Cold-Inducible RNA-Binding Protein but Not Its Antisense lncRNA Is a Direct Negative Regulator of Angiogenesis In Vitro and In Vivo via Regulation of the 14q32 angiomiRs—microRNA-329-3p and microRNA-495-3p

Inhibition of the 14q32 microRNAs, miR-329-3p and miR-495-3p, improves post-ischemic neovascularization. Cold-inducible RNA-binding protein (CIRBP) facilitates maturation of these microRNAs. We hypothesized that CIRBP deficiency improves post-ischemic angiogenesis via downregulation of 14q32 microRNA expression. We investigated these regulatory mechanisms both in vitro and in vivo. We induced hindlimb ischemia in Cirp−/− and C57Bl/6-J mice, monitored blood flow recovery with laser Doppler perfusion imaging, and assessed neovascularization via immunohistochemistry. Post-ischemic angiogenesis was enhanced in Cirp−/− mice by 34.3% with no effects on arteriogenesis. In vivo at day 7, miR-329-3p and miR-495-3p expression were downregulated in Cirp−/− mice by 40.6% and 36.2%. In HUVECs, CIRBP expression was upregulated under hypothermia, while miR-329-3p and miR-495-3p expression remained unaffected. siRNA-mediated CIRBP knockdown led to the downregulation of CIRBP-splice-variant-1 (CIRBP-SV1), CIRBP antisense long noncoding RNA (lncRNA-CIRBP-AS1), and miR-495-3p with no effects on the expression of CIRBP-SV2-4 or miR-329-3p. siRNA-mediated CIRBP knockdown improved HUVEC migration and tube formation. SiRNA-mediated lncRNA-CIRBP-AS1 knockdown had similar long-term effects. After short incubation times, however, only CIRBP knockdown affected angiogenesis, indicating that the effects of lncRNA-CIRBP-AS1 knockdown were secondary to CIRBP-SV1 downregulation. CIRBP is a negative regulator of angiogenesis in vitro and in vivo and acts, at least in part, through the regulation of miR-329-3p and miR-495-3p.


Introduction
Peripheral arterial disease (PAD) is caused by occlusions of the arterial vasculature in the lower limbs, mainly the femoral artery, resulting in deprivation of blood flow and, thus, of oxygen and nutrients to the lower extremities [1]. Current treatment options include angioplasty procedures with stent placement and bypass surgery [2]. However, many patients with advanced PAD are not or no longer eligible for these therapies [3,4]. Therefore, novel therapeutic approaches are still required. In patients with PAD, endogenous

Blood Flow Recovery and Neovascularization after Hindlimb Ischemia (HLI) Surgery in Cirp −/− Mice
To investigate the role of CIRBP in neovascularization in vivo, an HLI model was induced in Cirp knockout (Cirp −/− ) and wild-type (WT) C57BL/6 mice, and blood flow recovery was evaluated over time. No differences were observed in blood flow recovery between WT mice and Cirp −/− mice over 28 days ( Figure 1A,B). covery was evaluated over time. No differences were observed in blood flow recovery between WT mice and Cirp −/− mice over 28 days ( Figure 1A,B).
To visualize arteriogenesis, immunohistochemical staining for α-smooth muscle actin (α-SMA) in the adductor muscle was performed ( Figure 1C), and the number and diameter of α-SMA positive arterioles were quantified. Collateral density and the size of α-SMA positive arterioles of the ligated paws were similar between WT mice and Cirp −/− mice ( Figure 1D,E), which is in line with the results presented in Figure 1A.
To monitor the effects of Cirp deficiency on angiogenesis, capillary formation was evaluated in the soleus muscles at 28 days after induction of ischemia as visualized by CD31 staining ( Figure 1F). The CD31 + area represents the density of capillaries in the ischemic muscle. Compared to the unligated (right) paw, there was more capillary formation in the ligated (left) paw as shown in both WT mice (65.2% increase, p = 0.009) and Cirp −/− mice (100.5% increase, p = 0.003). Moreover, the CD31 positive area in the ligated paw of Cirp −/− mice was significantly higher than in WT mice (34.3% increase, p = 0.004), while there was no difference in the unligated paw between the two groups ( Figure 1G).  To visualize arteriogenesis, immunohistochemical staining for α-smooth muscle actin (α-SMA) in the adductor muscle was performed ( Figure 1C), and the number and diameter of α-SMA positive arterioles were quantified. Collateral density and the size of α-SMA positive arterioles of the ligated paws were similar between WT mice and Cirp −/− mice ( Figure 1D,E), which is in line with the results presented in Figure 1A.
To monitor the effects of Cirp deficiency on angiogenesis, capillary formation was evaluated in the soleus muscles at 28 days after induction of ischemia as visualized by CD31 staining ( Figure 1F). The CD31 + area represents the density of capillaries in the ischemic muscle. Compared to the unligated (right) paw, there was more capillary formation in the ligated (left) paw as shown in both WT mice (65.2% increase, p = 0.009) and Cirp −/− mice (100.5% increase, p = 0.003). Moreover, the CD31 positive area in the ligated paw of Cirp −/− mice was significantly higher than in WT mice (34.3% increase, p = 0.004), while there was no difference in the unligated paw between the two groups ( Figure 1G).

Ex Vivo Angiogenic Sprouting in Aorta Rings
To study the effects of Cirp deficiency on angiogenesis ex vivo, aorta ring assays were performed. Two different vascular endothelial growth factor (VEGF) concentrations (i.e., 10 and 30 ng/mL) were used to induce sprouting in either wild-type or Cirp deficient aorta rings. Aorta rings from Cirp −/− mice developed more sprouts compared to the WT control rings, both in rings incubated with 10 (55.4% increase in sprouts) or 30 ng/mL VEGF (79.6% increase in sprouts) ( Figure 2). Quantification of the sprouting demonstrated that for both VEGF concentrations, this difference was statistically significant (10 ng/mL VEGF, p = 0.004; 30 ng/mL VEGF, p < 0.001).

Ex Vivo Angiogenic Sprouting in Aorta Rings
To study the effects of Cirp deficiency on angiogenesis ex vivo, aorta ring assays were performed. Two different vascular endothelial growth factor (VEGF) concentrations (i.e., 10 and 30 ng/ml) were used to induce sprouting in either wild-type or Cirp deficient aorta rings. Aorta rings from Cirp −/− mice developed more sprouts compared to the WT control rings, both in rings incubated with 10 (55.4% increase in sprouts) or 30 ng/mL VEGF (79.6% increase in sprouts) ( Figure 2). Quantification of the sprouting demonstrated that for both VEGF concentrations, this difference was statistically significant (10 ng/mL VEGF, p = 0.004; 30 ng/mL VEGF, p < 0.001).

Figure 2.
Ex vivo angiogenesis: representative images and quantification of neovessel sprouts from 7 day collagen-embedded aorta rings (40 rings per condition from 3 WT and 3 Cirp −/− mice), treated without or with vascular endothelial growth factor (VEGF) at 10 and 30 ng/mL. Data are presented as the mean ± SEM; ** p < 0.01; *** p < 0.001 by independent sample Student's t-tests.

MicroRNA Expression in Cirp −/− Mice
MiR-329-3p and miR-495-3p are both highly associated with angiogenesis and regulated by CIRBP as has been shown previously [21,25]. To study if and how miR-329-3p and -495-3p expression changes in Cirp −/− mice after induction of hindlimb ischemia, the HLI surgery was repeated, and the mice were sacrificed at 1 day, 3 days, and 7 days after surgery, followed by RNA isolation from the soleus muscle from both the left and right paws. These early timepoints were chosen as ischemia had not resolved yet and angiogenesis was still ongoing. Although the expression of miR-329-3p and miR-495-3p increased in the ischemic soleus muscles of both mouse strains at 7 days, their expression increased significantly less in the Cirp −/− mice (40.6% less, p = 0.028 for miR-329-3p; 36.2% less, p < 0.001 for miR-495-3p; Figure 3A,B). This reduced upregulation became even more clear when looking at the ratio of expression in the ligated over the non-ligated paws, where the ratio only increased in the WT mice and not in the Cirp −/− mice (p = 0.027 for miR-329-3p; p = 0.0736 (trend) for miR-495-3p; Figure 3C,D). These differences were most evident at day 7. Ex vivo angiogenesis: representative images and quantification of neovessel sprouts from 7 day collagenembedded aorta rings (40 rings per condition from 3 WT and 3 Cirp −/− mice), treated without or with vascular endothelial growth factor (VEGF) at 10 and 30 ng/mL. Data are presented as the mean ± SEM; ** p < 0.01; *** p < 0.001 by independent sample Student's t-tests.

MicroRNA Expression in Cirp −/− Mice
MiR-329-3p and miR-495-3p are both highly associated with angiogenesis and regulated by CIRBP as has been shown previously [21,25]. To study if and how miR-329-3p and -495-3p expression changes in Cirp −/− mice after induction of hindlimb ischemia, the HLI surgery was repeated, and the mice were sacrificed at 1 day, 3 days, and 7 days after surgery, followed by RNA isolation from the soleus muscle from both the left and right paws. These early timepoints were chosen as ischemia had not resolved yet and angiogenesis was still ongoing. Although the expression of miR-329-3p and miR-495-3p increased in the ischemic soleus muscles of both mouse strains at 7 days, their expression increased significantly less in the Cirp −/− mice (40.6% less, p = 0.028 for miR-329-3p; 36.2% less, p < 0.001 for miR-495-3p; Figure 3A,B). This reduced upregulation became even more clear when looking at the ratio of expression in the ligated over the non-ligated paws, where the ratio only increased in the WT mice and not in the Cirp −/− mice (p = 0.027 for miR-329-3p; p = 0.0736 (trend) for miR-495-3p; Figure 3C,D). These differences were most evident at day 7.  3 for the untreated group, n = 4 for the HLI-1d group, n = 3 for the HLI-3d group, and n = 3 for the HLI-7d group) and Cirp −/− mice (n = 3 for the untreated group, n = 4 for the HLI-1d group, n = 3 for the HLI-3d group, and n = 3 for the HLI-7d group). Expression was normalized to U6. (C,D) The relative expression of mature miRNA-329 and -495 from mice after HLI surgery at different timepoints, calculated as the ratio of the left (ischemic) over the right (non-ischemic) paw. Data are presented as the mean ± SEM; * p < 0.05; *** p < 0.001 by two-way ANOVA.

Total CIRBP, CIRBP Splice Variants, and lncRNA-CIRBP-AS1 Expression in Hypothermia
The gene structure of the human CIRBP is shown in Figure 4A. Previous studies showed that CIRBP expression increased under cellular stress conditions including mild hypothermia [9]. To demonstrate that hypothermia upregulates CIRBP-targeted miRNAs and also the different CIRBP splice variants and antisense lncRNAs in HUVECs, HUVECs were subjected to mild hypothermia (32 °C) either for 24 or 48 hours. Total CIRBP expression increased after both 24 and 48 hours compared to the normothermic condition (37 °C) by 205.5% (p = 0.029; Figure 4B) and 68.7% (p = 0.028; Figure 4F), respectively. CIRBP-SV1 expression increased by 83.1% at 24 hours (p = 0.12) and by 67.1% at 48 hours (p = 0.14) under hypothermia, although this upregulation was not statistically significant. However, the remaining three splice variants were not altered consistently over time under hypothermia, neither were significant changes in microRNA expression observed over time ( Figure 4C,D,G,H), indicating that upregulation of CIRBP did not lead to additional mi-croRNA processing.
lncRNA-CIRBP-AS1 did show a trend towards a 61.7% increase in expression under hypothermia after 24 hours (p = 0.096; Figure 4E) and was even further upregulated by 130.4% after 48 hours (p = 0.097) compared to normothermia ( Figure 4I). for the HLI-1d group, n = 3 for the HLI-3d group, and n = 3 for the HLI-7d group) and Cirp −/− mice (n = 3 for the untreated group, n = 4 for the HLI-1d group, n = 3 for the HLI-3d group, and n = 3 for the HLI-7d group). Expression was normalized to U6. (C,D) The relative expression of mature miRNA-329 and -495 from mice after HLI surgery at different timepoints, calculated as the ratio of the left (ischemic) over the right (non-ischemic) paw. Data are presented as the mean ± SEM; * p < 0.05; *** p < 0.001 by two-way ANOVA.

Total CIRBP, CIRBP Splice Variants, and lncRNA-CIRBP-AS1 Expression in Hypothermia
The gene structure of the human CIRBP is shown in Figure 4A. Previous studies showed that CIRBP expression increased under cellular stress conditions including mild hypothermia [9]. To demonstrate that hypothermia upregulates CIRBP-targeted miRNAs and also the different CIRBP splice variants and antisense lncRNAs in HUVECs, HUVECs were subjected to mild hypothermia (32 • C) either for 24 or 48 h. Total CIRBP expression increased after both 24 and 48 h compared to the normothermic condition (37 • C) by 205.5% (p = 0.029; Figure 4B) and 68.7% (p = 0.028; Figure 4F), respectively. CIRBP-SV1 expression increased by 83.1% at 24 h (p = 0.12) and by 67.1% at 48 h (p = 0.14) under hypothermia, although this upregulation was not statistically significant. However, the remaining three splice variants were not altered consistently over time under hypothermia, neither were significant changes in microRNA expression observed over time ( Figure 4C,D,G,H), indicating that upregulation of CIRBP did not lead to additional microRNA processing. (C,D,G,H) primary microRNA (pri-miR), precursor microRNA (pre-miR), and mature microRNA expression levels of miR-329 and miR-495 were measured and normalized to U6; (E,I) CIRBP gene antisense long non-coding RNA (lncRNA-CIRBP-AS1) expression level was measured and normalized to GAPDH. Data are shown as the relative expression compared to the normothermic group, presented as the mean ± SEM; * p < 0.05 by one-sample t-tests (one-tail).

Angiogenesis Assays and RNA Expression after CIRBP Knockdown
To investigate the role of CIRBP in vitro, HUVECs were treated with small interfering RNA (siRNA) targeted to CIRBP, and the angiogenic potential was assessed using both scratch-wound healing assays and tube-formation assays. Scratch-wound healing showed an increased (2.8-fold, p = 0.004) angiogenic potential after siRNA--CIRBP transfection ( Figure 5A,B). To exclude any effects of potential CIRBP knockdown-induced cell proliferation, the expression of proliferating cell nuclear antigen (PCNA) mRNA was assessed and, indeed, no differences were observed compared to the negative control siRNA (Figure 5C). In addition, tube-formation assays also showed improvements in the number of segments (97.2%, p = 0.046), branches (65.4%, p = 0.016), and total length (27.7%, p = 0.014) after CIRBP knockdown ( Figure 5H,I). lncRNA-CIRBP-AS1 did show a trend towards a 61.7% increase in expression under hypothermia after 24 h (p = 0.096; Figure 4E) and was even further upregulated by 130.4% after 48 h (p = 0.097) compared to normothermia ( Figure 4I).

Angiogenesis Assays and RNA Expression after CIRBP Knockdown
To investigate the role of CIRBP in vitro, HUVECs were treated with small interfering RNA (siRNA) targeted to CIRBP, and the angiogenic potential was assessed using both scratch-wound healing assays and tube-formation assays. Scratch-wound healing showed an increased (2.8-fold, p = 0.004) angiogenic potential after siRNA-CIRBP transfection ( Figure 5A,B). To exclude any effects of potential CIRBP knockdown-induced cell proliferation, the expression of proliferating cell nuclear antigen (PCNA) mRNA was assessed and, indeed, no differences were observed compared to the negative control siRNA ( Figure 5C). In addition, tube-formation assays also showed improvements in the number of segments  After the scratch assays, cells were collected for RNA isolation to assess the expression of total CIRBP, CIRBP-SVs, and lncRNA-CIRBP-AS1 as well as the two microRNAs and their precursors. As shown in Figure 4A, the siRNA-CIRBP was designed to target all splice variants. Total CIRBP levels were indeed knocked down by 87.3% (p = 0.002, Figure 5D). However, when looking at the individual splice variants, only CIRBP-SV1 was knocked down (99.0% decrease, p = 0.002), whereas the other splice variants remained unaffected ( Figure 5D). Furthermore, a trend towards decreased lncRNA-CIRBP-AS1 expression (39.5% decrease, p = 0.078) was also observed when CIRBP was knocked down ( Figure 5G). Although upregulation of CIRBP expression through hypothermia did not increase microRNA expression, a significant decrease (71.2%; p = 0.02; Figure 5F) was observed in the expression of miR-495-3p under CIRBP knockdown. Surprisingly, miR-329-3p was not significantly downregulated in response to CIRBP knockdown, as it was in Cirp-deficient mice. We did observe a trend towards accumulation of the primary microRNA, pri-miR-329-1, which is in correspondence with CIRBP's effects on microRNA biogenesis [25].

Scratch-Wound Healing in CIRBP Knockdown or lncRNA-CIRBP-AS1 Knockdown HUVECs after 4 Hours of siRNA Treatment
As knockdown of both CIRBP and its antisense had similar effects on microRNA expression and angiogenesis, as well as on each other, it cannot be concluded which of the two was the main effector. Therefore, the scratch-wound healing experiments described above were repeated using a much shorter transfection time of 4 h. We confirmed that a shorter transfection time would allow for direct effects on gene expression but not yet for indirect effects (Supplemental Materials Figure S1).

(I) Total
CIRBP expression and its splice variant=s after the scratch assay, normalized to GAPDH. (J-L) After the scratch assay, the pri-miRNA, pre-miRNA, and mature microRNA expression levels of miR-329 and miR-495 were measured and normalized to U6; lncRNA-CIRBP-AS1 were measured and normalized to GAPDH. Data show the percentage compared to siRNA and the negative control group. Data are presented as the mean ± SEM; * p < 0.05; ** p < 0.01; one-sample t-tests (two-tail).

Discussion
In this study, we confirmed the previous findings that deficiency in CIRBP leads to enhanced angiogenesis in a murine hindlimb ischemia model. We demonstrated that during ischemia, CIRBP contributed to the regulation of the angiomiRs, miR-329-3p and miR-495-3p, both in vivo and in vitro. Importantly, we demonstrated increased angiogenic activity upon CIRBP knockdown in human endothelial cells. Furthermore, we investigated the complex CIRBP gene sequence in humans more closely and looked in detail at the regulation of different CIRBP splice variants as well as a regulatory antisense lncRNA, lncRNA-CIRBP-AS1. The hindlimb ischemia model is a classical method to investigate post-ischemic neovascularization, commonly employed as a model for PAD in humans [26]. Even though Cirp −/− mice display enhanced angiogenesis after induction of ischemia, faster blood flow recovery in Cirp −/− mice compared to wild-type C57/BL6 mice was not observed. Most likely, this can be attributed to the fact that during hindlimb ischemia and PAD, blood flow recovery depends more strongly on arteriogenesis than on angiogenesis [27]. Nonetheless, enhanced angiogenesis can still be highly beneficial for PAD patients, as the available blood is distributed better throughout the affected tissues and as enhanced angiogenesis can prevent non-healing wounds and ulcers, which are a common and serious complication in PAD patients [28]. Indeed, it was shown in Cirp −/− mice that the inflammatory response in tissue wound healing was faster than in wild-type mice. More CD31 expression, as a marker of endothelial cells and, thus, angiogenesis, was observed in the wounds of Cirp −/− mice [29], although the exact mechanism of action for CIRBP in wound healing was not elucidated. The pro-inflammatory function of CIRBP under stress conditions has already been reported by Qiang et al. [30]. Furthermore, CIRBP was reported to influence changes in leukocyte recruitment and macrophage polarization in direction to regenerative M2-like macrophages, thus regulating angiogenesis and tissue regeneration [5]. Moreover, the authors reported that CIRBP binds TLR4, MD2, and the TLR4/MD2 complex, which are known to stimulate neovascularization and can be found on both macrophages and neutrophils [31]. However, CIRBP acts through other angiogenesis-related pathways as well and, furthermore, the CIRBP gene in humans has a more complex structure than in mice, as it has several splice variants that result in different proteins as well as an antisense lncRNA. Their role has not been investigated in angiogenesis yet.
In previous studies, our group showed that CIRBP regulates the processing and expression of microRNAs, miR-329-3p and miR-495-3p [25], which both play an important role in angiogenesis [21]. We expected CIRBP to affect angiogenesis by regulating the expression of miR-329-3p and miR-495-3p. Importantly, we also reported enhanced macrophage attraction and influx into the vessel wall upon miR-495-3p inhibition [22], which could help explain the effects observed in the studies by Kübler et al. described above. In the current study, we showed that CIRBP was regulated under hypothermic conditions as was reported previously [7,[9][10][11][12]. The novelty of this study, however, is that we showed this in human vascular endothelial cells that were subjected to hypothermic conditions, which frequently occurs in PAD patients [1]. The expression of the 14q32 microRNAs, miR-329-3p and miR-495-3p, did not increase under hypothermia in HUVECs, however, even though the expression of their reported post-transcriptional regulator CIRBP increased. In contrast, silencing of CIRBP, using an siRNA, did lead to a decrease in miR-495-3p expression. Likely, the processing rate of 14q32 microRNAs was already at an optimum level under normothermic conditions, which can explain that an increase in CIRBP did not induce more processing of precursor microRNAs. The dramatic decrease in CIRBP expression following siRNA treatment, on the other hand, did result in insufficient processing of miR-495-3p. Surprisingly, expression of miR-329-3p was still unaffected in vitro in human cells, even though we did observe a trend towards accumulation of the primary microRNA, pri-miR-329-1, which would correspond with the inhibition of microRNA biogenesis. In vivo in mice, on the other hand, we could clearly see that ischemia-induced upregulation of both miR-329-3p and miR-495-3p was blocked quite efficiently in Cirp −/− mice.
When we speculate what these findings could mean for human PAD, CIRBP expression would likely be increased, as patients suffer from cold extremities [1]. We found that hypothermia caused CIRBP expression upregulation. Such an increase in CIRBP, likely accompanied by continuously high miR-329-3p and possible miR-495-3p expression, would inhibit efficient angiogenesis and wound healing in PAD patients, with, for example, nonhealing ulcers. Meanwhile, others have reported that warm temperatures (42 • C) can decrease CIRBP expression in male germ cells [32]. Therefore, one could imagine that hyperthermia leads to decreased CIRBP expression in the leg, followed by subsequent increases in angiogenesis. Thus far, heat therapy was shown to be of potential benefit to PAD patients by enhancing leg blood flow and improving muscle function [33,34].
We further assessed the potential pro-angiogenic features of CIRBP knockdown in human primary endothelial cells. Indeed, cell-migration assays as well as tube-formation assays in HUVECs showed an increase in angiogenic potential following CIRBP downregulation. These findings are in line with those found in vivo. Looking at human cells, however, also allowed us to look into changes in the CIRBP splice variants (CIRBP-SVs) and its antisense long noncoding RNA (lncRNA-CIRBP-AS1). CIRBP has four splice variants that alter the coding sequence. Of these four variants, CIRBP-SV1 specifically showed a trend towards upregulation after 24 and 48 h of hypothermia. When we knocked down CIRBP using an siRNA, we again observed specific knockdown of CIRBP-SV1. This was unexpected, as the binding site of the siRNA was predicted to target an mRNA sequence that is present in all four splice variants. Although we cannot explain the siRNA's preference for CIRBP-SV1, we can conclude that the majority of CIRBP's effects on angiogenesis were elicited through CIRBP-SV1.
The antisense strand of the human CIRBP gene also encodes a long noncoding RNA, lncRNA-CIRBP-AS1, the function of which has not yet been elucidated. We observed a trend towards upregulation of lncRNA-CIRBP-AS1 under hypothermic conditions, although the response to hypothermia was slower than that of CIRBP itself. After CIRBP knockdown, lncRNA-CIRBP-AS1 expression decreased, and after lncRNA-CIRBP-AS1 knockdown, CIRBP expression decreased; inhibition of lncRNA-CIRBP-AS1 with an siRNA resulted in simultaneous downregulation of total CIRBP and of CIRBP-SV1, and CIRBP-SV3 in particular. Furthermore, lncRNA-CIRBP-AS1 knockdown resulted in a decreased expression of both mature miR-329-3p and miR-495-3p. More importantly, silencing of lncRNA-CIRBP-AS1, like CIRBP itself, resulted in increased angiogenesis in HUVECs.
Likely, a positive feedback loop supports the transcription of the CIRBP locus, where CIRBP and lncRNA-CIRBP-AS1 induce each other's expression. In order to elucidate which of the two is the main effector in the enhanced angiogenic potential of HUVECs, we separated their mutual effects by shortening the siRNA transfection duration, and we found that CIRBP knockdown still had the same effect on angiogenesis and regulation of microRNA expression, while lncRNA-CIRBP-AS1 knockdown could no longer improve angiogenesis, even though miR-329-3p expression was already reduced significantly. Combining these findings, we conclude that the effects of lncRNA-CIRBP-AS1 downregulation on angiogenesis were most likely indirect. We can also conclude, however, that lncRNA-CIRBP-AS1 can directly impact CIRBP expression and, likely, also miR-329-3p expression. The mechanisms behind this regulation remain to be determined.
In conclusion, our findings confirm the previously reported increase in post-ischemic angiogenesis in Cirp −/− mice. We showed that CIRBP directly regulated the angiomiRs microRNAs, miR-329-3p and miR-495-3p, in vivo. In addition, we validated these findings in human primary endothelial cells for the first time. Furthermore, we showed that CIRBP-SV1 was the splice variant that predominantly regulates CIRBP's effects on both microRNA expression and angiogenesis. Finally, the lncRNA-CIRBP-AS1 can also impact angiogenesis, but these effects are likely caused by directing changes in CIRBP-SV1 and, subsequently, miR-329-3p and miR-495-3p.

Animal Experiments
All animal experiments were approved by the Committee on Animal Welfare of the Leiden University Medical Center (Leiden, The Netherlands) and were performed in accordance with the Directive 2010/63/EU of the European Parliament and Dutch government guidelines. Cirp −/− embryos on a C57BL6/J background were kindly provided by Jun Fujita's lab (Kyoto University, Kyoto, Japan) [35]. WT C57BL/6 mice (n = 10, male, aged 8-10 weeks) and Cirp −/− mice (n = 11, male, aged 8-10 weeks) were bred in the LUMC's in-house breeding facility and had free access to water and regular chow.

HLI Model
Before surgery, mice were anesthetized via an intraperitoneal injection of midazolam (5 mg/kg; Roche Diagnostics, Almere, The Netherlands), medetomidine (0.5 mg/kg; Orion, Espoo, Finland), and fentanyl (0.05 mg/kg; Janssen Pharmaceuticals, Beerse, Belgium). Unilateral HLI was induced by double ligation of the left femoral artery, proximal to the superficial epigastric artery and proximal to the bifurcation of the popliteal and saphenous artery. After surgery, mice were given a subcutaneous injection of flumazenil (0.5 mg/kg, Fresenius Kabi, Utrecht, The Netherlands) and atipamezol (2.5 mg/kg, Orion) to antagonize anesthesia. Buprenorphine (0.1 mg/kg, MSD Animal Health, Boxmeer, The Netherlands) was given after surgery for pain relief [36].

Laser Doppler Perfusion Measurements
Blood flow recovery to the paw was measured over time using laser Doppler perfusion imaging (LDPI) (Moor Instruments, Axminster, United Kingdom) at day 0 (before and after ligation), 3, 7, 10, 14, 21, and 28. Before measurements, mice were anesthetized with an intraperitoneal injection of midazolam (5 mg/kg, Roche Diagnostics) and medetomidine (0.5 mg/kg, Orion). Mice were placed in a double-glassed pot that was perfused with water at 37 • C for 5 min prior to each measurement. LDPI measurements in the ligated paw were normalized to measurements of the unligated paw as an internal control. After LDPI, anesthesia was antagonized by subcutaneous injection of flumazenil (0.5 mg/kg) and atipamezole (2.5 mg/kg). At day 28, after the last LDPI measurement, mice were injected with fentanyl (0.05 mg/kg) and sacrificed via retro-orbital bleeding. The proximal half of the adductor muscle was harvested and fixed in 4% formaldehyde. The distal half of the adductor muscle and the soleus muscle were harvested and snap-frozen on dry-ice.

Immunohistochemical Staining
Adductor muscles were embedded in paraffin, and 5 µm thick sections were cut for histological analysis. Smooth muscle cells were stained with primary antibody mouse anti-mouse α-SMA (Dako, 1:1000). Rabbit anti-mouse HRP (Dako, 1:300) was used as the secondary antibody. Slides were scanned with the Pannoramic MIDI digital slide scanner (3DHistech). The number and lumen diameter of α-SMA positive vessels were analyzed by Pannoramic viewer software (3DHistech, version: 2.3) with 20× magnification (3 sections per limb per mouse). The smallest diameter of vessel was measured in the picture as described previously [37].

Isolation of Venous Endothelial Cells (HUVECs)
Primary human vascular cells were isolated as described earlier by Welten et al. [21]. In brief, for HUVEC isolation, the vein was inserted with a cannula and flushed with sterile PBS. The vessel was infused with 0.075% collagenase type II (Worthington) and incubated at 37 • C for 20 min. The collagenase solution was collected, and the vessel was flushed with PBS in order to collect all detached endothelial cells. The cell suspension was centrifuged at 400× g for 5 min, and the pellet was resuspended in HUVEC culture medium (EBM-2 Basal Medium (CC-3156) and EGMTM-2 SingleQuots TM Supplements (CC-4176), Lonza, Walkersville, MD, USA). HUVECs were cultured in plates coated with 1% fibronectin from bovine plasma (Sigma, Amsterdam, The Netherlands).

Hypothermic HUVEC Cell Culture
Primary HUVECs were seeded in 12-well plates coated with 1% fibronectin at 100,000 cells per well in culture medium. After overnight incubation at 37 • C, cells were washed with PBS and new media were applied before putting the plates in the right incubator: the normothermic incubator at 37 • C and the hypothermic incubator at 32 • C, both humidified at 5% CO 2 and 20% O 2 . After 24 or 48 h, cells were washed with PBS, and 0.5 mL TRIzol/well was added for RNA isolation. Each single condition was performed in triplicate, and the hypothermia experiment was performed three independent times.

CIRBP and lncRNA-CIRBP-AS1 Knockdown with siRNA Transfection In Vitro
Primary HUVECs were seeded in 12-well plates coated with 1% fibronectin at 100,000 cells per well in culture medium. After 24 h, cells were washed with PBS, and each well was incubated with 900 µL Opti-MEM medium with 10% NBSCi and 1% penicillin/streptomycin and, after 10 min of incubation, 100 µL of transfection medium (94 µL Opti-MEM with 3 µL of Lipofectamine RNAiMax (Life Technologies, Bleiswijk, The Netherlands) and 3 µL of siRNA) was added. The final siRNA concentration per well was 30 nM. The siRNAs used were siRNA-CIRBP (sasi-172352), siRNA-lncRNA-CIRBP-AS1 (sasi-208901), and siRNA negative control (Mission Universal Negative Control #1) (all Sigma-Aldrich, Amsterdam, The Netherlands). After addition of transfection agents, cells were put in the incubator at 37 • C for the required time.

Migration Assay-Scratch Wound Healing
Primary HUVECs were seeded in 12-well plates coated with 1% fibronectin at 150,000 cells per well in HUVEC culture medium. After 24 h, the media were replaced with transfection medium as previously described. Then, transfection was conducted, and a scratch wound was performed across the diameter of each well using a p200 pipette tip. Next, cells were washed with PBS and fresh starving medium, and EBM-2 (Lonza) containing only 0.2% FBS and 1% gentamicin amphotericin of the provided BulletKit was added. In order to monitor scratch-wound closure, live phase-contrast microscopy (Axiovert 40C, Carl Zeiss, Oberkohen, Germany) was used for taking pictures at 0 and 18 h after introducing the scratch wound. In addition, a live cell microscope (Leica AF6000, Leica Microsystems, Tokyo, Japan) was used for taking picture every 4 h after scratch until 20 h in the timeline experiment. Pictures were taken in the same location at two positions in each well. Where necessary, pictures were contrast-enhanced using Microsoft PowerPoint. Scratch size was calculated using the wound healing tool macro for ImageJ. Finally, cells were washed with PBS, and 0.5 mL TRIzol/well was added for RNA isolation. Each single scratch assay condition was performed in triplicate, and the scratch-wound healing assay was performed three independent times.

Tube-Formation Assay
Tube-formation assay was performed using HUVECs at passage three. At confluence, cells were transfected as described above with Lipofectamine RNAiMax and siRNA-CIRBP, lncRNA-CIRBP-AS1, or siRNA negative control. After 24 h, cells were counted and seeded on solidified Geltrex TM (ref: A14132-02, Gibco) in a 96-well plate at 15,000 cells per well. Photos were taken using live phase-contrast microscopy at 12 h after seeding and quantified using the ImageJ Angiogenesis Analyzer. Each single tube-formation assay was performed in 6 wells per condition, and the tube-formation assay was performed three independent times.

Aorta Ring Assay
Thoracic aortas were isolated from Cirp −/− and wild-type mice, aged 4-5 weeks, after exsanguination under anesthesia via an intraperitoneal injection of midazolam (5 mg/kg; Roche Diagnostics), medetomidine (0.5 mg/kg; Orion), and fentanyl (0.05 mg/kg; Janssen Pharmaceuticals). Vessels were washed with Opti-MEM medium containing 1% penicillin/streptomycin and were cut transversely. Consequently, aorta rings were obtained 0.5-1 mm in width and incubated in Opti-MEM medium with 1% penicillin/streptomycin overnight at 37 • C. Collagen (type I, Merck Millipore, Darmstadt, Germany) was diluted to a concentration of 1 mg/mL with DMEM and 1% penicillin/streptomycin, and the pH was adjusted with 5 N NaOH to 7. Aortic rings were placed in 96-well plates coated with 75 µL collagen matrix as described previously [39]. After 1 h, 150 µL Opti-MEM supplemented with 2.5% FBS (PAA Laboratories, Pasching, Austria), penicillin-streptomycin (PAA Laboratories), and with or without vascular endothelial growth factor (in-house production and purification) of 10 or 30 ng/mL were added to the designated wells. Media were refreshed on days 3 and 5. Microvessel outgrowth was quantified after 7 days on photographs taken by live phase-contrast microscopy (Axiovert 40C, Carl Zeiss). The counting of microvessels started from a specific point on the ring, and each microvessel emerging from the ring was counted as a sprout, and individual branches arising from each microvessel counted as a separate sprout, working around the ring clockwise.

RNA Isolation
RNA isolation of cultured cells or tissue was performed by standard TRIzol-chloroform extraction according to the manufacturer's instructions (Thermo Fisher Scientific, Wilmington, DE, USA). RNA concentrations were measured using the Nanodrop TM 1000 Spectrophotometer (Thermo Fisher Scientific).

MicroRNA Quantification
For microRNA quantification of miR-329-3p and miR-495-3p, RNA was reversed transcribed using the Taqman TM MicroRNA Reverse-Transcription Kit (Thermo Fisher Scientific) and, subsequently, quantified using microRNA-specific Taqman TM qPCR kits (Thermo Fisher Scientific) on the VIIa7 (Thermo Fisher Scientific). MicroRNA expression was normalized against U6 small nuclear RNA.

Statistical Analyses
Data are presented as the mean ± SEM. Indicated differences had the following levels of significance: * p <0.05; ** p < 0.01; *** p < 0.001. All tests were performed with a significance level of α = 0.05.
One-sample t-tests were performed to test differences between treated groups that were expressed relative to the negative control treatment, which was set to 100%. Onesample t-tests (two tail) were used in the knockdown experiments, scratch assay, and tubeformation assay. One-sample t-tests (one tail) were used in the hypothermia experiment.
Differences in scratch wound healing and PCNA levels between groups were assessed using independent sample Student's t-tests.
Two-way ANOVA tests were performed to detect statistically significant differences among multiple groups. These tests were used to compare miRNA-329-3p and -495-3p expression levels of the soleus of mice subjected to HLI surgery at different timepoints.

Author Contributions:
All authors contributed to the study conception and design. Material preparation, and data collection and analysis were performed by E.A.C.G., L.Z. and M.R.d.V. The first draft of the manuscript was written by E.A.C.G. and L.Z. and revised by M.R.d.V., P.H.A.Q., J.W.J., and A.Y.N. All authors commented on previous versions of the manuscript. All authors have read and approved the final manuscript.