In two parallel, separate studies, Joe El Khoury and we (a group led by Bruce Lamb and including Sungho Lee, Nick Varvel, and myself) crossed CX3CR1 KOs to APP-PS1 mice (using distinct APP-PS1 models, ours from Matthias Jucker; El Khoury’s from Dave Borchelt) and monitored amyloid deposition. Our results were entirely concordant (using slightly different methods of analysis): there was a strong, gene dosage-dependent decrease in amyloid deposition in the CX3CR1 KO mice. This decrease was not associated with evident change in APP expression, nor in processing. Further, there were fewer microglia associated with each core plaque in the CX3CR1 KOs. The hypothesis was that CX3CR1 KO microglia are more efficient at amyloid phagocytosis, therefore clearing more with fewer cells. Since then, Bruce’s lab has in vitro data to support this hypothesis. These findings (obtained independently by our lab and that of El Khoury) are neither concordant nor discordant with those from Herms et al: their assessment of insoluble Aβ appears to show a non-significant reduction in the KO mice...  Read more

In two parallel, separate studies, Joe El Khoury and we (a group led by Bruce Lamb and including Sungho Lee, Nick Varvel, and myself) crossed CX3CR1 KOs to APP-PS1 mice (using distinct APP-PS1 models, ours from Matthias Jucker; El Khoury’s from Dave Borchelt) and monitored amyloid deposition. Our results were entirely concordant (using slightly different methods of analysis): there was a strong, gene dosage-dependent decrease in amyloid deposition in the CX3CR1 KO mice. This decrease was not associated with evident change in APP expression, nor in processing. Further, there were fewer microglia associated with each core plaque in the CX3CR1 KOs. The hypothesis was that CX3CR1 KO microglia are more efficient at amyloid phagocytosis, therefore clearing more with fewer cells. Since then, Bruce’s lab has in vitro data to support this hypothesis. These findings (obtained independently by our lab and that of El Khoury) are neither concordant nor discordant with those from Herms et al: their assessment of insoluble Aβ appears to show a non-significant reduction in the KO mice (Fig 1I), although the small number of animals assessed might preclude statistical significance.

In another study from Bruce’s group (Kiran Bhaskar), CX3CR1 KO mice (crossed with mice ‘humanized’ for tau [hTau mice]) showed worse tau pathology, dependent on IL1 and p38MAP kinase and resulting in cognitive impairment.

In response, then, to the key question about microglia in general and CX3CR1 in particular, it appears that the altered microglial reaction in CX3CR1 KO mice is a double-edged sword, producing better amyloid phagocytosis and worse tau pathology.

Combining the two aspects of AD pathology (in the triple-Tg) and focusing on a novel assay for neuron loss (monitoring with two-photon imaging), Herms et al. showed benefit related to absence of CX3CR1. Their work represents (to our knowledge) the first evidence for neuron loss in the triple-Tg AD model, and one which would not be observed using stereology (1.8 percent of neurons within one month). It remains uncertain why a uniform 1.8 percent neuron loss would not, however, be recognized if it persisted for six to 12 months. The authors’ hypothesis that microglial activation precedes neuron loss and therefore is causative needs further study: injury to neurons activates microglia, and it can clearly be seen in Fig. 1 (compare day 0 in 1c and 1e) that the +/- microglia are already activated. This conclusion becomes even more solid when one considers that the -/- microglia have two copies of GFP, while the +/- microglia have one and, if imaged similarly, would appear smaller and less prominent.

In summary, Herms et al. have shown neuronal cell loss in an AD model using two-photon imaging, and have provided evidence that microglial CX3CR1 is involved, somehow, in that process. The relationship to amyloid deposition or toxicity, or to tau pathology, needs to be studied further at the mechanistic level.

View all comments by Richard Ransohoff

The recent report from the Herms group offers new insight into the enigmatic relationship between microglia and AD pathobiology. The authors have focused on whether fractalkine receptor on microglial cells participates in neuronal loss using Frank LaFerla’s 3xTg-AD model. The novelty in this paper is really twofold: demonstration of in vivo neuronal loss in real-time, and new biology showing the role of microglial fractalkine receptor (CX3CR1) in mediating this neuronal death. The authors should be commended for taking such an elegant approach, utilizing two-photon intravital imaging. It is interesting that these authors observe neuronal loss within two weeks in fractalkine receptor-sufficient 3xTg-AD mice. This report comes on the heels of another recent Nature Neuroscience paper from Mathias Jücker’s group, where those authors used a ganciclovir cd11b suicide gene approach to destroy microglia in a transgenic APP/PS1 mouse model of AD for two to four weeks. Surprisingly, those authors did not detect altered cerebral amyloidosis or amyloid-associated neuritic dystrophy in AD...  Read more

The recent report from the Herms group offers new insight into the enigmatic relationship between microglia and AD pathobiology. The authors have focused on whether fractalkine receptor on microglial cells participates in neuronal loss using Frank LaFerla’s 3xTg-AD model. The novelty in this paper is really twofold: demonstration of in vivo neuronal loss in real-time, and new biology showing the role of microglial fractalkine receptor (CX3CR1) in mediating this neuronal death. The authors should be commended for taking such an elegant approach, utilizing two-photon intravital imaging. It is interesting that these authors observe neuronal loss within two weeks in fractalkine receptor-sufficient 3xTg-AD mice. This report comes on the heels of another recent Nature Neuroscience paper from Mathias Jücker’s group, where those authors used a ganciclovir cd11b suicide gene approach to destroy microglia in a transgenic APP/PS1 mouse model of AD for two to four weeks. Surprisingly, those authors did not detect altered cerebral amyloidosis or amyloid-associated neuritic dystrophy in AD model mice that were microglia-deficient. When taking the Jücker report together with this present work, one wonders whether there are not AD mouse model-specific effects of microglia. Of course, the only way to answer such a question would be to reproduce both sets of findings in other AD animal models.

I’d like to comment on the present authors’ data showing that fractalkine receptor-sufficient microglia increase in velocity when moving toward the neurons that are marked for death prior to the actual neuronal loss. Perhaps one of the more penetrating questions is, Are microglia initiating neuronal loss or acting at a point downstream, but still on the pathway to, neuronal death? I am sure that we will continue to grapple with this and other questions that have been prompted by this interesting work.

View all comments by Terrence Town