Background
Intracranial bleeding episodes, or haemorrhagic transformations (HTs), are common in ischaemic stroke [
1]. They can occur spontaneously and, according to large meta-analyses, are present in 24.2% of placebo-treated and 32.5% of recombinant tissue-type plasminogen activator (rt-PA)-treated patients [
2]. However, only those bleeding events which lead to worsening of neurological outcomes, so-called symptomatic intracerebral haemorrhages (sICH), have direct impact on the safety assessment of rt-PA treatment.
sICH is the most serious complication of thrombolysis with rt-PA in stroke, often resulting in devastating consequences. It occurs in ~ 6% of thrombolysed patients and carries 50% mortality rate [
1,
3], causing reluctance among some practitioners to employ rt-PA-induced thrombolysis [
4,
5]. Furthermore, a lack of research exploring the most appropriate management of sICH prevents the establishment of standardised guidelines for treatment of this condition [
6]. There is therefore a need to identify those who are at particular risk and to develop research tools to specifically combat this emergency situation.
HTs during stroke are not uniformly presented but classified radiologically as haemorrhagic infarctions (HI-1 and -2, characterised by petechial bleeding) or parenchymal haematomas (PH-1 or -2, defined as haemorrhage occupying < 30 or > 30% of the ischaemic area with mild or significant space-occupying effect, respectively) [
1,
3,
7]. While all types of HTs can be accompanied by neurological decline, it is mainly the PH-2 type that independently predicts neurological deterioration at 24 h and poor outcomes at 3 month post stroke [
1,
8], particularly if it coincides with other time and neurological criteria [
9].
Notably, many animal studies which assess rt-PA-related HTs do not classify the type of haemorrhage or determine whether it plays a role in functional outcome post-treatment [
10‐
15]. Even in rodent studies that use the clinical HT classification, no correlations are performed to evaluate whether brain haemorrhage actually contributes to neurological decline [
16‐
20]. Taken together, the nature of HTs in current pre-clinical stroke models needs further examination to allow refined simulation of the human rt-PA-related sICH.
Regardless of its outcome, for blood extravasation to occur the blood-brain barrier (BBB) must be compromised. With this in mind, considerable pre-clinical and clinical data suggests that rt-PA disrupts the BBB, potentially contributing to subsequent HT [
21]. This occurs via plasmin-dependent or -independent effects of rt-PA on various cellular and acellular components of the BBB [
22,
23].
Members of the low-density lipoprotein receptor (LDLR) superfamily, in particular LDLR-related protein 1 (LRP-1) [
24], have long been considered to serve as t-PA receptors in the CNS [
24,
25]. Many t-PA-associated BBB breakdown processes during stroke involve t-PA engagement with LRP-1, an interaction which drives upregulation of matrix metalloproteinase (MMP)-3 [
13], MMP-9 [
26,
27] and vascular endothelial growth factor (VEGF) in brain endothelial cells [
28] and microglia [
29], activation of the platelet-derived growth factor (PDGF)-CC [
12] and Rho-kinase pathways [
30] as well as direct cleavage of LRP-1 [
31] in astrocytes and control of vascular tone [
32] via stimulation of smooth-muscle cells [
33].
The pan LDLR antagonist, receptor-associated protein (RAP) [
34], is a small intracellular chaperone which prevents premature ligand binding to LRP-1 and also to other LDLRs such as megalin (gp330; LRP-2), the very low density lipoprotein (VLDL) receptor and the LDLR itself [
34,
35]. RAP improved functional outcome and HT in rats after stroke [
36] and reduced rt-PA-mediated BBB disruption [
28,
31] and brain haemorrhage [
13] in permanent and thrombotic mouse models of middle cerebral artery occlusion (MCAo), respectively. Yet, the efficacy of RAP against rt-PA damage in transient mechanical MCAo, which best enables focusing on off-target effects of rt-PA, has not been reported.
Here, we simulated rt-PA-triggered HT formation and examined its characteristics by appearance (HI or PH) and functional consequences in a mouse MCAo model. We further tested the potential of RAP to reduce rt-PA-induced BBB disruption and bleeding complications. Our results highlight the challenges in pre-clinical simulation of sICH and interestingly suggest that the use of RAP as a sole therapeutic agent might offer benefits in stroke despite being insufficient for effective protection of the BBB from rt-PA during thrombolysis.
Results
rt-PA administration is approved within 4.5 h from stroke onset. Beyond this time window, its risks, primarily for development of sICH and/or oedema following breach of the BBB, outweigh the potential clinical benefits. In an attempt to mimic these occurrences in the mouse we first characterised the degree of BBB disruption (extravasation of albumin and haemoglobin into the brain) and functional deficit in response to rt-PA under sham conditions and in our MCAo stroke model. In line with other pre-clinical reports [
12,
26], intravenous rt-PA (10 mg/kg) administered to uninjured animals did not induce albumin (Fig.
1a) or blood (Fig.
1b) accumulation in the mouse brain 24 h post treatment, suggesting that rt-PA does not affect the mouse BBB under resting conditions (note that no subtraction of the contralateral from the ipsilateral value was performed in shams). In contrast, transient stroke as short as 0.5 h followed by intravenous treatment at reperfusion disrupted the BBB and increased albumin content in the ipsilateral hemisphere by 24 h post MCAo (observed as positive values after subtraction of the contralateral hemisphere) in both vehicle and rt-PA-treated mice, an effect which was dramatically amplified by 29- and 33.1-fold, respectively, when stroke duration was extended to 4 h (
p < 0.0001, Fig.
1c). Notably, albumin accumulation at 24 h did not significantly differ between rt-PA and vehicle in both 0.5 and 4 h stroke periods, indicating that a similar degree of BBB permeation to plasma proteins developed with or without the thrombolytic (Fig.
1c) (see “
Discussion”). Brain haemoglobin levels, however, were significantly increased by 24 h in mice receiving rt-PA at 4 h, but not at 0.5 h post MCAo or in vehicle-treated animals (
p < 0.05, Fig.
1d), suggestive of development of significant HTs only during delayed use of rt-PA. Indeed, macroscopic examination of perfused brains 24 after 4 h MCAo revealed no bleeding or sporadic small petechiae in brains of vehicle-treated animals but extended bleeding with multiple sites of haematomas in most rt-PA-treated mice (Fig.
1e). Hence, our stroke model successfully simulated the time-dependent shift of rt-PA from a safe to unsafe intervention as judged by formation of substantial HT.
We then examined whether these extensive bleeding complications were contributing to functional outcome; in other words, if they could be considered ‘symptomatic’. Noticeably, none of our functional assessments performed 24 h after MCAo (neurological scoring, hanging wire and ANY-maze tests) correlated with the brain haemoglobin levels measured at this time point, irrespective of vehicle or rt-PA (Fig.
1f). This analysis shows that HTs associated with rt-PA in these experimental stroke settings are not predictive of deficit severity and cannot be considered symptomatic. Importantly, autopsies performed on animals which died before the 24 h end-point (transcardially-perfused where possible) did not reveal consistent HTs, including in rt-PA-treated animals (not shown), suggesting that HT was not playing a major role in their functional deterioration and early death.
LDLR blockade by RAP may provide a unifying solution to a number of rt-PA modes of action which lead to BBB disruption and HT during thrombolysis [
22,
23]. We therefore tested the effects of RAP against rt-PA-induced BBB breakdown and formation of HT as well as mortality and functional outcome in our 4 h transient thread occlusion protocol, which to the best of our knowledge has not been attempted before. As observed in our initial characterisation (Fig.
1), rt-PA treatment induced a similar increase in albumin extravasation as seen for vehicle-treated mice (Fig.
2a, b) but caused a significant upsurge in the brain haemoglobin content measured 24 h after stroke (
p < 0.05, Fig.
2c, d). Administration of RAP (2 mg/kg) at reperfusion with or without rt-PA did not significantly decrease total brain albumin levels (Fig.
2a), yet a trend for reduction of total brain haemoglobin was observed (
p = 0.18, Fig.
2c). Normalisation of the ipsilateral value to the contralateral baseline for each mouse (i.e. fold increase analysis) revealed near-significant reductions in cerebral albumin (
p = 0.054, Fig.
2b) and haemoglobin content (
p = 0.06, Fig.
2d) by RAP. Indeed, focal haematomas and diffuse bleeding across the infarct could often be observed in rt-PA + RAP-treated brains 24 h post MCAo, as seen with rt-PA alone albeit to a milder extent (Fig.
2e), in line with the biochemical findings. As mentioned above, our autopsies of mice which died before the 24 h end-point and were not included in the biochemical analysis did not establish a contributory relationship between HT and mortality, reducing the possibility of underestimation of the RAP effect on the BBB. Overall, these data suggest that RAP does not protect the BBB from rt-PA to an extent where blood cells and blood protein extravasation can be effectively inhibited, but may still harbour a partial capacity to attenuate rt-PA-induced development of HT.
Interestingly, despite its weak activity on the BBB, RAP treatment with or without rt-PA resulted in a marked decrease in mortality within 24 h from stroke onset (from 71.4% overall mortality in vehicle and rt-PA groups to 21% in RAP-treated groups,
p < 0.05, Fig.
3a). RAP administration also resulted in a functional benefit since rt-PA + RAP-treated animals received on average significantly lower neurological scores than mice treated with rt-PA alone (
p < 0.05, Fig.
3b). A strong trend for improved grip strength was further demonstrated with RAP in vehicle-treated (
p = 0.07 by t-test, Fig.
3c), but not in rt-PA-treated animals. Notably, none of the mobility parameters assessed by ANY-maze at 24 h were influenced by RAP (Fig.
3d–f). Taken together, these observations suggest that LDLR blockade during severe stroke could offer fundamental benefits such as improved (short-term) survival and attenuation of deficit, which may not be related to rt-PA, BBB breakdown or brain haemorrhage.
Discussion
Because severe intracranial haemorrhage is the main complication limiting thrombolysis with rt-PA in stroke [
1,
3,
6], we attempted to simulate rt-PA-driven HT in a mouse stroke model and to test whether LDLR blockade could protect against its occurrence. However, since only symptomatic bleeding episodes (sICH) are pertinent for the safety assessment of rt-PA, we further evaluated if HTs obtained in the mouse actually carry the functional and structural characteristics of the clinical condition, such as neurological deterioration and development of parenchymal haematoma type-2 (PH-2), respectively [
8]. This distinction is rarely made in rodent stroke studies but may be important for improvement of research efforts.
In line with pre-clinical reports by others [
17,
18,
41], rt-PA caused substantial HTs in our study by 24 h post stroke, but only when given 4 h after stroke initiation. Rodent studies therefore mimic the basic clinical observation showing lower incidence of severe intracranial bleeding at onset to treatment (OTT) ≤ 90 min and a shift towards higher risk of dangerous bleeding outcome at later OTTs [
2], which usually occurs within 24–36 h [
6]. Interestingly, despite the increase in rt-PA-mediated HT after prolonged occlusion, we saw no differences between rt-PA- and vehicle-treated groups in albumin extravasation into the brain within 24 h regardless of stroke duration. These observations provide valuable insight into the most significant activities of rt-PA on the BBB, as they presumably stem from differential opening of the BBB to varying degrees, allowing passage of different sized particles depending on the severity of disruption. Albumin (and other plasma proteins) can penetrate the brain paracellularly through damaged tight junctions or transcellularly via transcytosis [
39], while passage of red blood cells into the brain requires complete capillary disintegration. Hence, it seems that the long occlusion duration (4 h) is sufficient to substantially compromise the BBB for extravasation of plasma proteins regardless of rt-PA (or, that the effect of prolonged MCAo on albumin extravasation is robust enough to reach saturation within 24 h—the time of our evaluation—and mask early acceleration in this process by rt-PA). However, only rt-PA seems to drive the vessel further towards final catastrophic breakdown, permitting the passage of whole blood into the brain. Such vessel disintegration must involve mechanical rupture and total loss of key structural elements of brain capillaries like the lumen-forming endothelial cells and the supportive basement membrane; accordingly, robust rt-PA-associated activities such as plasmin- and MMP-dependent proteolytic degradation of basal lamina and tight junctions [
22], worsening of cerebral inflammation by reduction of regulatory T-cells [
42] and recruitment of other immune cells [
21] as well as changes to blood pressure and vascular tone [
32] should logically become therapeutic candidates in the context of sICH prevention after thrombolysis.
Despite successful modelling of the temporal bleeding formation which occurs in humans, HTs developing post-rt-PA in the mouse did not resemble clinical sICH in their appearance and functional consequence. These bleeding events, while substantial, appeared more as scattered (HI-1) or more confluent yet heterogeneous petechiae (HI-2) rather than parenchymal hematomas, characterised by a homogenous and dense haemorrhage creating a mass effect and occupying a large area of the infarct [
8]. Functionally, no cause-effect relationship could be established between HT and deficit severity. Few possible contributors may account for this outcome, including a lack of sensitivity of the functional assessment methods, the absence of longitudinal HT and deficit monitoring throughout the first 24 h to identify earlier associations (before mortality occurs) or the need to administer rt-PA by infusion rather than a bolus (the latter described in rodent studies with similar outcomes [
16,
17]). Nevertheless, as vehicle-treated mice generally suffered from high mortality similar to rt-PA-treated animals (Fig.
3) and displayed severe deficits 24 post 4 h MCAo (Figs.
1,
3), a lead explanation is a ‘ceiling’ effect, referring to such intense brain damage developing in the mouse as a result of complete MCA shutdown over many hours (required, however, to simulate rt-PA-dependent HT), that it cannot be modified any further when significant haemorrhage occurs [
9]. This notion is highly plausible, since a typical infarct size achieved in mice by the suture MCAo model is approximately an order of magnitude larger than the infarct produced by an average human stroke (by percentage of the hemisphere affected). Equivalent massive infarcts in humans are usually accompanied by substantial oedema, progressive infarct expansion, brain herniation or pan-hemispheric destruction, with minimal functional recovery [
43]. A prolonged transient mouse stroke might therefore be closer in outcome to a malignant human stroke, where a devastating outcome develops with or without HT [
43]. Taken together, our study highlights substantial limitations in pre-clinical modelling of rt-PA-triggered sICH by the popular suture model of transient MCAo in mice. A more suitable protocol to study this phenomena in rodents could consist of transient occlusion of a smaller brain territory for longer periods of time (e.g. 6–9 h) before rt-PA treatment, accompanied by periodic magnetic resonance imaging (MRI). Such approach may reduce baseline deficit but allow the occurrence of substantial bleeding in severely-impaired blood vessels and their progressive monitoring. Worsening of neurological symptoms as HT occurs should be incorporated in models specifically dealing with sICH and its prevention.
A second interesting finding of our study was that LDLR blockade with the pan LDLR antagonist RAP had only a weak capacity to influence BBB-related outcomes of rt-PA treatment, yet it offered marked improvement in short-term survival and more limited attenuation of deficit severity. One might further speculate that RAP-mediated improvements in neurological function is in fact underestimated, since only animals undergoing 4 h MCAo which survived up to 24 h post-occlusion were assessed; earlier functional assessment, before mortality occurred, would have included critically ill animals and may have resulted in better correlation between function and mortality.
Regarding the BBB, while we indeed noticed a non-significant trend of decreased brain albumin with RAP, this observation overall lies in contrast to other studies demonstrating reduction of blood and blood protein extravasation into the brain by intravenous [
13,
28] or intracerebroventricular [
31,
36] administration of RAP, with or without rt-PA treatment. While differences in stroke models most likely contribute to these discrepancies (i.e. rat [
36] vs. mice, thrombotic [
13,
36] vs. mechanical or permanent [
28,
31] vs. transient), our results can be logically explained by the multifaceted nature of rt-PA actions, some of which are LDLR-independent. In particular, while most plasminogen-independent pathways for rt-PA-induced BBB breakdown identify LRP-1 as a rt-PA receptor (see “
Background”) [
22,
23], more recently-discovered, plasmin-dependent cascades have partial or no reliance on LDLRs, including rt-PA/plasmin-driven activation of plasma kallikrein [
11], bradykinin [
44], monocyte chemoattractant protein (MCP)-1 [
45] and Rho-kinase 2 [
46]. The numerous mechanistic avenues by which rt-PA disrupts the BBB during thrombolysis implies that a sole therapeutic agent will probably be unable to achieve sufficient barrier protection. Instead, combination therapy of substances that address multiple pathways simultaneously, for example inhibitors of cell-surface receptors (Imatinib [
12], RAP [
13,
28], Icatibant [
44]) together with plasma and intracellular enzymes antagonists (GM6001 [
20], BPCCB [
11] and fasudil/KD025 [
30,
46], respectively) should be considered, similar to growing approaches in cancer therapy.
Importantly, rt-PA signalling via LDLRs, in particular LRP-1, also plays a key role in neuronal calcium haemostasis via regulation of the N-methyl-
d-aspartate receptor function [
37,
47], linking rt-PA and endogenous t-PA to pathological brain processes such as excitotoxicity and neurotoxicity [
47]. Hence, even in the absence of efficient protection of the BBB, RAP might still reduce neuronal damage and infarct size, as already shown in rats [
36]. Together with other effects of LDLR blockade, such as a decrease in t-PA-induced brain MMP-9 levels [
29], the net consequence of RAP treatment could be a reduction in brain damage to a level where animal survival and gross neurological deficits improve, in line with our observations.
Authors’ contributions
BN formed the concept, designed the experiments and assisted in their execution, analysed data, generated the figures and drafted the manuscript. BRB performed experiments and provided critical technical expertise, analysed data and contributed intellectual input. HH performed experiments and provided critical technical expertise. CGS and RLM supervised the work and provided critical expertise in drafting the manuscript. All authors read and approved the final manuscript.