Preclinical evidence of remote ischemic conditioning in ischemic stroke, a metanalysis update

Remote ischemic conditioning (RIC) is a promising therapeutic approach for ischemic stroke patients. It has been proven that RIC reduces infarct size and improves functional outcomes. RIC can be applied either before ischemia (pre-conditioning; RIPreC), during ischemia (per-conditioning; RIPerC) or after ischemia (post-conditioning; RIPostC). Our aim was to systematically determine the efficacy of RIC in reducing infarct volumes and define the cellular pathways involved in preclinical animal models of ischemic stroke. A systematic search in three databases yielded 50 peer-review articles. Data were analyzed using random effects models and results expressed as percentage of reduction in infarct size (95% CI). A meta-regression was also performed to evaluate the effects of covariates on the pooled effect-size. 95.3% of analyzed experiments were carried out in rodents. Thirty-nine out of the 64 experiments studied RIPostC (61%), sixteen examined RIPreC (25%) and nine tested RIPerC (14%). In all studies, RIC was shown to reduce infarct volume (− 38.36%; CI − 42.09 to − 34.62%) when compared to controls. There was a significant interaction caused by species. Short cycles in mice significantly reduces infarct volume while in rats the opposite occurs. RIPreC was shown to be the most effective strategy in mice. The present meta-analysis suggests that RIC is more efficient in transient ischemia, using a smaller number of RIC cycles, applying larger length of limb occlusion, and employing barbiturates anesthetics. There is a preclinical evidence for RIC, it is safe and effective. However, the exact cellular pathways and underlying mechanisms are still not fully determined, and its definition will be crucial for the understanding of RIC mechanism of action.

Acute ischemic stroke (AIS) is the world's second leading cause of mortality and the major cause of disability in adults worldwide 1 . The main revascularization therapies for AIS are thrombolysis with recombinant tissueplasminogen activator (tPA) and endovascular thrombectomy. Unfortunately, many patients cannot benefit from those therapies due to mainly narrow therapeutic window and they can also induce ischemia-reperfusion injury (IRI) 2,3 . The development of novel therapeutic strategies is needed to extend therapeutic windows and to mitigate further brain injury.
Neuroprotective therapies have a great potential to not only increase the benefits of available reperfusion therapies but also to provide an advisable medical procedure for AIS patients who are not eligible for current treatments 4 . But translation of strategies targeting neuroprotection to the clinical practice has failed so far, despite extensive clinical trials 5 . In this scenario, a promising therapeutic approach, but insufficient traveled avenue, is the remote ischemic conditioning (RIC) 6 . Murry et al. first introduced the ischemic preconditioning therapy in a canine myocardial infarction model in 1986 7 . Since then, it has been repeatedly confirmed in animal models that ischemic preconditioning is a powerful endogenous protective strategy against IRI of multiple organs, including heart, brain and kidneys 8 . A significant breakthrough was the discovery that ischemic conditioning induction to a remote organ from the site of severe ischemia can also protect target tissue 9 . RIC consists of brief episodes of ischemia/reperfusion (I/R) in a distant organ, such as a limb, that can provide protection to the ischemic brain. It can be applied either before ischemia (pre-conditioning; RIPreC), during ischemia (per-conditioning; RIPerC) Data analysis. All outcomes were transformed into effect sizes by using the studies' reported statistics, mean and standard deviation or standard error, or results from analyses including t-tests, analysis of variance, correlations, regressions, and linear mixed-effects models.
The primary outcome was defined as the percentage of volume infarct reduction between RIC and control groups. The meta-analysis was conducted using the packages 'tidyverse' , 'meta' , 'metafor' and 'dmetar' of the R 4.0 software 20 . Studies presenting mean infarct size with standard deviation (SD) or standard error of mean (SEM) values in both intervention and control groups, were extracted for the meta-analysis. The effect size included was the difference in the mean percentage change (control-intervention), presented as the mean percentage change (95% confidence interval) in the infarct size of intervention group with respect to controls. As the SEM of the difference in percentage change was not reported, we first calculate the SD for each group (SEM*√n) to obtain the SEM of the difference (√[(SD12/n1) + (SD22/n2)]). Since the characteristics and methods of the interventions used in the studies are different, a random-effects model with the inverse variance method was performed to calculate the mean effect size. Forest plots were performed to show individual and global effect sizes.
Statistical heterogeneity across studies was evaluated using the Cochran's Q test and I2 statistic. I2 estimates the percentage of variation between all studies that is due to heterogeneity rather than chance; I2 > 50% is considered as substantial heterogeneity. The function find.outliers of the ' dmetar' package was used to explore for possible outliers and the function InfluenceAnalysis was used to detect studies with a high influence on the overall results.
A Baujat Plot was performed to plot the overall heterogeneity contribution and the influence on pooled results for each study in the meta-analysis. As heterogeneity was highly presented in the study and we performed www.nature.com/scientificreports/ meta-regressions and subgroup analyses to explore the effects of the different characteristics on the percentage change in infarct volume. Q statistic was used to assess difference in the subgroup analysis and random-effects linear regression models were performed to assess correlations with quantitative variables. The bubble function of the package 'meta' was used to plot meta-regression outputs. Some quantitative variables (duration of cerebral ischemia, number of cycles and cycle duration) were divided into groups to have a different approach in a subgroup analysis. The analysis was also stratified by animal species, running separate meta-analyses for checking if the effects of some characteristics on the infarct size were different according to the animal tested.
Ethical statement. This article does not contain any studies with human participants or animals performed by any of the authors.

Results
The results of the retrieved literature and selection process are presented in Fig. 1. The initial search identified 286 studies, of which 124 duplicates were removed, leaving 162 studies. After screening by title and abstract, 102 articles were rejected when exclusion criteria were applied: hemorrhagic stroke, articles not related to cerebral ischemia or limb RIC, articles related to humans, animal models with comorbidities, articles not available in English, review articles, letters, proceedings paper and a book chapter. The full text of the remaining 60 articles www.nature.com/scientificreports/ were read and 10 studies were excluded because there was no available data. Finally, fifty studies were included in the meta-analysis with data on 64 experiments.
The key characteristics of the included studies are presented in Table 1. Sixty-one of the 64 (95.3%) experiments we carried out in rodents: 52 used rats (81.2%) and nine studied mice (14.1%). Primates were only used in three experiments (4.7%). Almost all studies were performed in young and healthy animals. The majority of the studies used solely male (93.7%) animals and 6.3% of studies used solely female animals. The majority of publications induced transient focal cerebral ischemia (87.5%) with 20-120 min of middle cerebral artery occlusion (MCAo). Most studies included in this meta-analysis induced cerebral ischemia by intraluminal filament (84.3%) and seven studies used permanent models of middle cerebral artery (10.9%).
The most commonly RIC protocol employed was three to four repetitions of 5-15 min I/R using a pressure cuff, applied on one (37.5%) or two (60.9%) limbs to observe a neuroprotective effect. Thirty-nine of the 64 experiments studied RIPostC (61%), sixteen examined RIPreC (25%) and nine tested RIPerC (14%). In 58 studies, RIC was performed as a single application (90.6%) and six studies employed multiple applications (9.4%). The anesthetic used varied between studies, being chloral hydrate the most used (36%) ( Table 1) (Fig. 2). However, high heterogeneity between studies was detected (I 2 = 90.1%; Q = 635.72, df = 63, p < 0.0001). The variance of the distribution of the effect sizes in this samples was T 2 = 167.06 (Table 2). Figure 3A presented results of the influence analysis. The study that contributed to a higher heterogeneity 21 and the most influential study on the overall results 22 were identified (Fig. 3A).
When RIC was applied to one limb (− 39.1%; 95% CI − 45.7 to − 32.5%) the effect was similar to when it was applied to two limbs (− 38.8%; 95% CI − 43.2 to − 34.5%). Only in one study, which used monkeys, RIC was applied to four extremities (+ 5.0%; 95% CI − 18.4 to + 28.3%) 23 . The results of this study within the meta-analysis showed significant differences (p = 0.013) regarding the number of limbs. However, if this study was not included in the meta-analysis, differences were not observed on the variable number of limbs (p = 0.604). Initiation of RIC was not related with the infarct volume reduction (− 0.461%; 95% CI − 1.180 to − 0.258%; p = 0.205) (Fig. 3E). Finally, significant sex-differences were observed in experiments performed on rats but not on mice. Experiments performed on male animals obtained higher proportion of volume reduction than experiments performed on female animals (− 41.9%; 95% CI − 45.6 to − 38.2% vs. − 28.8%; 95% CI − 44.8 to − 12.8%; p = 0.002), which it might indicate a different RIC's mechanism of action by sex.
Effect of anesthetic on infarct volume reduction. Up to nine different anesthetic strategies were used in the experiments. Among them, chloral hydrate was used in 22 experiments and isoflurane was used in 20, both were the most represented. The combination of ketamine and propofol was only used in monkey experiments. Significant differences were observed in experiments performed on mice (p < 0.001), due to anesthetic strategy. In addition, when we compared the two most frequent used types of anesthesia, chloral hydrate (− 43.37%; 95% CI − 48.73 to − 38.00%) showed a greater infarct volume reduction than isoflurane (− 34.76%; 95% CI − 40.48 to − 29.05%, (p = 0.022). Figure 4 illustrates the schematic representation of suggested underlying mechanisms of RIPreC, RIPerC and RIPostC. Selected studies had also described molecular and cellular processes involved on RIC. Diagram showed the different mechanisms grouped by cellular processes related with ischemic damage: oxidative stress, inflammation, hemodynamics, immune response, autophagy, and apoptosis. However, many molecular pathways were described, none was translated to humans. Special consideration should be given to four spots where no data was reported: no autophagic pathway was related to RIPreC and RIPerC molecular underlying mechanisms were not described on apoptosis, oxidative stress, and immune response.

Pathophysiology of RIC effects.
RIPreC would decrease oxidative stress through the release of endothelin-1 and the increase of H 2 S, Nrf2, HIF-1α, SOD1 and HO1 [24][25][26][27][28] . It would also reduce neuroinflammation by modulating the expression of HIF-1α, HIF-2α and activating the Notch1 and NF-KB pathways 25,29,30 . Apoptosis has been shown to be reduced when preconditioning is applied by regulating the JAK2/STAT3 signalling pathway 31 . Also, an improvement of brain www.nature.com/scientificreports/ edema and downregulation of the expression of AQP4 is observed 32,33 . Several studies have shown that RIPreC modulates the immune response decreasing the levels of IL-10, IL-6 and TNFα in the blood [34][35][36] . RIPerC would inhibit the autophagy process by increasing Bcl-2 phosphorylation 37-39 , decrease inflammation through incrementing Notch and NICD expressions 40 and increase collateral circulation [41][42][43][44] . Finally, RIPostC would decrease brain edema and blood-brain barrier permeability via upregulating eNOS, decreasing MMP-9 and increasing claudin-5 expression 41 www.nature.com/scientificreports/ The solid vertical line represents a mean difference of 0 or no effect. Points to the left of the line represent a reduction in infarct volume, and points to the right of the line indicate an increase. Each square around the point effect represents the mean effect size for that study and reflects the relative weighting of the study to the overall effect size estimate. The larger the box, the greater the study contribution to the overall estimate. The weight that each study contributed is in the right-hand column. MD mean difference, CI confidence interval.  www.nature.com/scientificreports/

Discussion
This systematic review and meta-analysis summarized the evidence on the protective effects of RIC on infarct volume in preclinical stroke models. A total of fifty studies with data on 64 experiments were included, which involved 941 animals. In all studies, the reduction in infarct volume in RIC groups compared to control was 38.4%. Our results suggested that RIC is more efficacious in transient than permanent ischemia, applying a smaller number of RIC cycles, using a RIC cycle length of ≥ 15 min, using one or two limbs, employing barbiturates anesthetics and in male animals. The majority of papers in this review used rodents, predominately rats. Despite being the most applicable animal models for research related to stroke, the demand for larger models, such as rabbits and even nonhuman primates, is increasing to better understand the disease and RIC mechanism of action 64 .
Most RIC studies used transient focal cerebral ischemia with intraluminal suture stroke model because it closely mimics the human ischemic stroke 65 . The optimal conditioning protocol for RIC to elicit organ protection remains unknown. Less than three cycles or more than 15 min of treatment intensity can have a significant role in ischemic neuroprotection. However, more than three ischemic cycles or cycles < 5 min did not have such a neuroprotective effect. The present evidence suggests that there may be a minimum threshold value for the neuroprotective effect of RIC. RIC was beneficial in all three temporal variants after its initial application: RIPreC, RIPerC and RIPostC. Despite this, RIC was found most effective when delivered after stroke injury (RIPostC) followed by the application during stroke (RIPerC). Both approaches are suitable to be translated to patients, where RIC would be applied during ambulance transportation once admission at the emergency room is done or during the first 24 h after the stroke. The preclinical evidence supported the current clinical trials on-going on RIPerC and RIPostC. Interestingly, the reduce in infarct size is related to neurological functional improvement. www.nature.com/scientificreports/ Most studies performed RIC as a single application. A single bout of RIC activates at least 2 distinct time frames of neuroprotection against I/R injury of the brain. The initial neuroprotection is short-lasting (2 h) and occurs immediately after RIC 66 . The delayed form of neuroprotection reappears after 12-24 h and lasts 48-72 h 67 . In addition to the short-lasting benefits of a single bout of RIC, long-term benefits may be induced with repeated daily conditioning 54 . A limited number of studies have explored the effect of repeated RIC in an animal model for brain ischemia 28,30,68,69 . RIC reduced infarct volume in both male and female animals but provided significantly more protection in males. It must be pointed out that only four studies examined the effect of RIC in female animals, so more experimental research on female animals should be done to determine the RIC effects on female animals.
Both rat and mice studies demonstrated significant statistical reduction in infarct volume in RIC groups compared to controls. Subgroup analysis shown that in mice experiments, there was a significant interaction with RIPerC. Subspecies analysis showed no significant interaction with duration of ischemia and number of RIC cycles. However, our analysis demonstrated > 100 min of ischemia to be more effective than < 90 min in rats. Similarly, 60 min of ischemia was more powerful than < 60 min in mice. We found 3 and > 3 cycles to be equally effective in rats, being < 3 the most beneficial. Conversely, > 3 cycles in mice provided a greater neuroprotection. These differences might be related with the total ischaemic dose (cycle number and duration). Interestingly, in rats, doses above 15 min were more effective, while in mice the opposite occurs. The shorter the length of each RIC cycle, the better reduction of the infarct size.
In all species, significant sex-differences were observed in experiments performed on rats but not on mice, showing a significant effect on males. This observation would be explained by the interaction of female's hormones with the RIC's molecular cascade and that most of the studies were performed in male mice. Taking in consideration the sex differences is particularly important because of the translational goal, and it could lead to better treatments for cerebrovascular diseases if RIC might have a differential sex-effect.
Our analysis supports the previous findings of no significant differences in RIC effect when it was applied on one or two limbs 70 . We also noticed a reduction of efficacy if isoflurane is used during surgical procedure 36,71 . Signaling protective pathways associated with the induction of brain ischemic tolerance are known for the inhalational anesthetics, however very little is known about the infused ones. Clinical and experimental studies of the anesthesia effect on ischemic preconditioning should be conducted in the future to determine its effect.
Although the exact mechanisms by which RIC reduces ischemic injury in the brain remain unclear, the currently accepted hypothesis is that preconditioning, perconditioning and postconditioning are all involve in both humoral and neural mechanisms 12 . RIC has been successfully reproduced by dozens of experimental laboratories but translation to the human clinical setting is still a challenge 6 . Despite many clinical trials shown protection to the heart, large randomized controlled trials found no improvement in clinical outcome and mortality in patients undergoing coronary bypass grafting 72 . Several trials are currently ongoing to explore the effects of RIC in ischemic stroke patients 73 . Data from these trials will help to better understand the effectiveness of RIC in AIS patients and will guide potential future implementation of RIC in the clinical practice.
The current systematic review and meta-analysis is the most recent revision of the literature on preclinical studies of RIC. It has considered three RIC strategies individually to define its effects independently, by contrast two recent systematic reviews and meta-analyses 74,75 considered only two RIC strategies (RIPreC vs RIP-erC + RIPostC; RIPreC vs RIPostC). We have considered a subgroup analysis by species (mice, rats, monkeys) because of the vascular hemodynamics of each specie. A detailed summary of the three systematic reviews is provided on Table 3.
Some potential limitations should be stated. First, a large proportion of studies included in the meta-analysis use young male rodents with an absence of animals with co-morbidities which may inhibit the effects of RIC and a lack of adults/aged animals. In clinical studies, RIC would be used to treat aged persons with hypertension, diabetes and dyslipidaemia, which are not represented in preclinical models at this time. Second, considering that the incidence of stroke is higher among women compared to men, with women experiencing poorer outcomes, it is imperative to include more females in future studies. Third, anesthesia during RIC delivery is another concern because it is reportedly neuroprotective in preclinical models of stroke. Fourth, apart from infarct volume, we did not perform the meta-analysis of neurological function outcome because it was reported in a wide variety of different tests which make the analysis weak and deficient, and the high variability might be complicated to understand.

Conclusion
This article, to our knowledge, is the first meta-analysis of RIC in preclinical stroke models that includes data on lesion volume, neurological impairment and mechanisms involved in RIC. This study demonstrated that RIC is a feasible and safe strategy and supported the ability of RIC to reduce infarct size and improve neurological function. However, the present study detected moderate statistical heterogeneity across studies influenced by species. Precise knowledge of RIC optimal dosage, the effects of comorbidities, sex and anesthesia is yet to be found. Further investigation in pre-clinical characterization of the RIC protocol obeying animal research guidelines is needed so that it can be successfully translated to humans.