Tissue inhibitor of metalloproteinase-4 deletion in mice impacts maternal cardiac function during pregnancy and postpartum

Abstract

Reversible physiological cardiac hypertrophy of the maternal heart occurs during pregnancy and involves extracellular matrix (ECM) remodeling. Previous mouse studies revealed that changes in ECM molecules accompany functional changes in the left ventricle (LV) during late pregnancy and postpartum. We evaluated the effect of global Timp4 deletion in female mice on LV functional parameters and ECM molecules during pregnancy and the postpartum period. Heart weights normalized to tibia lengths were increased in Timp4 knockout (Timp4 KO) virgin, pregnant, and postpartum day 2 mice compared with wild types. Serial echocardiography performed on pregnancy days 10, 12, and 18 and postpartum days (ppds) 2, 7, 14, 21, and 28 revealed that both wild-type and Timp4 KO mice increased end systolic and end diastolic volumes (ESV, EDV) by mid to late pregnancy compared with virgins, with EDV changes persisting through the postpartum period. When compared with wild types, Timp4 KO mice exhibited higher ejection fractions in virgins, at pregnancy days 10 and 18 and ppd2 and ppd14. High-molecular weight forms of COL1A1 and COL3A1 proteins in LV were greater in Timp4 KO virgins, and COL1A1 was higher in late pregnancy and on ppd2 compared with wild types. With exceptions, Timp4 KO mice during late pregnancy and the early postpartum period were able to maintain stroke volume similar to wild-type mice through increased ejection fraction. Although TIMP4 deletion in females exhibited altered ECM molecules, it did not adversely affect cardiac function during first pregnancies and lactation.

NEW & NOTEWORTHY Pregnancy and lactation increase volume load on the heart. Defects in cardiac remodeling during pregnancy and postpartum can result in peripartum cardiomyopathy. TIMPs participate in cardiac remodeling. The present study reports the cardiac function in Timp4 knockout adult female mice during pregnancy and lactation. Timp4 knockout females at many time points have higher ejection fraction to maintain stroke volume. Global deletion of Timp4 was not detrimental to maternal heart function during first pregnancies and lactation.

INTRODUCTION

During pregnancy, the maternal cardiovascular system undergoes physiological adaptations such as increased blood volume to ensure adequate perfusion of maternal, placental, and fetal tissues (1). To accommodate the heightened volume load, the heart hypertrophies accompanied by increased ventricular internal dimensions, left ventricle (LV) wall thickness, and LV mass (2). This hypertrophy is physiological and reverses during the postpartum period although the exact timing of return to baseline has not been fully described, and permanent changes in cardiac geometry have been demonstrated with multiparity (3). Peripartum or postpartum cardiomyopathy can develop when there is inappropriate cardiac remodeling or its reversal, resulting in reduced ejection fraction with or without ventricular dilatation (4, 5). This pathological condition can progress into heart failure without prompt treatment.

For cardiac remodeling to occur during pregnancy, the heart must reorganize the extracellular matrix (ECM) to allow the cardiomyocytes to expand (6, 7). The cardiac ECM is composed of collagens and noncollagenous proteins, proteoglycans, and glycoproteins (8). Physiological remodeling of collagens, the most abundant extracellular proteins, is accomplished by balancing their degradation and new synthesis (9). Collagen degradation is predominantly carried out by matrix metalloproteinases (MMPs) (10). MMPs are regulated by their endogenous inhibitors, tissue inhibitors of metalloproteinases (TIMPs), and it is the combined actions of MMPs and their inhibitors that determine the extent of ECM remodeling (8). Conditions favoring collagen deposition over degradation can lead to fibrosis, however, when assessed histologically, there is little evidence of fibrosis in the heart of normal pregnancy (7, 11–13). Heart failure in patients from several etiologies including postpartum cardiomyopathy exhibits alteration in ECM pathway genes (14).

Collagen type I and III are the most abundant collagens in the heart ECM (15, 16). Collagen I and III subunit mRNAs and proteins are altered in maternal rodent hearts during the peripartum period, but the direction of regulation is not necessarily consistent (17, 18). MMP1, MMP2, MMP9 proteins and Mmp3 mRNA also showed altered levels in rodent LVs during late pregnancy or immediately postpartum compared with nonpregnant animals (7, 18, 19). In addition, Timp1 mRNA and its protein show consistent transient peripartum increases in these rodent studies. Late pregnant rats exhibited a small but significant increase in cardiac TIMP4 protein (7), and a transient Timp4 mRNA decrease in maternal LVs of wild-type mice occurred on the day after parturition (18). These data support the concept that the MMP-TIMP system may play an active role in the cardiac ECM remodeling during pregnancy and the peripartum period.

TIMP4 has been demonstrated to play a role in cardiac remodeling under pathological conditions, where matrix metalloproteinases are also altered. TIMP4 can inhibit MMP2, MMP14, and MMP26 (20). Use of Timp4-deficient C57BL/6 male mice showed TIMP4 was required for cardiac ischemia-reperfusion injury recovery, and that Timp4-deficient myocardium from ischemic-injured mice had increased MMP2 activity (21). In another study, Timp4 KO mice, on an FVB background, subjected to LV pressure overload had lower survival rates than wild types or mice overexpressing TIMP4 in the heart (22). In a pathological model of cardiac volume overload, male mice exhibited decreased Timp4 mRNA along with increased diastolic LV diameter and reduced ejection fraction (23). Male Timp4 KO mice (C57BL/6 background) exhibit cardiac changes of increased ventricular mass and increased posterior wall thickness with advanced age (20 mo old) (24). Unlike pathological models of acute volume overload, pregnancy causes a more gradual volume overload condition that is influenced by changes in the hormonal milieu of pregnancy (1, 25). The data with these aforementioned models indicate an important role for TIMP4 in pathological cardiac remodeling, however, whether TIMP4 loss affects the functional remodeling of the heart during the physiological volume overload condition of pregnancy and postpartum lactation is unknown.

A few human studies have evaluated serum TIMP and MMP levels during normal and abnormal pregnancies (26, 27). There are no studies of human cardiac tissue and ECM mRNAs and proteins during pregnancy. Genetic rodent models are useful for delineating the role of specific ECM proteins in cardiac function during different physiological states such as pregnancy. In the current study, we aimed to determine if Timp4 KO adult female mice functionally remodel their hearts differently than wild-type mice during pregnancy and the postpartum period, and whether the loss of TIMP4 in females alters ejection fraction. We used serial echocardiography during pregnancy and the first postpartum month (through the pup weaning period) to assess LV function in wild-type and Timp4 KO mice. In addition, we evaluated mRNAs for ECM components including TIMPs, MMPs, and collagens and proteins of interest to determine if Timp4-deficiency altered their expression during late pregnancy and postpartum compared with virgin mice.

METHODS

Animals and Experimental Groups

The animal protocol was approved by the University of South Carolina Institutional Animal Care and Use Committee and all university guidelines were followed for animal use. Virgin wild-type C57BL/6 mice were purchased from Jackson laboratory (No. 000664, Bar Harbor, ME). Global Timp4 KO mouse founders (C57BL/6 background) were obtained from Dr. Zamaneh Kassiri, University of Alberta (21). Animals were maintained under standard conditions and bred by homozygous pairings at the University of South Carolina animal resource facility. For each genotype, starting at 9 wk of age, the estrous cycles of virgin female mice intended for mating were tracked by morning vaginal lavage and cell distribution as previously described (28). To produce age-matched controls for pregnancy, another group of virgin mice for each genotype had their estrous cycles tracked beginning at ∼3 mo of age. Mouse estrous cycles were tracked for at least two consecutive cycles and a virgin echocardiogram was performed in diestrus. Diestrus was chosen for virgins as a period of low estrogen, stable energy expenditure, and based on similar previous study designs (19, 29, 30). For virgin mice that were mated, a single female mouse was paired with a single male mouse on the day after the virgin baseline echocardiogram. Mice were checked daily in the morning for the presence of a copulation plug or the presence of sperm in the vaginal lavage, which when present was denoted day 0 of pregnancy or embryonic day 0 (ed0). Pregnancy was confirmed on day 10 (ed10) and/or day 12 (ed12) using ultrasound (see Echocardiography) by the presence of live fetuses in the uterus. Male mice were removed after pregnancy confirmation on ed12. Dams that gave birth were not disturbed until postpartum day 2 (ppd2) to reduce the chance of pup cannibalism that we previously observed on ppd1.5 (18), and pups were weaned on ppd21 after the maternal echocardiogram was performed. The experimental groups for each genotype were euthanized as virgins in diestrus, and on day 18 of pregnancy (ed18), ppd2, and ppd28. Mice not gestating at least four fetuses (for the ed18 group), not giving birth to at least four pups (ppd2 and ppd28 groups) or not lactating four pups (ppd28 group), were not included in subsequent analyses. A flow chart of the experiment is shown in Fig. 1 with the number of animals euthanized at each reproductive status time point and exclusions. At the time of euthanasia, mice were euthanized by isoflurane inhalation until the toe pinch reflex was lost, followed by cervical dislocation. Euthanized mice were weighed, the heart was removed and flushed with PBS to remove blood, squeezed lightly to remove fluid, and gently blotted before weighing intact. The LV wall was dissected then cut into two pieces, snap frozen, and stored at −75°C until RNA and protein were isolated. Tibias were removed and measured after 3 days of soaking in 1% NaOH.

Figure 1.

Experimental flow chart showing the total number of animals undergoing baseline cardiac ultrasound (virgin echo), number of animals euthanized at each reproductive status time point, and the number of animals excluded. “Used” indicates the animals in Fig. 2 and that left ventricle (LV) tissue was saved for mRNA and protein studies. ed, embryonic day; ppd, postpartum day.

Echocardiography

Adult female mice underwent transthoracic two-dimensional echocardiography to measure LV internal diameters using a parasternal short-axis view. All diestrous virgin mice had a baseline echocardiogram. Females that were mated also had serial echocardiography performed on ed10, ed12, and ed18 and ppd2, ppd7, ppd14, ppd21, and ppd28. For groups of virgins, ed18, ppd2, and ppd28 mice, euthanasia was performed within 1–3 h of their last ultrasound, and the estrous cycle stage of nonpregnant mice was recorded. For echocardiography, adult females were initially anesthetized using 3%–3.5% inhaled isoflurane (Zoetis, Kalamazoo, MI) in oxygen and maintained at 1%–1.5% isoflurane during sonography. Hair was removed from the thoracic region and abdominal region (for pregnant mice) before sonography. Sonography was performed using a VEVO 3100 high-resolution imaging system (FUJIFILM VisualSonics, Toronto, ON, Canada) with a 50-mHz transducer. Data for mice were collected in the heart rate range of 425–435 beats/min, with average heart rate being 429 ± 3.9 for C57BL/6 wild-type mice and 427 ± 3.2 for Timp4 KO mice. M-mode measurements were performed at midpapillary level. With the use of postacquisition routines (Vevo LAB 5.5.1 software; FUJIFILM VisualSonics) and established algorithms, the measures for LV volumes, ejection fractions, cardiac output, and cardiac index (mL/min/g) were computed (31, 32). Power analyses determined that a minimum of eight mice were needed for echocardiography studies. Echocardiogram analyses were performed by both blinded and nonblinded laboratory members. Bland–Altman analysis (Graphpad Prism v.9.4, San Diego, CA) of LV percent ejection fraction was used as a measure of interobserver variability (33) and indicated minimal bias of 0.89% with an standard deviation of bias 3.11 and 95% limits of agreement −5.19 to 6.98%.

Quantitative PCR

Messenger RNA levels for Timp1–4, Mmp2, -3, -9, -13, -14, and -15 and collagens Col1a1, Col3a1, Col5a1, and Col8a1 and control gene Rplp0 were measured by qPCR, and ECM mRNAs were chosen because they have been shown to be regulated or lack regulation during pregnancy and postpartum (7, 17–19). One piece of frozen LV tissue was homogenized in Ribozol reagent (VWR Life Science, Atlanta, GA) using a dounce homogenizer and cleared at 12,000 g at 4°C to pellet debris. Subsequent RNA isolation was performed from the cleared supernatant using the Direct-zol RNA MiniPrep Plus kit with DNase I treatment (Zymo Research, Irvine, CA) according to the manufacturer’s instructions. RNA was quantified at A260 using a NanoDrop One (Thermo Fisher Scientific, Waltham, MA). Approximately 200 ng of RNA per reaction was used for cDNA synthesis using the iScript cDNA synthesis kit (Bio-Rad, Hercules, CA) according to the manufacturer’s instructions. All PCR primers were purchased from Qiagen (Thermo Fisher Scientific) except for Col1a1 that was synthesized by Integrated DNA Technologies (Coralville, IA) as previously described (18). Primer amplicons were confirmed as single products and efficiencies determined in a previous study (18). qPCR was performed with 5 ng cDNA, primer mix, and Sso-Advanced Universal SYBR Green Supermix (Bio-Rad) and an I-cycler (Bio-Rad). Cycling conditions were 95°C for 1.5 min, followed by 35 cycles of 95°C for 15 s, 60°C of 15 s [58°C for Col1a1 (18)], and 72°C for 30 s followed by a 10-min final extension at 72°C and cycling to generate melt curves. Amplifications were performed with two or more replicate wells. The standard curve method was used to determine mRNA amounts (34). Target mRNA amounts were extrapolated from standard curves of their serially diluted amplicon using their threshold cycle (Cq) values. Target mRNA amounts were then normalized to their respective Rplp0 mRNA amounts, which were also determined from their Cq values and standard curves. Ratios were multiplied by 100.

Protein Isolation and Western Blot Analysis

One piece of frozen LV tissue was homogenized with 15–20 strokes of a dounce homogenizer in cold lysis buffer containing 50 mM Tris (pH 7.4), 1% Igepal, 0.5% sodium deoxycholate, 0.1% SDS, 150 mM NaCl, 1 mM EDTA, and 1× Halt Protease and phosphatase inhibitor (Thermo Scientific) on ice. Protein extracts were isolated after a 30-min incubation on ice and subsequent centrifugation at 15,000 g. Protein amounts in supernatants were quantified using Bio-Rad Dye reagent. Proteins (50 µg) were separated by electrophoresis using precast Criterion TGX 4%–20% acrylamide gels (Bio-Rad). To detect high-molecular weight forms of collagen protein, nonreducing conditions were used as described (35). Nonreducing conditions lacked reducing agent in the Laemmli sample buffer, and samples were not heated. SDS was absent from gels but was present in the running and sample buffers. For MMP2 and MMP9 immunoblots, reducing conditions were used by adding 2-mercaptoethanol to the sample and heating at 100°C for 8 min. Proteins were transferred onto Amersham Hybond P PVDF membranes (MilliporeSigma, Burlington, MA) by Criterion blotter tank transfer. Immunoblotting was performed by blocking membranes in 5% nonfat milk in 1× Tris-buffered saline containing 0.05% Tween-20 (1× TTBS) for 1 to 2 h. The primary antibodies used were rabbit anti-COL1A1 (Bioss, 10423 R 1:500), COL3A1 (Invitrogen, PA5-92066 1:500), COL8A1 (Invitrogen, PA5-97604 1:500), MMP2 (Bioss, bs-4599R 1:500), MMP9 (Abcam, Waltham, MA ab38898 1:500), and α-tubulin (11224-1-AP, 1:2,000, Proteintech, Rosemont, IL). Bioss and Invitrogen antibodies were purchased from Thermo Fisher Scientific. Primary antibodies were diluted in blocking solution and incubated overnight at 4°C with rocking and followed by washing three times with 1× TTBS (10 min each). The secondary antibody used was affinity purified goat anti-rabbit IgG (A0545, 1:8,000 to 1:10,000, MilliporeSigma) and incubated in blocking solution for 1 h followed by five washes with 1× TTBS (10 min each). Membranes were stripped and reblocked between antibody probings. Proteins bands were visualized using Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific) and blue X-ray film (Research Products International, Mt Prospect, IL). Proteins bands were quantified using UN-SCAN-IT gel 7.1 (Silk Scientific, Orem, UT). Proteins were normalized to α-tubulin from the same blot (55 kDa).

Zymography

LV protein extracts (50 µg) were electrophoretically separated using nonreducing conditions and 8% SDS-PAGE gels containing type A gelatin [1 mg/mL; MilliporeSigma (36)]. Gels were incubated in 2.5% Triton X-100 buffer, 50 mM Tris (pH 7.5) to renature MMPs and rocked overnight with incubation buffer, consisting of 1% Triton X-100, 50 mM Tris (pH 7.5), 5 mM CaCl₂, and 1 µM ZnCl₂ at 37°C. Gels were stained for 45 min with staining solution, consisting of 40% methanol, 10% acetic acid, 5 mg/mL Coomassie Blue R-250, and destained (40% methanol, 10% acetic acid) until light bands could be clearly observed against a dark background. The coomassie band at ∼75 kDa was used as a loading control. Gels were imaged with a Bio-Rad Molecular Imaging Gel Doc XR⁺ and white regions, and the coomassie bands were quantified using UN-SCAN-IT. Each gel contained a virgin, ed18, ppd2, and ppd28 sample from each genotype. Different gels containing different mice were each normalized to the wild-type ppd28 sample on the same gel.

Statistics

Statistical analyses were performed using Graphpad Prism Software version 9.4. Weight, mRNA, protein, and zymography data were ln-transformed and subjected to two-way ANOVA, to analyze the main effects of reproductive status group (time point) and genotype and their interaction. Reproductive status was analyzed post hoc by Tukey’s multiple comparison test when its main effect or the interaction was significant. Sidak’s multiple comparison test was used post hoc when genotype main effect or interaction was significant to compare the genotypes on each reproductive status day. Echocardiographic data that included serial measurements of pregnant and postpartum mice were analyzed and nontransformed by mixed-effects analyses with matching followed by Tukey’s and Sidak’s tests as described earlier. P < 0.05 was considered significantly different. Cardiac output and cardiac index that were determined for mice on the day of euthanasia were nontransformed and analyzed by two-way ANOVA as described for nonserial measurement data. Data are presented as means ± SD.

RESULTS

Mouse Reproductive Characteristics

The mean age of mice at euthanasia was similar between genotypes, 112 ± 16.2 days for wild-type and 116 ± 17.6 days for Timp4 KO mice. We evaluated several reproductive end points between the two genotypes including time to first litter, gestational duration, fetal number at euthanasia (ed18 group), number of pups born, live pups on ppd2, and number of pups weaned for adult females used in this study including breeders; there were no significant differences (Supplemental Table S1). Analyses of estrous cycle length and phase duration in virgin mice revealed similar overall length of estrous cycle (wild types, 5.3 ± 0.7 vs. Timp4 KO, 5.2 ± 0.6 days). Analyses of proestrus, estrus, and metestrus/diestrus phases revealed a significantly reduced estrus phase duration in Timp4 KO mice compared with wild types (wild types, 1.8 ± 0.5 vs. Timp4 KO, 1.5 ± 0.5 days, P = 0.003, ∼7 h difference), with no significant difference in the duration of other phases. For postpartum mice shown in Fig. 2 that were euthanized at ppd2, wild-type mice were 12.5% in proestrus, 12.5% estrus, and 62.5% combined metestrus/diestrus, whereas Timp4 KO mice exhibited 0% in proestrus, 20% estrus, and 80% combined metestrus/diestrus. For mice euthanized at ppd28, wild-type mice exhibited 12.5% in proestrus, 62.5% in estrus, and 25% combined metestrus/diestrus, whereas Timp4 KO mice exhibited 11% proestrus, 22% estrus, and 67% combined metestrus/diestrus. The difference in estrus percentage between genotypes could be due to the finding in virgins that Timp4 KO have shorter estrus phase than wild types.

Figure 2.

Heart weights (HWs), body weights (BWs), and heart weights normalized to body weights (HW/BW) or tibia lengths (HW/TL) from virgin, late pregnant (embryonic day 18, ed18), and postpartum days 2 and 28 (ppd2 and ppd28, respectively) wild-type and Timp4 knockout (KO) mice. Ln-transformed data were analyzed by two-way ANOVA to determine the effects of reproductive status and genotype on weights and ratios. Post hoc comparisons of different reproductive days within each genotype were evaluated by Tukey’s multiple comparison test. Post hoc comparisons for each reproductive status day between genotypes was performed using Sidak’s multiple comparison test. Bars indicate means ± SD, and dots represent individual mice. n = 8–11 mice/group. Within genotype, P < 0.05, ^adiffers from the virgin, ^bdiffers from ed18, and ^cdiffers from ppd2. *P < 0.05, difference between genotypes for the same reproductive status day.

Heart and Body Weights in Wild-Type and Timp4 KO Mice in Different Reproductive States

Heart weights (HWs), body weights (BWs), and heart weights normalized to either body weight (HW/BW) or tibia length (HW/TL) were evaluated in wild-type and Timp4 KO mice as virgins in diestrus, at late pregnancy (ed18), and at early (ppd2), and later (ppd28) postpartum days (Fig. 2 and Table 1). Two-way ANOVA was used to assess main effects of reproductive status, genotype, and their interaction. There were significant simple main effects of reproductive status on HW, BW, HW/BW, and HW/TL (P < 0.0001 for each). There were significant simple main effects of genotype on HW, BW, and HW/TL (P < 0.0001 for each), but not HW/BW. There were no significant interactions of reproductive status and genotype for HW, BW, HW/BW, or HW/TL. Tukey’s post hoc comparison of reproductive status days within genotype indicated that in wild types, HWs by themselves in wild types were increased at ed18 (P = 0.03), ppd2 (P = 0.004), and ppd28 (P < 0.0001) compared with virgins, and the ppd28 group was increased versus the pregnant group (ed18, P = 0.0003). In Timp4 KO mice, ppd2 and ppd28 HWs (P < 0.0001 for each) were increased compared with virgins, and the ppd28 (P = 0.006) group was increased compared with ed18. Within wild types, compared with virgins BWs were increased on ed18 (P < 0.0001), ppd2 (P = 0.03), and ppd28 (P < 0.0001) and ppd2 and ppd28 (P < 0.0001 for each) were decreased compared with ed18, and ppd28 (P = 0.04) was increased compared with ppd2. For Timp4 KO mice, BWs were heavier than virgins at ed18, ppd2, and ppd28, and decreased at ppd2 and ppd28 compared with ed18 (P < 0.0001 for each). When heart weights were normalized for their respective body weights, wild-type mice were lower on ed18 (P < 0.0001) and ppd28 (P = 0.03) higher than virgins, and ppd2 and ppd28 higher than ed18 (P < 0.0001 for each). For Timp4 KO mice, HW/BW means were lower for ed18 than virgins, and ppd2 and ppd28 higher than the ed18 group (P < 0.0001 for each). Heart weight was also normalized by tibia length, as tibia length is more stable than body weight especially during pregnancy and the early postpartum period. In wild-type mice, HW/TL means were higher in ppd2 (P = 0.0004) and ppd28 (P < 0.0001) mice compared with virgins, and ppd28 (P = 0.002) were higher compared with ed18 mice. A similar pattern of HW/TL mean differences was observed within Timp4 KO mice with ppd2 and ppd28 (P < 0.0001 for each) mice higher compared with virgins, and the ppd28 (P = 0.009) group higher compared with ed18 mice.

Table 1.

Characteristics of mice

	Reproductive Status Group
Parameter	Virgin	ed18	ppd2	ppd28
Heart weight, mg
Wild type	99.38 ± 15.24	111.55 ± 10.00	120.14 ± 13.27	135.88 ± 13.20
Timp4 knockout	112.49 ± 7.05	125.91 ± 10.44	140.95 ± 14.90	147.29 ± 9.21
Body weight, g
Wild type	20.58 ± 1.66	33.52 ± 3.66	22.86 ± 1.43	25.47 ± 1.73
Timp4 knockout	22.49 ± 1.66	36.10 ± 2.64	27.47 ± 2.15	27.49 ± 2.03
Heart weight/body weight, mg/g
Wild type	4.82 ± 0.49	3.34 ± 0.29	5.26 ± 0.48	5.35 ± 0.63
Timp4 knockout	5.01 ± 0.28	3.49 ± 0.19	5.13 ± 0.33	5.37 ± 0.30
Heart weight/tibia length, mg/mm
Wild type	6.02 ± 0.82	6.56 ± 0.44	7.23 ± 0.80	7.74 ± 0.79
Timp4 knockout	6.68 ± 0.35	7.38 ± 0.65	8.14 ± 0.91	8.50 ± 0.55
Cardiac output, ml/min†
Wild type	12.71 ± 3.34	16.77 ± 2.67a	16.24 ± 2.97	19.07 ± 4.16a
Timp4 knockout	16.17 ± 2.60*	19.20 ± 1.35	17.23 ± 2.63	18.18 ± 3.18
Cardiac index, ml/min/mg body wt§
Wild type	0.63 ± 0.14	0.52 ± 0.09	0.67 ± 0.09b	0.74 ± 0.12b
Timp4 knockout	0.72 ± 0.10	0.53 ± 0.05c	0.63 ± 0.09	0.67 ± 0.15

Values are means ± SD of characteristics as shown in Fig. 2, as well as other statistical comparions. ed18, embryonic day 18; ppd2, postpartum day 2; ppd28, postpartum day 28.

^†Two-way ANOVA reproductive status, P = 0.0003, and genotype, P = 0.04;

^*differs between genotypes by Sidak’s test, P = 0.04;

§two-way ANOVA reproductive status, P < 0.0001.

^aDifferent from virgin within the same genotype by Tukey’s test, P = 0.03 (ed18), and P = 0.0002 (ppd28);

^bdifferent from ed18 within the same genotype by Tukey’s test, P = 0.04 (ppd2), and P = 0.001 (ppd28);

^cdifferent from virgin within the same genotype by Tukey’s test, P = 0.003 (ed18).

When comparing genotypes by Sidak’s post hoc test, HWs were increased in Timp4 KO mice compared with wild types at virgin (P = 0.01), ed18 (P = 0.04), and ppd2 (P = 0.004) time points, but not at ppd28. BW were significantly higher in Timp4 KO virgin (P = 0.048) and ppd2 (P < 0.0001) mice than wild types. When heart weight was normalized for body weight, there were no differences between genotypes. When heart weight was normalized for tibia length, values for Timp4 KO mice were increased compared with wild types at virgin (P = 0.03), ed18 (P = 0.04), and ppd2 (P = 0.03) time points, but not at ppd28.

Echocardiographic Assessment

Serial echocardiography was performed on adult mice to determine LV structure and function before pregnancy, and during pregnancy and postpartum (Fig. 3 and Supplemental Fig. S1). In addition, virgin mice in diestrus that were not paired underwent cardiac ultrasound to serve as the baseline control group. Mixed-effects analyses was used to assess the baseline (virgin) and serial echocardiographic measurements for main effects of reproductive status and genotype and their interaction. Simple main effects analyses for reproductive status showed significant effects on LV end-systolic volume (ESV), end-diastolic volume (EDV), stroke volume (SV), and ejection fraction (EF) (P < 0.0001 for each). Simple main effects analyses for genotype showed significant effects on ESV (P < 0.0001), EDV (P < 0.02), SV (P = 0.0004), and EF (P < 0.0001). Reproductive status and genotype showed a significant interaction only for ESV (P = 0.04).

Figure 3.

Left ventricular (LV) functional parameters from echocardiography in virgin; pregnancy days 10, 12, and 18 [embryonic day 10 (ed10), ed12, and ed18, respectively], and postpartum days 2 (ppd2), 7, 14, 21, and 28 wild-type and Timp4 knockout (KO) mice. Bars indicate means ± SD. Per genotype, the virgin group consists of all virgin mice used for the experiment. For virgin, ed10, ed12, ed18, ppd2, ppd7, ppd14, ppd21, and ppd28, n = 38, 25, 27, 22, 19, 8, 8, 8, and 8 wild-type mice per group, respectively, and n = 53, 31, 32, 29, 22, 9, 9, 10, and 10 for Timp4 KO groups, respectively. EDV, end-diastolic volume; EF, ejection fraction; ESV, end-systolic volume; SV, stroke volume. LV echocardiographic data collected by single (virgins) or serial measurements of pregnant and postpartum mice were analyzed by Mixed-Effects Analyses with matching followed by Tukey’s test for comparisons of within genotype differences between reproductive status days, and Sidak’s test for comparisons between the two genotypes at the same reproductive status day. Within genotype, P < 0.05, ^adiffers from the virgin, ^bdiffers from ed10, ^cdiffers from ed12, ^ddiffers from ed18, ^ediffers from ppd2, and ^gdiffers from ppd14. *P < 0.05, difference between genotypes for the same reproductive status day.

For comparison within genotype by reproductive status by Tukey’s test, wild-type ESV for ed18 (P = 0.0007), ppd2 (P = 0.004), ppd7 (P = 0.001), ppd14 (P = 0.0003), ppd21 (P = 0.01), and ppd28 (P < 0.0001) were significantly higher than virgins. In addition, wild-type ESV for ed18 (P = 0.005), ppd2 (P = 0.02), ppd7 (P = 0.004), ppd14 (P = 0.001), ppd21 (P = 0.03), and ppd28 (P < 0.0001) were significantly higher than ed10. Furthermore, wild-type ESV at ed18 (P = 0.02), ppd7 (P = 0.01), ppd14 (P = 0.003), and ppd28 (P < 0.0001) were higher than ed12; ppd28 ESV was also higher than ed18 (P = 0.01) and ppd2 (P = 0.009). In Timp4 KO mice, ESV for ed18 (P = 0.004), ppd7 (P = 0.0002), ppd21 (P = 0.01), and ppd28 (P = 0.001) were higher than virgins, and ed18 (P = 0.04), ppd7 (P = 0.001), ppd21 (P = 0.046), and ppd28 (P = 0.007) were higher than the ed10 group. When comparing genotypes by Sidak’s post hoc test, LV ESV was lower in Timp4 KO mice compared with wild types as virgins (P = 0.004), ed18 (P = 0.003), ppd2 (P = 0.0009), ppd14 (P = 0.001), and ppd28 (P = 0.0006) time points, and the ed10 group approached significance (P = 0.052).

For comparison within genotype by reproductive status by Tukey’s test, wild-type EDV for ed18, ppd2, ppd7, ppd14, ppd21, and ppd28 (P < 0.0001 for each) were higher than virgins. Wild-type EDV was increased in the ed18 (P = 0.003), ppd2 (P = 0.04), ppd7 (P = 0.0004), ppd14 (P < 0.0001), ppd21 (P = 0.0001), and ppd28 (P < 0.0001) groups compared with ed10. In addition, wild-type EDV was higher at ed18 (P = 0.02), ppd7 (P = 0.003), ppd14, (P = 0.0002), ppd21 (P = 0.0008), and ppd28 (P < 0.0001) than ed12. Furthermore, wild-type ppd28 EDV was also higher than ed18 (P = 0.003) and ppd2 (P = 0.0004). In Timp4 KO mice, compared with virgins, EDV was higher for all reproductive status groups from ed12 through ppd28 (P < 0.001 for ppd2, and P < 0.0001 for all other groups) and ppd7 (P = 0.0003), ppd14 (P = 0.0002), ppd21 (P = 0.0007), and ppd28 (P = 0.02) were higher than the ed10 group. In addition, Timp4 KO EDV was higher at ppd7 (P = 0.04) and ppd14 (P = 0.03) than ed12 and ppd7 (P = 0.01), ppd14 (P = 0.01), and ppd21 (P = 0.03) higher than in ppd2 group. When comparing genotypes by Sidak’s post hoc test, LV EDV was significantly lower only in ppd28 Timp4 KO mice (P = 0.008) when compared with wild types.

Post hoc comparison of SV by reproductive status within wild types showed LV SV was higher than virgins in the ed12 (P = 0.02), ed18 (P = 0.01), ppd7 (P = 0.01), ppd14 (P = 0.002), ppd21 (P < 0.0001), and ppd28 (P = 0.002) groups. Within Timp4 KO mice, SV was higher than virgins in the ed10 (P = 0.0008), ed12 (P = 0.001), ed18 (P = 0.001), ppd7 (P = 0.005), ppd14 (P < 0.0001), and ppd21 (P < 0.0001) groups. Furthermore, in Timp4 KO mice the ppd14 group SV was higher than the ed10 (P < 0.0001), ed12 (P < 0.0001), ed18 (P < 0.0001), ppd2 (P < 0.0001), and ppd28 (P = 0.002) groups. Comparing genotypes, SV was higher in the Timp4 KO ed10 (P = 0.045) and ppd14 (P = 0.005) groups than in wild types.

Mean EF in virgin wild-type mice was 65.0 ± 11.90 (n = 38) and in virgin Timp4 KO mice was 77.2 ± 11.77 (n = 53). For wild types, the lowest mean ejection fraction was 51.4 ± 7.87 (n = 8) at ppd28 and the highest was 66.5 ± 9.82 (n = 25) at ed10. For Timp4 KO, the lowest mean ejection fraction was 64.7 ± 12.57 (n = 9) at ppd7 and the highest was 79.4 ± 10.89 (n = 31) at ed10. Post hoc comparison of LV EF by reproductive status within wild types showed ppd28 was lower than virgins (P = 0.04), ed10 (P = 0.02), and ed12 (P = 0.03) groups. Within Timp4 KO mice, EF was lower in the ppd7 (P = 0.02) group than virgins, and ed18 (P = 0.04), ppd7 (P = 0.006), and ppd28 (P = 0.03) had lower EF than the ed10 group. When comparing genotypes, EF was higher in the Timp4 KO mice as virgins (P < 0.0001) and ed10 (P = 0.0005), ed18 (P = 0.002), ppd2 (P = 0.0003), and ppd14 (P = 0.005) groups compared with wild types.

Means and individual data points for LV internal diameter during systole and diastole (LVIDs and LVIDd, respectively) used for ESV and EDV calculations, and LV posterior wall thickness during systole and diastole (LVPWTs and LVPWTd) are shown in Supplemental Fig. S1. Two-way ANOVA showed a significant main effect of genotype (P < 0.0001) on LVPWTs. LVPWTs was significantly higher in Timp4 KO mice for virgins (P = 0.0005) and ed12 (P = 0.04) and ed18 (P = 0.001), and ppd2 (P = 0.0006). There were no differences for LVPWTd.

The cardiac outputs (CO) and cardiac indexes (CI) for mice euthanized on specific days and shown in Fig. 2 are reported in Table 1. Two-way ANOVA showed significant simple main effects of reproductive status (P = 0.0003) and genotype (P = 0.04) for CO, whereas for CI the only significant main effect was for reproductive status (P < 0.0001). Within wild-type, CO of the ed18 (P = 0.03) and ppd28 (P = 0.0002) groups were higher than virgins. For CI wild-type, ppd2 (P = 0.04) and ppd28 (P = 0.001) were higher than the late pregnancy ed18 group. Within Timp4 KO mice, CI for the ed18 group (P = 0.003) was lower than virgins. Between genotypes, CO differed for virgins (P = 0.04) only.

LV mRNA Levels for Timps

The mRNAs for Timps 1–4 in LVs of mice on the day of euthanasia were evaluated from virgins in diestrus, late pregnancy, and at postpartum days 2 and 28 (Fig. 4). A two-way ANOVA was performed to analyze the effect of reproductive status and genotype for each mRNA. There was no significant interaction of reproductive status and genotype for any Timp mRNA. Simple main effects analyses showed that reproductive status only had a significant effect on Timp1 mRNA (P < 0.0001), but not other Timp mRNAs. Simple main effects analyses showed genotype had a significant effect on Timp1 (P = 0.029), Timp2 (P = 0.012), and Timp4 (P < 0.0001) mRNAs, but not Timp3 mRNA. Post hoc comparisons within genotype, indicated wild-type ppd2 LV Timp1 mRNA levels were higher than virgin (P = 0.0004), ed18 (P = 0.025), and ppd28 (P = 0.04) mice, whereas in Timp4 KO mice the ppd2 Timp1 mRNA was greater than virgin mice only (P = 0.002). Timp2, Timp3, and Timp4 mRNAs did not differ with reproductive status on the days examined. Timp4 KO mice exhibited negligible Timp4 mRNA. When comparing genotypes, Timp1 and Timp3 mRNAs did not significantly differ (Fig. 4). Timp2 mRNA differed significantly between genotypes only at ppd28 (P = 0.019), with mRNA levels being lower in Timp4 KO mice. As expected, Timp4 mRNA levels were different between the genotypes for all reproductive status days (P < 0.0001).

Figure 4.

Left ventricular (LV) mRNA levels of Timp1, 2, 3, and 4 for wild-type and Timp4 knockout (KO) mice in different reproductive states including virgin, late pregnant (embryonic day 18, ed18), and postpartum days 2 and 28 (ppd2 and ppd28, respectively). mRNAs were quantified by qPCR and standard curve method using for Rplp0 mRNA for normalization. Bars with lines indicate means ± SD, and dots represent individual mice. n = 8 to 9 animals/group. Ln-transformed data were analyzed by two-way ANOVA to determine the effects of reproductive status and genotype on mRNA levels. Post hoc comparisons of different reproductive days within each genotype were evaluated by Tukey’s multiple comparison test. Post hoc comparisons of mRNA levels for each reproductive status day by genotype were performed using Sidak’s multiple comparison test. P < 0.05, ^asignificantly different compared with virgins, ^bdiffers from ed18, and ^cdiffers from ppd2. *P < 0.05, difference between genotypes for the same reproductive status day.

LV Collagen mRNAs and Proteins

We examined specific LV mRNAs and proteins for collagens I, III, and VIII as their mRNAs and/or proteins have previously exhibited regulation during pregnancy or postpartum in rodents (12, 17, 18). Wild type and Timp4 KO euthanized as virgins, late pregnant, and postpartum days 2 and 28 were evaluated. Col5a1 mRNA was included because it was not previously regulated in pregnant or postpartum wild-type mice (18). The mRNAs for collagens Col1a1, Col3a1, Col8a1 (Fig. 5A), and Col5a1 (Supplemental Fig. S2) were analyzed by qPCR, and data were analyzed by two-way ANOVA for main effects of reproductive status and genotype, and their interaction. Simple main effects analyses of reproductive status revealed significant effects on Col3a1 (P < 0.0001), Col5a1 (P = 0.02), and Col8a1 (P < 0.0001) mRNA levels, but not Col1a1 mRNA levels. Simple main effects analyses of genotype showed a significant effect on Col3a1 (P = 0.006), but not other collagen mRNAs examined, although the effect on Col1a1 mRNA (P = 0.057) was just above the significance limit. There were no significant interactions of reproductive status and genotype for any of the collagen mRNAs examined.

Figure 5.

Quantification of specific collagen mRNAs and proteins in the left ventricles (LVs) of wild-type and Timp4 knockout (KO) mice in different reproductive states including virgin, late pregnant (embryonic day 18, ed18), and postpartum days 2 and 28 (ppd2 and ppd28, respectively). A: target collagen subunit mRNAs were quantified by qPCR and standard curve method using for Rplp0 mRNA for normalization. B: nonreducing gels and Western blot analysis were used to identify high-molecular mass collagen subunits. Kilodaltons represent location of molecular mass markers. Two different mice are shown per group, and each row represents images from the same gel, film, and exposure time. The brackets on the left indicate the region quantified. C: Western blot data were quantified and normalized for tubulin monomer (55 kDa). Dots represent individual animals, and bars and lines indicate means ± SD. Ln-transformed data were analyzed by two-way ANOVA to determine the effects of reproductive status and genotype on mRNA levels. Post hoc comparisons of different reproductive days within each genotype were evaluated by Tukey’s multiple comparison test. Post hoc comparisons of mRNA levels for each reproductive status day by genotype were performed using Sidak’s multiple comparison test. For mRNA, n = 8 to 9 mice/group. For protein n = 4 mice/group. P < 0.05, ^asignificantly different compared with virgin; ^bdiffers from ed18; and ^cdiffers from ppd2. *P < 0.05, difference between genotypes for the same reproductive status time point.

Post hoc Tukey’s analyses were used to evaluate mRNA levels for differences between the reproductive states of the mice within each genotype (Fig. 5A and Supplemental Fig. S2). LV Col3a1 mRNA levels were greater in wild types at ppd2 (P < 0.0001) and ppd28 (P = 0.04) than virgins, and ppd2 was greater than ed18 (P = 0.0005). Interestingly, Col3a1 mRNA was not altered in Timp4 KO mice by reproductive status. In both wild-type and Timp4 KO mice, LVs Col8a1 mRNA was higher in the ppd2 groups than their virgin (P = 0.0006 for both genotypes) and ed18 (P < 0.0001 for both genotypes) groups. For wild types, the ppd28 group Col8a1 mRNA level was higher (P = 0.02) than the ed18 group, and the ppd28 group with Timp4 KO mice Col8a1 mRNA level was less than (P = 0.006) than the ppd2 group. Col1a1 and Col5a1 mRNA levels did not vary with reproductive status. When comparing the wild-type and Timp4 KO genotype mRNA levels by Sidak’s multiple comparison test for the same reproductive status day, Col3a1 mRNA was lower in Timp4 KO mice at ppd2 (P = 0.02) compared with wild types. There were no other collagen mRNA level differences.

As collagen proteins and their mRNAs do not necessarily mirror each other’s expression, we evaluated protein levels for collagens I, III, and VIII. Nonreducing gel conditions were used to evaluate high molecular weight forms of collagens I, III, and VIII with antibodies to COL1A1, COL3A1, and COL8A1 subunits (Fig. 5B). Collagens can appear as multiple bands under these conditions on Western blots due to the presence of individual subunit molecules or complexes (35). COL1A1 bands ranged from ∼100 to 160 kDa that likely included procollagen 1α1 and collagen 1α1 and its intermediate forms (37, 38). COL3A1 was observed as two bands ∼130–150 kDa, which is predicted to include the procollagen 3α1 chain (39). Col8a1 antibody detected ∼188 kDa band reflecting complexes and an ∼80 kDa monomeric form (40, 41). Higher molecular weight complexes (>200 kDa) were not detected. Western blot proteins were quantified relative to α-tubulin monomer (Fig. 5C). Relative protein levels were analyzed by two-way ANOVA as described for mRNA above. Simple main effects analyses showed reproductive status had significant effects on COL1A1 protein levels (P < 0.04), but not COL3A1 and COL8A1 protein levels. Simple main effects analyses of genotype showed significant effects on COL1A1 (P < 0.0001), COL3A1 (P = 0.001), and COL8A1 (P = 0.04) protein levels. There were no significant interactions of reproductive status and genotype for any of the collagen proteins examined.

Post hoc Tukey’s analyses were used to evaluate collagen protein subunit levels for differences between the reproductive states of the mice within each genotype and showed that COL1A1 protein in ppd28 wild types were higher than ed18 (P = 0.01) mice, but no other differences were present for collagen proteins. When comparing genotypes by Sidak’s analyses for each reproductive status day, COL1A1 protein was significantly higher in the Timp4 KO mouse LVs than wild types for virgin (P = 0.003), ed18 (P = 0.003), and ppd2 (P = 0.003) mice, but not ppd28 mice. COL3A1 protein was higher in virgin Timp4 KO mice than virgin wild types (P = 0.02) but did not differ on other days. COL8A1 levels did not vary between genotypes on any day.

LV Matrix Metalloproteinase mRNAs and MMP2/9 Proteins

The mRNA levels for Mmp2, -3, -9, and 15 (Fig. 6A) and Mmp13 and 14 (Supplemental Fig. S2) in LVs of mice on the day of euthanasia were measured. Two-way ANOVA was used to assess the effects of reproductive status and genotype and their interactions on Mmp mRNA levels. A significant simple main effect of reproductive status was detected for Mmp2 (P = 0.04), Mmp3 (P = 0.005), Mmp9 (P = 0.047), and Mmp15 (P = 0.002) but not for Mmp13 or Mmp14 mRNA. A significant simple main effect of genotype was observed for Mmp2 (P = 0.04), Mmp3 (P = 0.0001), and Mmp9 (P = 0.0002), but not Mmp13, Mmp14, or Mmp15 mRNA levels. Post hoc Tukey’s analyses within genotypes revealed that in wild-type mice, Mmp15 mRNA was lower in late pregnancy (ed18) compared with virgins (P = 0.01) but increased back to virgin levels in ppd2 and ppd28 mice (P = 0.02 and P = 0.002, compared with ed18 mice). In Timp4 KO mice, LV Mmp3 mRNA levels were increased in ppd2 mice compared with virgin mice (P = 0.003) and decreased back to virgin levels in ppd28 mice (P = 0.007 compared with ppd2). No other significant within-genotype differences were detected. When comparing genotypes for specific reproductive status day by Sidak’s post hoc test, Mmp3 mRNA levels were lower in Timp4 KO mice in the virgin (P = 0.04) and ppd28 (P = 0.007) groups compared with wild types. Mmp9 mRNA was lower in the late pregnancy ed18 (P = 0.02) group in Timp4 KO mice when compared with wild types. No differences were observed between genotypes for Mmp13 and Mmp14 mRNAs.

Figure 6.

Quantification of specific Mmp mRNAs, matrix metalloproteinase (MMP)2, and MMP9 proteins and activities in the left ventricles (LVs) of wild-type and Timp4 knockout (KO) mice in different reproductive states including virgin, late pregnant (embryonic day 18, ed18), and postpartum days 2 and 28 (ppd2 and ppd28, respectively). A: Mmp mRNAs were quantified by qPCR and standard curve method using for Rplp0 mRNA for normalization. B: reducing conditions and Western blot analysis were used to detect MMP2 and MMP9 proteins. Kilodaltons represent approximate sizes of bands on the gel. Western blot data were quantified and normalized for tubulin. Two different mice are shown per group, and each row represents images from the same gel, film, and exposure time. The last ppd28 column shown was not quantified because of the artifact. C: quantification of MMP2 and MMP9 protein from Western blots. D: zymogram of LV protein. Light bands indicated represent MMP2 and -9 activities. The large Coomassie-stained band (75 kDa) was used as a loading control. Dots on graphs represent individual mice and bars and lines indicate means ± SD. mRNA, protein level, and MMP2 activity data were ln-transformed and analyzed by two-way ANOVA to determine the effects of reproductive status and genotype on mRNA levels. Post hoc comparisons of different reproductive days within each genotype were evaluated by Tukey’s multiple comparison test. Post hoc comparisons of mRNA levels for each reproductive status day by genotype were performed using Sidak’s multiple comparison test. For mRNAs, n = 8 to 9 mice/group. For Western proteins, n = 3 to 4 mice/group. For zymograms, n = 4 mice/group. P < 0.05, ^asignificantly different compared with virgin, ^bdiffers from ed18, and ^cdiffers from ppd2. *P < 0.05, difference between genotypes for the same reproductive status day,

MMP2 and MMP9 protein levels were measured by Western blot normalized to tubulin protein (Fig. 6, B and C). Two-way ANOVA revealed no main effects of reproductive status or genotype, and no interaction between these independent variables for MMP2 protein levels. For MMP9 protein levels, two-way ANOVA indicated significant main effects of reproductive status (P = 0.047) and genotype (P = 0.01), but no significant interaction; however, no post hoc differences were detected for reproductive status within genotype or between genotypes. Finally, we evaluated MMP2 and MMP9 activity in LV by gelatin zymography (Fig. 6D). Two-way ANOVA revealed a main effect of genotype (P = 0.04) on MMP2 activity, but no main effect of reproductive status. There was a significant interaction (P = 0.008) of reproductive status and genotype on MMP2 activity. Post hoc analyses showed within genotype, Timp4 KO MMP2 activity was lower (P = 0.045) in ppd28 than virgin mice. There were no differences in MMP2 activity among the wild-type mouse groups. When genotype was assessed post hoc, virgin Timp4 KO mice had higher (P = 0.006) MMP2 activity than virgin wild-type mice. MMP9 levels were very low and not consistently detected above background and thus were not quantified.

DISCUSSION

TIMP4 has been shown to be altered in cardiac pathologies including mouse models of myocardial infarction, pressure overload, and volume overload. Our initial hypothesis was that TIMP4 deficient females would show reduced cardiac function with the volume overload of pregnancy and lactation, but this was not supported by the data. Our current data indicated that Timp4 KO female mice had altered ejection fraction at many time points and maintained stroke volume for most time points, and this was accompanied by alterations in ECM-related mRNAs and their proteins. We cannot rule out direct effects on myocytes, as one study with transformed mouse cardiomyocytes in culture indicated TIMP4 overexpression increased contractility when compared with TIMP4 depletion (42). Whether specific proteins are responsible for the cardiac adaptation we observed in female Timp4 KO mice will require further detailed study.

We previously reported that wild-type C57BL/6 mice had higher HW/TLs at postpartum days 1.5 and 7 compared with virgin mice (18), which was also found here for postpartum mice of both genotypes. Heart weights were increased in Timp4 KO mice for all groups except the late pregnancy group. Normalizing heart weight for tibia length revealed higher ratios for Timp4 KO mice, thus body weight differences may not completely explain heart weight differences. Although C57BL/6 background Timp4 KO female weights have not been previously reported, a study of 8-wk-old Timp4 KO males revealed increased body weights compared with wild types (43), a pattern which the females in our study appear to follow, with the exception of postpartum day 28 mice. Interestingly, in a 12-wk feeding study, Timp4 KO males had higher food intake, lower percent body fat, and higher lean body mass than wild types. In addition, Timp4 KO males had lower nighttime energy expenditure and lower basal metabolic rate (43). A potential indicator that Timp4 KO female mice may have metabolic differences from wild types is their shorter estrus phase of the estrous cycle, as estrus uses more energy than proestrus or diestrus (30). Whether similar differences in metabolism are present in female Timp4 KO mice remain to be determined.

Increases in LV internal diameters during late pregnancy in humans and mice have been previously reported (18, 44–46). In the current study, both genotypes showed increased LV internal diameters with mid or late pregnancy that continued through the first postpartum month for diastolic measurements. Maintenance of systolic changes in LV internal diameters varied by genotype with more consistent increases for wild types. The Timp4 KO mice displayed smaller LV internal diameters and resulting volumes that when compared with wild types were significantly lower during systole for the majority of time points. In contrast, diastolic internal diameters only differed between genotypes on ppd28, 1 wk after weaning of pups. Wild-type mice had baseline ejection fractions similar to those previously reported for C57BL/6 female mice (47–49). Although ejection fraction was increased in Timp4 KO mice compared with wild types for the majority of time points, it likely was still in a healthy range. A comprehensive review of ejection fraction for female mice has not been reported, however, a review of studies with normal C57BL/6 male mice report mean ejection fraction as 69 ± 8.2 (50). We found that in Timp4 KO mice, stroke volume was highest at ppd14 and was higher than wild types. Postpartum days 7 through 14 represent a period of rapid pup growth and high energetic demands on the mother both for lactation and thermoregulation (51), which could be factors in the ppd14 difference between genotypes. Our findings in C57BL/6 wild-type mice differ from our previous study (18), in which we failed to observe any increase in SV with pregnancy through postpartum day 7. Our current work has the advantage over our previous shorter study in that the current mice underwent serial echocardiography so that the same mice were followed until their euthanasia day. This protocol allowed for more early data points. In addition, all mice were bred from the same starting group in our animal facility, whereas the Parrott et al. (18) study used a combination of mice purchased as timed-pregnant animals and mice ordered on different occasions. In our current study, we used pregnancy and postpartum mice that had at least 4 fetuses (ed18) or pups (ppd2 and ppd28) to have more uniformity in the pregnancy and lactation load on the heart, which was not done in our previous study (18).

Timp1 mRNA was elevated at ppd2 compared with virgins for wild-type mice, which is similar to reports for ppd0 and ppd1.5 C57BL/6 mice (18, 19). Timp1 mRNA elevation at ppd2 was conserved in Timp4 KO mice. In rats, TIMP1 protein was elevated in late pregnancy but not on postpartum day 7 (7). Studies of Timp1 KO mice indicate that TIMP1 is needed for maintaining cardiac geometry (31), thus its transient expression may regulate remodeling in the peripartum window. Although there were no differences in Timp2 or 3 mRNA levels for the different reproductive time points evaluated, Timp2 mRNA at ppd28 was lower in the Timp4 KO females than wild types. This would be a time when the volume overload on the heart resulting from lactation would be diminished as pups were weaned at ppd21. Like the virgin females in our study, male Timp4 KO mice exhibited similar Timp1–3 mRNAs under basal (sham) conditions. In contrast, Timp4 KO males did increase Timp2 mRNA in response to pressure overload (24).

Col1a1, Col3a1, and Col8a1 mRNAs were examined as we had previously observed changes in their expressions in wild-type mice with pregnancy and/or early postpartum (18). We evaluated Col5a1 mRNA as an unregulated control. Unexpectedly, we did not observe changes in Col1a1 mRNA in either genotype. We did observe the expected increase in Col3a1 mRNA in the postpartum wild-type hearts but not in late pregnancy. Interestingly, there were no significant changes in Col3a1 mRNA in the Timp4 KO mice at the days examined, and Timp4 KO mice had lower Col3a1 mRNA levels on ppd2 and ppd28 compared with wild types. Consistent with our previous study of wild-type mice, Col8a1 mRNA peaked in the early postpartum period (ppd2) for both wild-type and Timp4 KO mice (18).

As collagen mRNA and protein levels do not always agree with each other (17), we also evaluated the proteins for COL1A1, COL3A1, and COL8A1 subunits by immunoblotting. Initial experiments with reduced conditions failed to detect higher molecular weight forms of these proteins and thus nonreducing gel conditions were used as described by Iannarone et al. (35). With the nonreducing (49) conditions, we were able to detect high-molecular-weight procollagen forms of COL1A1 and COL3A1 (38). COL1A1 protein was greater in Timp4 KO virgins, pregnant, and postpartum day 2 but not ppd28 mice. COL3A1 protein was also higher in virgin Timp4 KO LV tissue. These molecular forms of collagen should reflect new collagen synthesis (37, 38). One possible interpretation of these findings is that the Timp4 KO mouse females may have a higher turnover of LV collagen because of the lack of TIMP4 availability to inhibit collagenases, and thus synthesize more COL1A1 and COL3A1 to compensate as needed. A more detailed analysis of collagen turnover coupled with histological collagen analyses would be needed to support this idea. In rats, collagen I decreased while collagen III protein increased during pregnancy as evaluated by traditional Western blot, however protein sizes were not reported (17). In contrast, another rat study found no increase in collagen content in pregnant hearts when assessed by picrosirius red staining (12). Likewise, collagen differences were not found in control (sham) male Timp4 KO mouse hearts when assessed by picrosirius red staining (21).

MMP encoding-mRNAs, proteins, and activities levels are frequently altered in pathological cardiac models along with ECM remodeling, particularly MMP2 and MMP9 (52). We examined several LV Mmp mRNAs that had previously demonstrated significant changes or trends during the peripartum period in wild-type C57BL/6 mice (18). Umar et al. (44) reported elevated Mmp2 mRNA levels in LV of late pregnancy (ed19/20), whereas as our prior and current study did not detect differences at ed18/19 in wild types (18). In the current study, we observed higher Mmp2 mRNA in ppd2 compared with virgins only in the LVs of Timp4 KO mice, with MMP2 protein levels not affected, and lower MMP2 activity at ppd28. Virgin Timp4 KO mice exhibited higher LV MMP2 activity than wild-type virgins. Taken together, MMP2 differences in activity in virgins could in part explain the need for Timp4 KO virgins to make more new collagen protein. The loss of TIMP4 would theoretically increase MMP2 activity if other TIMPs do not fully compensate. TIMP4 is known to inhibit active MMP2 directly and also prevent MMP2 activation indirectly through inhibition of MMP14 (53, 54). There were no alterations of Mmp14 mRNA levels in females within or between genotype to suggest an effect on this end point. Furthermore, MMP14 activity did not differ in Timp4 KO male hearts under control conditions suggesting it is not affected (21).

Two studies reported Mmp3 mRNA elevated in C57BL/6 wild types during late pregnancy and the early postpartum compared with virgin (18, 19). In our current study, increased Mmp3 mRNA was not observed for wild types, but Timp4 KO mice did exhibit increased Mmp3 mRNA at the early postpartum time point. Mmp3 mRNA was also lower at the nonpregnant and nonlactating time points in Timp4 KO, virgin and ppd28, when compared with wild types. Similarly, to previous work with wild types, differences in Mmp9 mRNA were not detected (18), however, ed18 Timp4 KO mouse LV tissue had lower Mmp9 mRNA than wild types, an effect that did not affect MMP9 protein levels at the same time point.

Our zymography findings in virgin female mice are consistent with studies of control male wild-type and Timp4 KO mouse hearts that exhibited low levels of MMP2 and MMP9 activity (21), however, unlike in males which showed no differences between genotypes our results found higher MMP2 activity in Timp4 KO virgins suggesting a sex difference for the genotype.

Limitations

A limitation of our study is that the Timp4 KO mice are global knockouts and thus have lacked TIMP4 protein in all tissues throughout the development and could reflect adaptations to the TIMP4 deficiency. An inducible tissue-specific Timp4 knockout would allow better assessment of the role of TIMP4 in the adult heart. Another limitation is that we chose to look at cardiac changes starting at day 10 of pregnancy (midpregnancy), and it is possible that measurable functional and molecular cardiac changes may have occurred earlier in pregnancy and therefore have been missed. We chose day 10 of pregnancy as a starting point for echocardiography because we wanted to visually confirm live fetuses (with heart beats) in the uterus to confirm viable pregnancies. In a prior study of wild-type mice, measurement of 21 mRNAs encoding ECM proteins found only two (Col1a1 and Col3a1) significantly altered at day 12 of pregnancy compared with virgins, thus we chose to focus on late pregnancy (ed18) for molecular studies. A recent study of pregnancy, cardiac dynamics, and metabolism in wild-type FVB mice has demonstrated changes in numerous gene pathways on day 8 of pregnancy (55). FVB mice have slightly accelerated gestational development compared with C57BL/6 mice (56). In addition, a meta-analysis of human maternal heart geometry during normotensive pregnancies showed cardiac geometric increases during the second trimester (46), suggesting midpregnancy was a good starting point for echocardiographic studies.

One caution is that serial echocardiography required mice to undergo repeated exposure to anesthesia which could have impacted our overall findings. As wild-type mice underwent the same procedures as knockouts our comparisons between genotypes should still be valid. In addition, a lack of histological analyses of cardiac sections is also a limitation of our study. It would be informative to have data on cell size, ECM fiber orientation, and distribution. This would be most important if there was reduced cardiac function suggesting fibrosis, yet young adult female Timp4 KO mice showed increased ejection fractions and mostly normal stroke volumes.

Conclusions

Young adult female Timp4 KO mice during their first pregnancies and the first postpartum month exhibit LV functional differences from wild-type mice, yet evidence is lacking for peripartum cardiac myopathy or any adverse cardiac activity. Mice with global loss of Timp4 in our study did not exhibit reduced ejection fraction, rather ejection fraction was increased or not different from wild-type mice. Primarily, systolic LV internal diameters and volumes in the female Timp4 KO mice were affected. Our work also provides further evidence that the mouse maternal heart does not completely return to the prepregnant state within 1 wk after pups are weaned suggesting that either it takes longer to completely remodel or that the hearts do not fully remodel to the prepregnant state. As we did not include nonlactating postpartum mice we cannot differentiate between these choices at this time. A recent study with C57BL/6 mice has shown that nonlactating postpartum females had cardiac parameters similar to nulliparous animals by 10.5 wk postpartum (57), indicating postpartum female mice without lactation do return to baseline with time. In summary, Timp4 KO mice exhibited genotype-specific differences in collagen I and III proteins, Timp2, Mmp3 and Mmp9 mRNAs, and MMP2 activity levels with different reproductive states. Differences in extracellular matrix molecules in Timp4 KO mice likely contribute to an adaptive response of these mice to maintain adequate stroke volume even with demands of pregnancy and lactation.

DATA AVAILABILITY

Data will be made available upon reasonable request.

SUPPLEMENTAL DATA

Supplemental Table S1: https://doi.org/10.6084/m9.figshare.20444427.

Supplemental Fig. S1: https://doi.org/10.6084/m9.figshare.20444445.

Supplemental Fig. S2: https://doi.org/10.6084/m9.figshare.20444439.

GRANTS

This work was supported by a SC INBRE Developmental Research Project award funded by National Institute of General Medical Sciences Grant P20GM103499 (to H.A.L.).

DISCLAIMERS

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

A.A.T. and H.A.L. conceived and designed research; A.A.T., A.N.E., and H.A.L. performed experiments; A.A.T., A.N.E., E.C.G., F.G.S., and H.A.L. analyzed data; A.A.T., A.N.E., E.C.G., F.G.S., and H.A.L. interpreted results of experiments; H.A.L. prepared figures; A.A.T. and H.A.L. drafted manuscript; A.A.T., A.N.E., E.C.G., F.G.S., and H.A.L. edited and revised manuscript; A.A.T., A.N.E., E.C.G., F.G.S., and H.A.L. approved final version of manuscript.