs) revealed that infiltrating leukocytes express (G) MMP-9 (arrowheads) and (H) MMP-7 (arrowheads), but (I) not MMP-2. J–M: Dual immunofluorescence confirmed that uNK cells and macrophages produce MMPs: (J) cell surface CD56 (red) and cytoplasmic MMP-9 (green) co-localized in cells in the vascular wall of remodeling vessels, (K) CD56 (red) and MMP-7 (green) also co-localized as did (L) CD68 (red) and MMP-9 (green) and (M) CD68 (red) and MMP-7 (green) (arrowheads). N: In situ zymography revealed gelatinase activity in leukocytes infiltrating remodeling vessels, (O) High magnification of (N). P–R: Negative controls: (P) Primary and secondary antibody was omitted to control for autofluorescence, (Q) the first secondary antibody was omitted to control for cross-reaction of the first primary and second secondary antibodies and, (R) mouse IgGs were applied to negative control sections to control for non-specific staining. Scale bars: 20 μm (J–R), 25 μm (A–C and G–I), and 50 μm (D–F).

Apoptosis of Vascular Cells

To determine the fate of the vascular cells during vessel remodeling, a marker of apoptosis (TUNEL) was examined. Discrete nuclear staining was detected in stage II remodeling vessels exhibiting disrupted and disorganized VSMC, with leukocyte infiltration (Figure 5A). Apoptotic cells were occasionally detected in stage III remodeling vessels (not shown). There were no stained cells in stage I unremodeled (Figure 5B) or stage IV remodeled vessels (Figure 5C). Dual immunostaining identified positive cells as a subset of VSMC and endothelial cells in actively remodeling vessels (Figure 5, D–H). α-SMA staining was detected in the cytoplasm; CD31 was detected on the cell surface and TUNEL in the nucleus. No green nuclear staining was detected in the TUNEL negative control (Figure 5I).

Open in a separate window

Figure 5

Identification of apoptotic vascular cells during remodeling. TUNEL immunostaining in (A) a stage II remodeling vessel, arrowheads highlight positive nuclei within the vascular wall. No TUNEL positivity was detected in (B) unremodeled or (C) remodeled vessels. D and E: Dual fluorescent staining for α-smooth muscle actin (SMA) and TUNEL, and for (G and H) CD31 and TUNEL revealed that proportion of TUNEL positive cells were also α-SMA positive and CD31 positive (arrowheads denote dual labeled cells). F: No TUNEL positivity was detected in unremodeled vessels. I: Dual stained TUNEL negative control: section was immunostained for α-SMA (red cells) and TUNEL negative control (FITC-labeled dUTPs applied but TdT enzyme omitted). Scale bars: 20 μm (D–I), 50 μm (A and B), and 100 μm (C).

Discussion

Remodeling of decidual spiral arteries is a complex and tightly regulated process that is critical to successful pregnancy, yet the mechanisms involved are poorly understood. This study aimed to delineate the sequence of events during vascular remodeling to shed light on the respective roles of decidual leukocytes and EVTs. We describe extensive dramatic disruption, disorganization, and loss of VSMC and endothelial cells in the early stages of vascular remodeling and importantly have shown that these events occur in the absence of detectable vEVTs within the vessel. Instead, these remodeling events were associated with a significant infiltration of decidual uNK and macrophages into the VSMC layers. No leukocyte infiltrate was apparent in intact vessels (classified as stage I) or in fully remodeled vessels relined with vEVTs (classified as stage IV). We have also demonstrated that uNK cells and macrophages infiltrating remodeling vessels produce MMP-7 and -9, which could be involved in the degradation of the vascular ECM. Furthermore we show that disrupted spiral artery VSMC and endothelial cells undergo apoptosis in vivo during leukocyte-associated remodeling.

Through detailed examination, we have identified four distinct stages of vascular remodeling, defined by the extent of disorganization and loss of the VSMC and endothelium, and the degree of vEVT colonization. These changes were independent of gestational age; instead the extent of remodeling is likely to reflect the proximity of the decidual sample to the placental attachment site. Although these are based on snapshots of the remodeling process, the degree of VSMC loss provides strong evidence for the temporal sequence of events. Development of a staging system provides a framework to determine the timing of vEVT association with remodeling events and to delineate the potential involvement of leukocytes. This system has enabled the identification of a discrete stage of remodeling, when vEVTs are absent. This is the most comprehensive report of decidual vascular remodeling to date and extends previous observations of subtle trophoblast-independent vascular remodeling.10

These observations provide strong correlative evidence that immune cells of the decidua could play a pivotal transient role in vascular remodeling. A role for uNK cells in spiral artery remodeling has been hypothesized,16,27 but never established in the human. Studies in the mouse have provided definitive evidence of a critical role for uNK cells in spiral artery remodeling. Mice deficient in uNK cells exhibit impaired spiral artery remodeling.12 These mice experience 50% fetal loss and the small surviving fetuses have reduced placental weights. Transplantation of bone marrow from severe combined immunodeficient donors to reconstitute the uNK cell population rescues the phenotype and results in normal vascular remodeling. Interferon-γ was identified as a key uNK cell-derived factor responsible for normal spiral artery remodeling,28 and important roles were defined for interleukin-15 and its signaling receptor (interleukin-2Rγ) in the differentiation of uNK cells in the mouse.29 In humans, uNK cells infiltrating artery walls during stage II of remodeling could be actively involved in the reorganization of the vascular wall as seen in the mouse. Macrophages also infiltrated the vascular wall in stage II but remained present in stage III, suggesting potential further roles in phagocytic clearance of apoptotic cells during more advanced remodeling. Cross talk between uNK cells and macrophages is likely to be involved in vascular remodeling during stages II and III. uNK cells and macrophages are able to interact to initiate an immune response, which occurs via the activating receptor NKG2D.30 NK cells can also advance the maturation and activation of macrophages by interferon-γ signaling.31 On the other hand, macrophages can activate NK cell proliferation and enhance cytokine secretion while limiting cytotoxicity.32 Such reciprocal interactions could facilitate and advance the remodeling process.

MMPs are known to be important for trophoblast invasion and are likely to be key proteases for ECM degradation during vascular remodeling. Here we report the expression of MMP-7 and -9, but not MMP-2, by uNK cells and macrophages infiltrating remodeling arteries, and demonstrate individual foci of gelatinase activity at a time when vascular ECM is visibly degrading. Expression of MMP-2, -7, and -9 has previously been detected in rodent and human peripheral NK cells where they contribute to NK migration to tumor metastases.33,34 While endometrial/decidual macrophages are an established source of MMPs,8 the production of MMP-2 and -9 by uNK cells has only recently been described.35 These findings suggest that leukocyte-derived MMPs contribute to vascular remodeling, consistent with their mediation of focal degradation of endometrial ECM during menstruation.36 Localized secretion of MMPs by leukocytes has been implicated in the pathogenesis of atherosclerosis,37,38 tumor angiogenesis,39 and metastasis.40 MMP-2 and −9 also alter chemokine bioactivity, thereby influencing leukocyte migration.41,42,43 In this way they could potentially mediate tight regulation of leukocyte trafficking during vascular remodeling. Furthermore, recent work by ourselves and others demonstrates that EVTs migrate in response to chemokines.44,45 Secretion of chemokines by immune cells within the remodeling vessels may subsequently attract invading vEVTs.

Previous studies have suggested iEVTs fulfill a priming role in vascular remodeling, enabling subsequent vEVT migration and action.46 In the present study iEVTs were detectable in decidual stroma in all stages, thus a role in vascular remodeling before vEVT arrival cannot be discounted. However, iEVTs were rarely observed intimately associated with the disrupted vessel wall in stage II. In addition, using an in vitro co-culture model we have demonstrated that apparently normal vascular remodeling events can occur in the absence of iEVTs.23 It is likely that in vivo, decidual leukocytes and iEVTs act co-operatively to ensure adequate spiral artery remodeling, potentially fulfilling overlapping or compensatory roles in this essential process.

Despite the abundance of uNK cells and macrophages throughout the decidua (Figure 1C), partially remodeled vessels were never detectable in decidua parietalis. These data suggest that leukocytes must be activated locally to infiltrate the vascular wall, perhaps by placentally derived factors. iEVTs surrounding decidual arteries, or vEVTs within the proximal portion of arteries may release paracrine factors that activate decidual leukocytes to infiltrate the artery walls. This is in contrast to vascular remodeling in the mouse, where leukocytes independently remodel spiral arteries. It is tempting to speculate that leukocytes fulfill a dominant role in the early stages of vascular remodeling, however we do not regard leukocyte-associated vascular remodeling as trophoblast-independent. Instead, as previously suggested,47 we hypothesize a greater degree of cooperation and interaction between leukocytes and EVTs in the sequential, progressive, and controlled transformation of spiral arteries.

The processes involved in loss of VSMCs and endothelial cells from the vascular wall were analyzed by examining apoptosis in remodeling vessels. Studies in vitro have shown that vEVTs can induce VSMC apoptosis.5,6 Currently there are no reports of apoptosis in remodeling vessels in vivo; and a recent report described the failure to identify apoptotic VSMCs.9 Using serial immunohistochemistry and dual immunofluorescence, we detected TUNEL-positive nuclei in VSMCs and endothelial cells of arteries exhibiting signs of VSMC disruption and loss (predominantly in stage II). Though such nuclei are not abundant, these studies support the loss of some vascular cells via apoptosis, before vEVT entry. The failure of previous studies to detect apoptosis in vascular remodeling may be related to the rapidity of cell death, and additionally to a focus on more advanced stages of vascular remodeling (trophoblast-associated), where we similarly were unable to detect apoptosis. Infiltrating uNK cells may be involved in triggering the vascular apoptosis, via death receptor-mediated pathways (eg, Fas ligand or tumor necrosis factor-α-related apoptosis-inducing ligand-mediated).48 Alternatively, MMP-mediated detachment from the ECM may trigger a form of apoptosis known as anoikis.49

Pre-eclampsia is more common is primiparous women, or in those with a new partner or long interpregnancy interval,50 suggesting an immunological component. This has been attributed to a failure of maternal immune tolerance to expression of fetal antigens by EVT at the decidual-placental interface.51 Interactions between EVT expressing HLA-C and uNK cells expressing killer immunoglobulin-like receptors are believed to be key for inducing tolerance.52 However, particular haplotype combinations of killer immunoglobulin-like receptors and paternally derived HLA-C can result in activation of uNK cells and limit trophoblast invasion and spiral artery remodeling.51,53 The potential role for uNK cells and macrophages in early vascular remodeling suggests an alternative or additional immune-related etiology of pre-eclampsia, if maternal immune cells fail to respond adequately to an EVT-mediated signal and fulfil a priming role for subsequent EVT invasion. Despite the fact that uNK cells are not present in the myometrium, decidual artery remodeling is likely to be a prerequisite for successful myometrial artery remodeling. Therefore inadequate decidual remodeling may result in impeded EVT invasion or altered phenotype, and subsequent impaired myometrial remodeling. Macrophages are present within the myometrium,14 and therefore may contribute to myometrial spiral arterial remodeling. Further work examining the placental bed in pathological pregnancies is required to confirm this hypothesis.

In summary, we have shown for the first time that extensive disruption and remodeling of the spiral arteries occurs before endovascular colonization by EVTs, coincident with vascular infiltration by uNK and macrophages. This highlights a potential active, but transient, role for leukocytes in spiral artery remodeling, consistent with early placentation events in the mouse. Further functional studies are required to evaluate leukocyte actions, but we provide evidence for induction of matrix degradation and apoptosis as putative mechanisms. Development of a staging system has aided in delineating the key instigators and sequence of events of vascular remodeling, which is critical if we are to define aberrations that may occur in disease. Though a majority of studies of the pathogenesis of pre-eclampsia have focused on abnormal EVT function or survival, some authors have pointed to the likelihood that alterations in the uterine environment are important in pregnancy pathology.54 We anticipate that these findings will open up new avenues for investigation to decipher the involvement of decidual immune cells in insufficient vascular remodeling and trophoblast invasion.

Acknowledgments

Footnotes

References