SORLA, a negative regulator of APP processing and major AD risk factor


Introduction.  SORLA was identified initially in quest for novel lipoprotein receptors in the human brain. However, possible contributions of this receptor to neuronal lipoprotein metabolism remained obscure. We were the first to suggest an alternative role for SORLA as neuronal receptor for APP that prevents breakdown of this precursor into neurotoxic Ab peptides, a pathological hallmark of AD (Andersen et al., PNAS, 2005; Andersen et al., Trends Neurosci, 2006). In the past 5 years, we substantiated the relevance of SORLA for amyloidogenic processing in patients and animal models. Based on this work, SORLA is now considered a major risk factor for AD.


SORL1 variants are genetic AD risk factors. One major aim of our work has been to validate the relevance of SORL1 (the gene encoding SORLA) as genetic risk factor for sporadic AD. Initially, we showed that SORLA levels are significantly reduced in brains of AD patients (Andersen et al., PNAS, 2005). Subsequently, we contributed to a large epidemiological study that demonstrated association of several SNPs in SORL1 with AD in Caucasian populations (Rogaeva et al., Nat Genet, 2007).

Finally, we identified two SNPs associated with poor receptor expression in AD brains. Both SNPs alter the SORL1 transcript sequence, switching from frequent to rare codon usage in the risk genotype. Studies in cultured cells confirmed less efficient translation of risk transcripts into protein, providing an explanatory mechanism how SORL1 genotypes cause poor receptor expression in the human brain (Caglayan et al., Arch Neurol, 2012). Finally, we documented dysfunction of a SORL1 gene variant as the possible cause of early-onset AD in humans (Caglayan, STM, 2014).


Loss of SORLA promotes AD-related processes in mouse models. Thus, a second major aim of our work has been to prove that low levels of SORLA in mouse models promote amyloidogenic processes (as proposed for AD patients).


Fig. 5.


To do so, we produced mice genetically deficient for Sorl1 and crossed them with several human APP transgenic lines commonly used as murine AD models. In support of our hypothesis, loss of SORLA significantly increased APP processing, Aβ production, and senile plaque deposition (Rohe et al., JBC, 2008; Dodson et al., J Neurosci, 2008).


SORLA is a sorting receptor for APP. Another major achievement has been the elucidation of the molecular mechanism of SORLA function in APP processing. Here, we focused on studies in cell lines and primary neurons expressing wild-type or trafficking mutants of the receptor. In brief, we demonstrated that SORLA acts as sorting protein interacting with APP via its luminal (Andersen et al., Biochemistry, 2006) and cytoplasmic domains (Spoelgen et al., J Neurosci, 2006). Interaction sequesters APP in the trans-Golgi network (TGN) preventing transport into cell compartments were processing by secretases occurs (Fig. 5). SORLA-dependent trafficking of APP requires the cytosolic adaptors PACS1, GGA, and retromer that sort SORLA between TGN and endosomes (Schmidt et al., JBC, 2007; Spoelgen et al., Neurosci, 2009; Fjorback et al., J Neurosci, 2012).


Therapeutic potential of SORLA.  So far, the role of SORLA (and of other modifiers) in AD has mainly been explored in transgenic cell lines and knockout mouse models. However, experimental systems with overexpression or total lack of modifier activity do not reflect the situation in patients where subtle alterations may affect neuropathology over the course of a lifetime. To test the clinical relevance of modest reductions in SORLA (as in individuals with SORL1 risk allele), we developed a tet-regulatable cell system to vary the ratio of APP and SORLA over a continuous concentration range. In a systems biology approach, we combined quantitative biochemical studies with mathematical modeling to establish a kinetic model of amyloidogenic processing, and to evaluate the influence by SORLA (Schmidt et al., EMBO J, 2011). Contrary to previous hypotheses, our studies demonstrated that secretases represent allosteric enzymes that require cooperativity by APP oligomerization for efficient processing. Cooperativity enables swift adaptive changes in activity with even small alterations in APP levels.

Fig. 6.


We also uncovered that SORLA impairs APP oligomerization, eliminating the preferred form of the substrate. Finally, we showed that a mere 2-fold reduction in receptor levels translates into 40% increased amyloidogenic processing, similar to what is seen in AD patients. Our data represent the first mathematical description of the contribution of genetic risk factors to AD, substantiating the relevance of subtle changes in SORLA levels for the disease.


We also explored the possibility of raising SORLA levels as therapeutic approach in AD. To do so, we aimed at identifying factors that control Sorl1 transcription in vivo. Using screening approaches in primary neurons, we identified brain-derived neurotrophic factor (BDNF) as inducer of Sorl1 that activates gene transcription 10-fold via the ERK pathway. In line with a physiological role as regulator of Sorl1, expression of SORLA was impaired in mouse models with genetic (Bdnf-/-) or disease-related loss of BDNF activity (Huntington’s disease). Intracranial application of BDNF reduced Aβ levels in wild-type but not in SORLA-deficient mice (Fig. 6) validating the therapeutic potential of increasing brain SORLA levels in AD (Rohe et al., J Neurosci, 2010).