Alzheimer"s disease (AD) is a heterogeneous neurodegenerative disorder and detailed understanding of its pathophysiologic processes at earlier phases is crucial for effective treatment development and prevention. In the present study, we investigated within the Cognitive Complaints Cohort (CCC) the changes in the CSF proteome of patients with Mild Cognitive Impairment (MCI) fulfilling criteria of high likelihood for AD (n=45), compared to MCI patients without any biomarkers of Aβ deposition or neuronal injury (n=23). This research strategy intended to control for nonspecific changes that might influence the proteome and thus sort out the alterations directly related to AD pathogenesis. CSF proteomics could discriminate patients with MCI due to AD from patients with MCI without biomarkers of Aβ deposition or neurodegeneration. In a second step, we looked at the existence of putative subtypes in patients with MCI due to AD using a clustering non-negative matrix factorization method. Results were validated by applying the same methodology on MCI patients with (n=92) and without (n=102) AD pathology from the European Medical Information Framework for Alzheimer"s Disease (EMIF-AD) cohort. Using the 164 and 170 selected proteins in the CCC and EMIF-AD cohorts, respectively, that differed significantly between patients with MCI due to AD and patients with non-degenerative MCI, two consistent clusters of AD patients at the MCI stage were found. Twenty (47.6%) patients were labelled in Cluster 1 and 22 (52.4%) were labelled in Cluster 2 in the CCC cohort, and 46 (50%) participants were labeled in Cluster 1 and 46 (50%) labeled in Cluster 2 in the EMIF-AD cohort. Distinct Gene Ontology profiles were found for the two clusters. Cluster 1 showed increased levels of proteins related to biological processes such as the immune system, acute inflammatory response, coagulation and lipid metabolism, whereas Cluster 2 showed involvement of neuronal and neurodevelopmental processes. There was a statistically significant concordance between the two cohorts regarding the contribution of different gene ontologies to Cluster 1 and Cluster 2 (Cohen"s kappa= 0.398, p<0.0001). Cluster 2 had higher levels of CSF tTau, CSF pTau and CSF Aβ42 levels as compared to Cluster 1. The results show that, in spite of slightly distinct diagnostic criteria and different mass spectrometry approaches used in the two cohorts, two consistent subtypes of AD patients emerge that could be described as (i) immune and blood-brain barrier dysfunction (Cluster 1) and (ii) hyperplasticity (Cluster 2), as previously proposed. These findings might be important for the design and interpretation of clinical trials since there might be an association between response to the therapy and the AD subtype to which the patients belong.