Characterization of a novel model of global forebrain ischaemia–reperfusion injury in mice and comparison with focal ischaemic and haemorrhagic stroke

Stroke is caused by obstructed blood flow (ischaemia) or unrestricted bleeding in the brain (haemorrhage). Global brain ischaemia occurs after restricted cerebral blood flow e.g. during cardiac arrest. Following ischaemic injury, restoration of blood flow causes ischaemia–reperfusion (I/R) injury which worsens outcome. Secondary injury mechanisms after any stroke are similar, and encompass inflammation, endothelial dysfunction, blood–brain barrier (BBB) damage and apoptosis. We developed a new model of transient global forebrain I/R injury (dual carotid artery ligation; DCAL) and compared the manifestations of this injury with those in a conventional I/R injury model (middle-cerebral artery occlusion; MCAo) and with intracerebral haemorrhage (ICH; collagenase model). MRI revealed that DCAL produced smaller bilateral lesions predominantly localised to the striatum, whereas MCAo produced larger focal corticostriatal lesions. After global forebrain ischaemia mice had worse overall neurological scores, although quantitative locomotor assessment showed MCAo and ICH had significantly worsened mobility. BBB breakdown was highest in the DCAL model while apoptotic activity was highest after ICH. VCAM-1 upregulation was specific to ischaemic models only. Differential transcriptional upregulation of pro-inflammatory chemokines and cytokines and TLRs was seen in the three models. Our findings offer a unique insight into the similarities and differences in how biological processes are regulated after different types of stroke. They also establish a platform for analysis of therapies such as endothelial protective and anti-inflammatory agents that can be applied to all types of stroke.


Results
The DCAL model is a reproducible and reliable method to induce global forebrain ischaemia using a simple and quick surgical method. To assess/confirm the extent of cessation and subsequent reperfusion of blood flow during the procedure, we i.v. injected Evans blue into the mice post bilateral carotid artery ligation as well as DCAL (transient occlusion of the right CCA) to determine the extent of ischaemia and reperfusion in this model. Following bilateral carotid artery ligation, Evan's blue is excluded from the vasculature of the brain spanning the frontal and parietotemporal lobes (Bregma 2 to − 4 mm) but not the cerebellum. This was uniform in both the left and right sides of the brain (DCAL-non-perfused) ( Figure S1A and B). When the regular DCAL procedure was performed i.e. the left carotid artery was clamped for 30 min and then allowed to reperfuse (DCAL-reperfused), significantly higher Evan's blue signal was detected in both the left ( Figure S1A) and right hemispheres (Figure S1B), suggesting strong reperfusion and therefore subsequent reperfusion injury. In sham and naïve animals, similar levels of Evan's blue was detected in both sides of the brain however, only the naïve group had significantly higher levels of Evan's blue compared to the non-perfused brain for the mid-section (0-2 mm anterior to bregma) of the left side of the brain ( Figure S1A) and the latter section (2-4 mm posterior to bregma) of the right side of the brain ( Figure S1B). This suggests that there could be lasting perfusion impairment following DCAL and sham procedures. This further explains why the mice subjected to the sham procedure have a degree of cerebral damage.
Neurological deficits are more pronounced while locomotor function is preserved in mice that have undergone DCAL compared to MCAo and ICH. As expected after DCAL, MCAo and ICH procedures mice had an overall higher median Bederson score compared to sham controls (Fig. 1a). However, compared to ICH and MCAo, DCAL stroke caused the highest median Bederson score among the models.
MCAo resulted in two-fold increase in footslips/m ( Fig. 1b) but deficits were significantly more pronounced after DCAL (Fig. 1b). There were also significant increases in footslips/m after ICH. In the open field assay, mice travelled 50% less distance after MCAo and ICH compared to respective sham animals (Fig. 1c) and also travelled significantly less compared to mice after DCAL. Interestingly, despite other evident neurological deficits, after DCAL, distance travelled and time mobile was not reduced compared to sham animals (Fig. 1c, d. Mice subjected to MCAo and ICH were substantially less mobile compared to sham animals (Fig. 1d). Further, these mice were also significantly less active when compared to mice after DCAL (Fig. 1d). DCAL produces smaller bilateral lesions relative to MCAo but not ICH. Infarcts were visible in both hemispheres in the DCAL model (Fig. 2a), however only the left hemisphere (site of permanent occlusion) had a significantly larger lesion compared to the sham animals (mean volume = 15.6 mm 3 ; Fig. 2a). As expected, MCAo stroke resulted in a unilateral infarct in the right hemisphere (mean = 42.3 mm 3 ; Fig. 2). When compared, MCAo yielded significantly larger lesion volume than the lesion volume caused by DCAL-stroke (n = 5; p = 0.0002). Despite comparable neurological deficits, the mean infarct volume after ICH was less than ischaemic stroke (mean = 12.3 mm 3 ; Fig. 2) but was similar to infarct volume in the DCAL model. Corresponding cresyl violet stained sections taken from the same brains clearly delineated the infarcts in all three stroke models (Fig. 2c).
Apoptotic activity is comparable between DCAL and MCAo models but higher after ICH. Both ischaemia and haemorrhage triggered apoptotic cell death in the brain at 24 h (Fig. 2b) www.nature.com/scientificreports/ pase 3/7 activity was significantly higher in both sides of the brain (Fig. 2b). After ICH, caspase activity increased 2 × in the haemorrhaged site, and notably, this model resulted in the highest apoptotic activity amongst the 3 models ( Fig. 2b) despite resulting the lowest lesion volume.

Albumin extravasation is increased at 24 h post injury after DCAL, MCAo and ICH. At 24 h
post-DCAL, BBB permeability increased in the site of transient occlusion, but there was a further two-fold increase in the side of permanent occlusion, consistent with increased lesion volume in this side of the brain (Fig. 3). BBB permeability was significantly increased after MCAo, albeit significantly less than the DCAL model (Fig. 3). Albumin extravasation was highest overall after ICH (Fig. 3). However, there is an obvious caveat in that albumin would increase if there was bleeding in the brain. Therefore, we measured and correlated albumin and haemoglobin levels in the peri-haematomal region. Haemoglobin was increased both at 3 and 24 h post-ICH ( Figure S2). At 3 h, albumin and haemoglobin levels correlated, confirming that the high amount of albumin was in part due to the accumulation of blood (n = 8; Pearson r = 0.68; p = 0.0116-not shown). However, at 24 h, there was no correlation between albumin and haemoglobin levels (n = 8; Pearson r = 0.27; P = 0.55-not shown), suggesting that albumin content detected within the brain parenchyma at this time was due to plasma protein extravasation as a result of increased BBB permeability.

VCAM expression varies at 3 h and 24 h post injury after DCAL, MCAo and ICH. At 3 h post
DCAL-stroke, mice had a 2-threefold upregulation of VCAM in both hemispheres ( Fig. 4a) but not at 24 h post injury (Fig. 4b). In the MCAo model, VCAM was significantly upregulated in both sides of the brain at 3 h (Fig. 4a) and interestingly it was still upregulated at 24 h post-MCAo (Fig. 4b). In contrast, in the ICH model, VCAM was not upregulated at either timepoint ( Fig. 4a, b).  www.nature.com/scientificreports/ and sevenfold respectively after DCAL (Fig. 5a, b). After MCAo, both C3aR and C5aR were increased tenfold  www.nature.com/scientificreports/ ( Fig. 5a, b). However, only C5aR was upregulated in the ICH model ( Fig. 5a, b). C5aR is also significantly more upregulated after DCAL stroke unlike MCAo and ICH models. Expression of CXCL-1, -2, -10 and -12 was not increased after DCAL ( Fig. 5c-f) but CXCL-1, -2 and -10 were drastically upregulated after MCAo ( Fig. 5c-f). Modest but significant upregulation of CXCL-1 and -2 was seen after ICH (Fig. 5c, d). However, expression of CXCL12 was significantly higher in the DCAL model when compared to MCAo and ICH. Remarkably, because of an evident sham effect in this model, the difference between DCAL and sham was not significant.
There were three-and fivefold increases in expression of TLR2 and TLR4 in the DCAL and MCAo models respectively (Fig. 5g, h) and DCAL stroke had the highest expression of TLR4 within the models. Unlike for ischaemic stroke, there were no significant increases of TLRs in the ICH model (Fig. 5g, h).
Expression of IL-1α and IL-6 significantly increased after DCAL (Fig. 5i-l) and was also the most upregulated in the stroke models. Similarly, in the MCAo model, IL-1α, IL-1β, IL-6 and TNFα increased exponentially while only TNF-α was upregulated at 24 h after ICH ( Fig. 5i-l). The inflammatory response in the DCAL-sham animals is once again evident for IL-1β and TNFα (Fig. 5j, l). IFN-γ was not upregulated at this timepoint in any of these models (not shown). HIF1α was only upregulated after ischaemic stroke and significantly upregulated after DCAL stroke when compared to MCAo (Fig. S3).

Discussion
Here we investigated and compared biochemical and neurobehavioural changes at 24 h in a new model of global forebrain ischaemia with focal stroke, including MCAo and ICH.
Global ischaemia models induce widespread hypoxic-ischaemic injury due to oligemia and I/R injury and recapitulate the neurological effects of cardiac arrest and profound systemic hypotension due to peripheral haemorrhage, strangulation or drowning 8 . The DCAL model is most similar to the transient BCCAo model 9 and the mouse model of global cerebral ischemia wherein the bilateral common carotid artery occlusion is combined Figure 5. Gene expression analysis of inflammatory genes highlights significant differences between all three stroke models. Real-time PCR analysis of RNA isolated from the site of injury at 24 h shows an overall increase of Complement receptors (a) C3aR and (b) C5aR expression in DCAL, MCAo and ICH models. There was only a significant increase in chemokines (c) CXCL-1 and (d) -2 after ICH, and CXCL-1, -2 and (e) -10 after MCAo but not in (f) CXCL-12. There was an overall increase in (g-h) TLR 2/4 gene expression after ischaemic stroke, but there was no discernible increase after haemorrhagic stroke (i-l) Inflammatory gene expression was high variable; only MCAo resulted in an upregulation of all the inflammatory genes except IFNγ. (Data is mean ± SEM; Sham: n = 5-9, Stroke: n = 5-8; *p < 0.05, **p < 0.001, ***p < 0.0001; t-test, sham vs stroke, #p < 0.05, ##p < 0.001, ###p < 0.0001, ####p < 0.00001, two-way ANOVA with Tukey post-hoc analysis, DCAL vs MCAo vs ICH). www.nature.com/scientificreports/ with isoflurane-induced hypotension 10 . Other variations include the bilateral common carotid artery stenosis (BCAS) 11 and bilateral internal carotid artery occlusion (BICAO) models 12 .
We showed that in this model blood flow is restricted across the frontal, parietal and temporal lobes of the brain during the ischaemic period. Without occlusion of the basilar artery, the cerebellum remains perfused (not shown) indicating patent and well-established posterior collateral supply via the Circle of Willis. Therefore it is a model of global forebrain ischaemia. After 30 min when the vascular clip on the right CCA is removed, both sides of the brain reperfuse immediately ( Figure S1). Therefore the DCAL model allows us to study the contribution of I/R injury following a transient period of ischaemia as well as provides the means to directly compare the effects of 30 min of global to focal cerebral ischaemia (MCAo model).
It appears that hypoperfusion occurs after permanent suturing of the left CCA for the sham procedure as the extent of Evan's blue perfusion in these animals is not as high as in naïve untreated animals. Further there are pathological changes, particularly in the inflammatory response, albumin extravasation and endothelial activation in the sham animals that are not seen in the MCAo and ICH sham groups. Indeed, oligemia following CCA ligation is known to cause an inflammatory response without resulting in a quantifiable infarct 13 . Therefore in the DCAL model, we can isolate effects of oligemia and I/R injury from hypoperfusion, although further experiments are needed to confirm this.
Neuronal cell death following global ischaemia occurs by a different mechanism compared to focal ischaemia 14 . At 24 h after DCAL-stroke, mean MRI-based infarct volumes were 15.6 mm 3 for the left side of the brain (ipsilateral to the site of permanent ligation) and approx. half i.e. 8.8mm 3 for the right side of the brain (site of transient occlusion). This demonstrates that although reperfusion injury may have been of greater magnitude on the right side of the brain, the left side of the brain suffers the added effect of oligemia which exacerbates cell death that is otherwise not caused by mild hypoperfusion likely seen in the sham cohort. The predominantly striatal lesions are consistent with published global ischaemic brain injury models such as BCCAo 15 . Ischaemia reduces the metabolic rate within the tissue and hence there is more cell loss in the striatum because it is the region of the brain with the highest degree of blood flow-glucose metabolism dissociation 16 . Notably, the overall infarct volume in both hemispheres was still significantly less than that of the unilateral focal infarct caused by MCAo (analysis not shown). Despite this, with a median Bederson score of 4, overall neurological outcome was worse in this cohort when compared to the MCAo cohort. Pronounced neurological deficit after DCAL-stroke was reinforced by the finding of increased footslips/m in this cohort, which was ~ 4.5 × than that observed in the MCAo cohort. However, these neurological impairments did not impact the mobility of these mice or their locomotor activity over a 5 min period. This can be explained by the hyperactive and anxious behaviour observed in these mice. Indeed, an increase in locomotor activity and anxiety has been reported in rodent models of global cerebral ischaemia 17 as well as in humans 18 , and has been attributed to post-ischaemic microglial activation in the brain 19 .
The intraluminal thread model used to cause MCAo is relatively non-invasive and results in a large infarct localised to striatal regions, particularly basal ganglia and cortex; it is typically considered as a clinically relevant model of large vessel occlusion 20 . At 24 h after 30 min transient MCAo, median Bederson score was 3 and mean lesion volume was 42.3mm 3 , consistent with a recent report that documented Bederson scores of 3-5 and mean lesion volume of 77.5mm 3 after 45 min of MCAo 21 . Mice subjected to MCAo stroke had quantifiable neurological deficits both in terms of distance travelled and time immobile when compared to the sham operated control mice. Additionally, there was a ~ threefold increase in the number of footslips per metre measured by the ANYMaze system further attesting to the significant neurological dysfunction caused by the MCAo model, and consistent with published reports 22 .
The collagenase injection model is most widely used and best recapitulates blood vessel rupture leading to bleeding, re-bleeding, and consequent haematoma expansion 23 . Although enzymatic disruption of the extracellular matrix is also observed in haemorrhagic infarctions and traumatic haemorrhages in humans, the use of collagenase has its drawbacks. By degrading collagen IV, a part of the extracellular matrix in the brain, collagenase itself affects BBB integrity, alters immune responses as well as causes significant inflammation, which can affect how ICH progresses. The collagenase model is nevertheless suited to this study because it is reproducible, produces a larger injury and causes greater and longer lasting neurological deficits 24 . We found that although the infarct volume was comparable to DCAL stroke, overall neurological dysfunction was less than both ischaemic models. However, similar to focal ischemia, mobility was significantly impaired after ICH.
Following ischaemia, caspase-3 is cleaved and activated and subsequently degrades several substrates in the cytoplasm and nucleus, precipitating in cell death 25 . Our findings are consistent with published data investigating both rat 26 and mouse 27 MCAo models; further, activated caspase-3 has also been detected in post-mortem brain tissue from humans who have sustained stroke following occlusion of a major artery 28 . The delayed wave of caspase-3 activity that has been observed at > 12 h post-MCAo defines the secondary expansion of the lesion by energy dependent caspase activation leading to apoptosis and therefore identifies the ischaemic penumbra 29 . After MCAo-stroke we found a 5 × increase in apoptotic activity within the ischaemic lesion. Increased caspase-3 activity is predominately localised to the CA3 region of the hippocampus and striatum 24 h after global ischaemia induced in the BCCAo model 9,30 . We observed that overall, after DCAL-stroke apoptosis in the ischaemic lesion was comparatively less than that of the MCAo cohorts. Caspase3/7 activity doubled in the site ipsilateral to the transient 30 min occlusion but was 3 × in the site ipsilateral to the permanent occlusion, accurately mirroring the infarct volume data. Increased cytochrome c, a pre-cursor to caspase-3 cleavage in the intrinsic apoptosis pathway has been reported at 24 h in rat ICH models involving autologous blood-injection 31 and increased apoptotic activity has been reported at the same time point in rat 32 and mouse models 33 of collagenase injections. Caspase activity was increased fourfold in the ipsilateral cortex in our model, which was ~ 2 × that seen in the MCAo and DCAL models at the same timepoint. After ICH, added pathophysiological processes such as mechanical damage caused by mass effect of extravasated blood, influx of blood components including complement molecules www.nature.com/scientificreports/ and coagulation factors, and haemoglobin toxicity 31 might be promoting apoptosis compared to that seen in ischaemia models. In the DCAL model, even though infarct volume is less, overall BBB breakdown is worse than in MCAo. While there is a modest increase in the side of the brain ipsilateral to temporary occlusion at 3 and 24 h, there is a significant 2 × increase in albumin extravasation in the side of the brain ipsilateral to the permanent occlusion at 24 h. There is a clear effect of the sham procedure on albumin levels in the DCAL model, demonstrating that hypoperfusion as a result of permanent ligation of the left CCA itself disrupts the neurovascular unit. Presumably, the increase in albumin seen in the side of the permanent occlusion is the net impact of oligemia/hypoxia and I/R injury, whereas that seen in the side of transient occlusion is purely due to I/R injury. BBB disruption is higher due to reperfusion injury in focal transient MCAo models when compared with permanent MCAo. After transient MCAo, there is a biphasic pattern of BBB permeability; there is an increase at 3 h 34 that is maintained up to 24h 35 , followed by another increase in BBB permeability 48 h after reperfusion 34 . Our data showing increased albumin extravasation 3 h after transient MCAo and persisting for 24 h is consistent with previous publications.
Following haemorrhagic stroke, bleeding in the brain triggers an influx of all blood proteins and therefore measuring BBB permeability is best done by intravenously injecting fluorescent tracers of different molecular weights after ICH and studying how they accumulate in the brain. Using a radiolabelled tracer, Lee and colleagues were able to show increased BBB permeability in rats subjected to intracerebral thrombin infusion 36 . Because our study was designed to compare ischaemic and haemorrhagic stroke, we studied albumin extravasation in all three models. Additionally, for the ICH model, by studying haemoglobin and albumin we were able to perform statistical correlation studies to demonstrate that at 24 h post-ICH there is a significant increase in BBB permeability in the peri-haematomal region. It is well known that BBB disruption after ICH is caused by the release of iron (due to breakdown of haemoglobin) after red blood cell lysis and subsequently results in formation of vasogenic oedema 37 .
Cerebral microvascular endothelial cells are rapidly activated after a stroke and upregulate a range of proinflammatory factors, including VCAM-1. VCAM-1 promotes adhesion and migration of peripheral leukocytes and BBB damage. Gauberti et al. conducted a thorough analysis of VCAM-1 using MR-imaging of microparticles of iron oxide (MPIOs) targeted to VCAM-1 via an antibody. They demonstrated that VCAM-1 is globally upregulated within 24 h after MCAo and persists for at least 5 days, thereby providing evidence of an inflammatory penumbra in this model 13 . Our findings confirm that VCAM-1 is indeed upregulated both in the ipsilateral and contralateral sides of the brain at 24 h post-MCAo. Gauberti et al. also confirmed enhanced VCAM-1 upregulation after ischaemia/oligemia 13 ; while our data show that VCAM-1 is upregulated in both sides of the brain in the DCAL model, and this is over and above that caused by oligemia likely seen in the sham cohort. Others have also confirmed VCAM-1 upregulation in global ischaemic brain injury 38 . Taken together, VCAM-1 upregulation seen at 3 h but not at 24 h after bilateral forebrain ischaemia via the DCAL model can be attributed predominantly to I/R injury.
However, our findings that VCAM-1 is not upregulated at 3 or 24 h in the haematomal or perihaematomal (Fig. 4a, b) region after ICH do not align with the haemorrhagic stroke data from Gauberti et al. 13 which showed a wide perihaematomal distribution of VCAM-1 in the acute phase of ICH in the same model as ours. This could be due to the high sensitivity of specific molecular localisation of VCAM and high resolution imaging in their study when compared to conventional immunoblotting methods that we utilised. Also, the dose of collagenase was 3 × more concentrated what we used, and hence corresponds to a more severe injury. Consistent with our findings, Liesz et al. also showed that VCAM-1 is not upregulated until 72 h post-ICH and only in the autologous blood injection model not the collagenase model 39 .
During ischaemic injury to the brain, free radical release activates the pro-inflammatory nuclear transcription factor NF-κB which triggers the transcription of genes encoding pro-inflammatory cytokines IL-1 α and β, IL-6, TNF-α. TNF-α and IL-1α in turn upregulate endothelial bound VCAM-1 16 . In addition to an increase in VCAM-1 mRNA (not shown), we found a significant increase in transcription of these cytokines within the lesion, 24 h after MCAo, consistent with published reports 40 . In contrast only IL-1 α and IL-6 were upregulated in the side of the brain corresponding to the transient occlusion at this timepoint after DCAL stroke. Although TNF-α is known to be upregulated in global brain ischaemia, it has a biphasic response peaking at 3 h and then again at 36 h 14 . In our data, it is interesting that there is lack of a significant change in TNF-α mRNA levels after the DCAL procedure primarily because of the sham effect wherein TNF-α levels in the brains of sham animals are markedly higher than that seen in the sham controls from the MCAo cohort. Similar to the albumin data, this further supports the pronounced inflammatory effect of oligemia and demonstrates that I/R injury does not trigger an additional increase in TNF-α mRNA at this timepoint. It remains to be determined whether there is an early increase in TNF-α corresponding to the VCAM-1 upregulation at 3 h post-DCAL as it is a known inducer of VCAM-1 40 . In contrast, after haemorrhagic stroke, only TNF-α was significantly upregulated at 24 h. In both the collagenase and autologous blood injection models of ICH, microglial produced IL-1, TNF-α and IL-6 peak early at 3 h, while IFN-γ peaks only at 7 days and coincides with the delayed invasion of lymphocytes after ICH 39 .
Chemokines can be secreted by a range of cell types following stroke 41 . Following focal ischaemia increased chemokine expression promotes leukocyte infiltration and rapid peri-infarct upregulation of CXCL1 and CXCL10 occurs has been observed in a rodent model of permanent ischaemia 42 . We found that CXCL-1, -2 and -10 were all upregulated after focal ischaemia but not in global forebrain ischaemia induced by the DCAL model. In contrast, CXCL12 and HIF-1α were highly upregulated only after DCAL albeit there is a strong sham effect for CXCL12. HIF-1α is a transcription factor commonly upregulated during cerebral ischaemia to help neuronal cells adapt to the lack of oxygen 43 . Similar to other models of carotid occlusion 30 , and consistent with the neural response on hypoxia, in the DCAL model we saw significant upregulation of HIF-1α. Also known as SDF-1 (stromal-derived factor-1), CXCL12 expression is known to be regulated by the HIF-1 α in endothelial cells in response to reduced oxygen tension 43 . This could explain why both HIF-1 α and CXCL12 are concomitantly www.nature.com/scientificreports/ upregulated in the mouse brain after DCAL. The increase in expression of CXCL12 in the sham animals further suggests that there is an effect of hypoxia due to permanent ligation of the left CCA in these animals. Following ICH, activated microglia release chemokines that promote recruitment of peripheral inflammatory cells and produce pro-inflammatory cytokines 44 . We found that expression of CXCL1 and CXCL2 was significantly elevated at 24 h post-ICH, potentially secreted by astrocytes and mediated via TLR2 which is induced at 6 h post-collagenase infusion 45 . However, we did not detect a significant increase in TLR2 expression at this timepoint and it remains to be determined whether TLR2 is upregulated at 6 h in our model. Neither did we find a significant change in TLR4 expression at this timepoint, unlike published reports showing significant upregulation of TLR4 in peri-haematomal tissue in the autologous blood injection model at the same timepoint 46 . In contrast, both TLR2 and TLR4 were significantly increased after focal and global ischaemia. During ischaemic injury, microglial TLR2 stimulation results in increases in TNF-α, IL-6 and IL-10, and TLR4 stimulation results in increases in TNF-α, IL-6, IL-10, CXCL-10 and IFN-β 47 . These data reveal the similarities in the molecular response to I/R injury regardless of the animal model chosen.
There is cumulative evidence that activation of complement anaphylatoxins is a key pathogenic mechanism of ischaemic brain injury. Under resting conditions in the brain, receptors for the complement C3a and C5a are predominately expressed by neurons, but under inflammatory conditions both glial cells and neurons express C3aR and C5aR. However, following focal ischaemic brain injury, C3aR and C5aR expression on endothelial cells and astrocytes was upregulated from 6-48 h, which promoted neutrophil recruitment and worsened brain injury 48 . C3aR upregulated under conditions of hypoxic ischaemic brain injury in neonatal rodents is conversely linked to improved outcome 49 . Hence although we show that C3aR expression is upregulated in both MCAo and DCAL models, we cannot ascertain if it is protective or detrimental without further studies. After ICH, thrombin activates the complement system to facilitate red blood cell lysis and C3aR has been linked to inflammatory cell infiltration in the autologous blood injection model of ICH 50 . We found a significant upregulation of expression of C5aR and not C3aR in our ICH model, at 24 h post-haemorrhage. C5aR activation is also associated with worsened outcome after ICH, and mice treated with C5a receptor antagonist showed less granulocyte infiltration into the lesion 51 .
In Table 1 we have summarised our findings about the DCAL model and how it compares to the MCAo and ICH models. It should be noted that we studied male C57Bl6 mice only and we will need further studies to confirm if the same trends will be obtained in female mice and other strains of mice.

Conclusions:
The initial triggers are different for ICH, ischaemic stroke and global ischaemic brain injury, and the reason for ensuing brain injury differs in each type of stroke. But as us and others have shown, all types of stroke have a final common pathway-cell death, endothelial activation contributing to arterial wall damage, inflammation and BBB breakdown. Certain interventions such as blood pressure lowering drugs have already been trialled in cases of ischaemic and haemorrhagic strokes with variable outcomes 52 . Anti-inflammatory or endothelial protective drugs are also likely to be protective for both ischaemic and haemorrhagic stroke. Patients with ischaemic cerebrovascular disease have a high risk of ICH, and vice versa. Thrombolysis after ischaemic stroke is also associated with increased risk of ICH 53 . For these reasons we established and characterised uniform readouts across rodent ischaemic and haemorrhagic stroke models. We found that although cell death was comparable between the global and focal ischaemia models, neurological outcome and albumin extravasation www.nature.com/scientificreports/ were significantly worse in the site of permanent occlusion after global forebrain stroke. Molecular responses including VCAM-1 upregulation and transcriptional upregulation of pro-inflammatory cytokines and TLRs/ complement differed vastly in the three models. These techniques and models can help develop new therapies for I/R injury that can be tested for safety in models of haemorrhagic stroke. They can also further aid in the pre-clinical development of pharmacotherapies that are safe for all stroke types and therefore preclude need for neuroimaging prior to administration.

Experimental procedures
In vivo experiments. Mice MCAo. Transient focal cerebral ischaemia (30 min) was induced using MCAo, as described previously 54 . Briefly, mice were anesthetized with isoflurane (1.5-2%), and rectal temperature was maintained at 37 °C. MCAo was induced by transecting the ECA, temporarily tying off the CCA and using the external carotid artery trunk as a side path to pass a suture through the internal carotid artery (ICA) to lodge in the junction of the anterior and middle cerebral arteries. A silicone rubber-coated monofilament suture was inserted via the right ECA until it obstructed the MCA, and the common carotid artery was allowed to perfuse for the duration of the ischemic period of 30 min. CBF was measured using transcranial laser Doppler flowmetry (Periflux System 5010; Perimed) in the centre (bregma coordinates: 2 mm posterior, 5 mm lateral) of the ischemic territory. For sham controls, vessels were visualized and cleared of connective tissue under anaesthesia. All animals were left to recover on the heat pad for at least 4 h after procedure.
Double carotid artery ligation (DCAL). Transient global forebrain ischaemia was induced by permanently ligating the left common carotid artery (CCA), and transiently clamping the right CCA. To permanently occlude the left CCA, fat and the surrounding connective tissue was dissected and a 2-3 cm length of 5-0 silk suture was tied around it, taking care not to include the nerve. After a 10 min rest period to minimise overstimulation of the vagal nerve, the right CCA was cleaned of fat and connective tissue, and a vascular clip was used to transiently clamp the artery for 30 min after which, it was removed and the animal was recovered for 24 h. This model was specifically designed to assess I/R injury independently of the effect of ischaemia/oligemia and therefore, for sham controls, the left CCA was permanently ligated, and the mice were allowed to recover. There were no deaths as an outcome of this procedure.
To determine the extent of ischaemia and reperfusion following DCAL, mice were subjected to DCAL procedure; in one cohort both arteries were ligated and in the other the right common carotid artery was transiently occluded as is done in the DCAL procedure. Sham procedure whereby the left CCA was permanently occluded, and naïve animals were also included. Evan's blue was then intravenously injected and allowed to circulate before mice were culled (Refer to supplementary methods for more information).
ICH. Mice were anaesthetised with 2,2-tribromoethanol (Avertin) and 0.15U of Collagenase VII (from Clostridium histolyticum; both from Sigma Aldrich Australia) was injected intracerebrally as described 24 at coordinates MV: 1.75, AP: 0.4, DV: 3.8 from bregma. For sham controls, 1 µl saline was injected at the same location, instead of collagenase. The syringe was held in place for 5 min and gently withdrawn, the hole in the skull was sealed with bone wax, and the animals were left to recover overnight with supportive care as above.
Euthanasia and tissue harvesting. Supplementary methods.
Behavioural testing. Behaviour was assessed 20-24 h post-stroke as detailed below.
• Open field and parallel rod floor assays: Mice were placed in a 30 × 30 cm black Plexiglas box or in a box containing narrow wires 1 cm apart as described 22 . Behavioural activity was tracked using AnyMaze software (Stoelting Co., USA).
Magnetic resonance imaging. At 24 h post-stroke, ex-vivo MRI was carried out as described previously 56 . Analysis was performed using ITK-SNAP 57 . Refer to supplementary methods for further information.
Cresyl violet staining. After ex vivo MRI, brains were cleaned of agar, embedded in paraffin and 5 µm sections were stained with Cresyl violet at the Monash Histology Platform.
Caspase-3/7 activity assay. The fluorogenic substrate (Ac-DEVD-AFC) was used, and the protocol was according to Notte et al 58 .