Modern ischaemic stroke management involves intravenous thrombolysis followed by mechanical thrombectomy, which allows markedly higher rates of recanalization and penumbral salvage than thrombolysis alone. However, <50% of treated patients eventually enjoy independent life. It is therefore important to identify complementary therapeutic targets. In rodent models, the salvaged penumbra is consistently affected by selective neuronal loss, which may hinder recovery by interfering with plastic processes, as well as by microglial activation, which may exacerbate neuronal death. However, whether the salvaged penumbra in man is similarly affected is still unclear. Here we determined whether these two processes affect the non-infarcted penumbra in man and, if so, whether they are inter-related. We prospectively recruited patients with (i) acute middle-cerebral artery stroke; (ii) penumbra present on CT perfusion obtained <4.5 h of stroke onset; and (iii) early neurological recovery as a marker of penumbral salvage. PET with 11C-flumazenil and 11C-PK11195, as well as MRI to map the final infarct, were obtained at predefined follow-up times. The presence of selective neuronal loss and microglial activation was determined voxel-wise within the MRI normal-appearing ipsilateral non-infarcted zone and surviving penumbra masks, and their inter-relationship was assessed both across and within patients. Dilated infarct contours were consistently excluded to control for partial volume effects. Across the 16 recruited patients, there was reduced 11C-flumazenil and increased 11C-PK11195 binding in the whole ipsilateral non-infarcted zone (P = 0.04 and 0.02, respectively). Within the non-infarcted penumbra, 11C-flumazenil was also reduced (P = 0.001), but without clear increase in 11C-PK11195 (P = 0.18). There was no significant correlation between 11C-flumazenil and 11C-PK11195 in either compartment. This mechanistic study provides direct evidence for the presence of both neuronal loss and microglial activation in the ipsilateral non-infarcted zone. Further, we demonstrate the presence of neuronal loss affecting the surviving penumbra, with no or only mild microglial activation, and no significant relationship between these two processes. Thus, microglial activation may not contribute to penumbral neuronal loss in man, and its presence in the ipsilateral hemisphere may merely reflect secondary remote degeneration. Selective neuronal loss in the surviving penumbra may represent a novel therapeutic target as an adjunct to penumbral salvage to further improve functional outcome. However, microglial activation may not stand as the primary therapeutic approach. Protecting the penumbra by acutely improving perfusion and oxygenation in conjunction with thrombectomy for example, may be a better approach. 11C-flumazenil PET would be useful to monitor the effects of such therapies.