Synaptic elimination and the complement system in Alzheimer’s disease
28
This conclusion gained further support when we analysed the result of the
burst stimulations. C3 KO mice displayed an increased synaptic response to
burst-stimulations (Fig. 5, paper II), probably due to the normalising effect on
release probability of the short-term plasticity induced by the burst.
To corroborate the electrophysiological findings, we made
immunohistochemical stainings of VGLUT1 in thin sections of hippocampal
tissue from WT and C3 KO mice. Again surprised, we found that C3 KO
mice had a lower density and smaller puncta of VGUT1 staining (Fig. 6,
paper II), thus implying fewer synapses. An explanation for this result that
would still fit together with the previous results, is, that in the C3 KO mice,
synapses are more numerous, but smaller, so small that they creep under the
detection level for immunohistochemistry. This rational comes with the
prediction that the VGLUT1 protein levels in C3 KO mice should not be
lower than in WT mice. Indeed, western blot and qPCR experiments showed
that VGLUT1 expression was not different between C3 KO and WT mice,
thus supporting the notion of more numerous but smaller synapses. In further
support of this notion, it has been shown that mice lacking C1q, a protein
found upstream of C3 in the classical complement cascade, in fact, have
smaller synaptic boutons (Chu et al. 2010).
Considering the importance of proper synapse elimination during
development, C3 KO mice should display some learning disabilities. We
confirmed this prediction by subjecting the animals to fear conditioning,
showing that the C3 KO mice were blessed with an impaired memory for the
foot shocks when the learning paradigm was hippocampus-dependent (Fig. 7,
paper II). Interestingly, this opens for the possibility that synaptic elimination
in the amygdala is not impaired by the lack of C3.
In the last experiment of this study, we explored the possibility of an
increased propensity for epileptiform activity in the brains of the C3 KO
animals. Such an effect was seen in the previous study of C1q deficient mice
(Chu et al. 2010) and could possibly be explained by the increased burst
response reported in the present study. This notion is further supported by the
fact that inhibitory GABAergic synapses do not seem to be eliminated in a
complement-dependent manner (Chu et al. 2010). Therefore we recorded
spontaneous activity in vivo in C3 KO and WT mice. We did, however, not
observe any spontaneous seizures in the C3 KO mice (Fig. 8, paper II).
In conclusion, these results implicate the complement system in the
elimination of glutamate synapses in the hippocampus. They also suggest that
the increased neuronal activity, an expected consequence of deficient
Jonny Daborg
29
elimination and too many glutamate synapses, is partially compensated by
reduced release probability, possibly via homeostatic plasticity.
4.3
Complement levels in AD CSF
Previous reports have shown that post mortem brain tissue from AD patients
have increased levels of complement mRNA (Yasojima et al. 1999), and that
AD CSF has increased levels of a certain isoform of C4b (Finehout et al.
2005), C3 and factor H (Wang et al. 2011). Since the complement system
seems to be involved in AD we evaluated the complement proteins, C3, C4
and CR1 as potential CSF biomarkers for AD.
C3 and C4 CSF levels were significantly higher in AD patients as compared
to stable MCI patients (Table 1, and Fig. 1, paper III). Since the core
biomarker patterns were similar between the control and stable MCI groups,
and the MCI-AD and AD groups (Table 1, paper III), these were merged into
a control/stable MCI group and an MCI-AD/AD group (Fig. 2, paper III).
Statistical comparison of these two groups showed that the CR1 levels were
significantly increased in the MCI-AD/AD group. C3 and C4 levels were not
significantly different between the groups, but there was a trend towards
higher C3 levels in the MCI-AD/ AD group (p = 0.068).
To evaluate if the levels of C3, C4, or CR1 could be of diagnostic utility,
receiver operating characteristic (ROC) curves were created for the different
analytes. The core biomarkers T-tau, P-tau and Aβ42 could differentiate AD
from controls with an area under the curve (AUC) of 0.96. When we added
the complement proteins to this model, the AUC did not improve (Fig. 3,
paper III). The variable importance in the projection (VIP) plots show the
contributions of the different analytes to the model (Fig. 3, paper III).
Although C3 and CR1 contributed to some extent, the magnitude was too
small to be of clinical use.
In addition to the primary analyses described above, we also investigated
potential correlations between the different variables in the two merged
groups (Table 2, paper III). The results showed that the levels of the
complement proteins correlated with age and albumin ratio, an indicator of
the blood-brain barrier function. Correlations were also seen between age,
MMSE and albumin ratio.
The correlation between complement levels and age has been reported in a
previous study (Wang et al. 2011) and is somewhat unsettling considering the
significant differences in age seen between the diagnostic groups in the