Elevated Plasma microRNA-206 Levels Predict Cognitive Decline and Progression to Dementia from Mild Cognitive Impairment

Round 1

Reviewer 1 Report

This is an interesting study that seeks to investigate the utility of measuring miRNAs in plasma as prognostic biomarkers for MCI and AD-dementia.  The first part of the manuscript reports the initial identification of miRNAs found to be at different levels across three subject groups (control, MCI and AD).  The authors then go on to examine relevant miRs in the context of predicting cognitive decline and further examination of microRNA-206 from serial samples collected approximately every two years.  While the idea of testing for and using microRNAs as biomarkers for AD is not novel per se, strengths of the manuscript include the longitudinal design, and novelty arises from use of the microfluidic device, which allowed the authors to reproduce their original RT-PCR findings, thus suggesting the possibility of a future practice cost-effective tool for microRNA analysis of blood. 

There were two main concerns with the manuscript:  Firstly the AD samples (n=25) were obtained from a biobank, while the control and MCI sample from the Vallecas project.  It is possible that differences in plasma storage conditions and the timeframe of storage could impact miR levels, resulting in the observed differences.  This could be tested via measuring e.g. microRNA-206 in additional control and MCI samples from the biobank. At the very least the storage conditions etc., should be described and discussed. 

Secondly, the longitudinal timeframe may have been insufficient.  The most promising results are derived from the experiment measuring baseline miR-206 and using this as a predictor for those likely to develop dementia, but at this point we do not know if those subjects really did develop dementia or not. Additionally, there was a concern that because the baseline assay of miR-206 (Fig 4b) did not differ between the first two groups, the biomarker assay seems limited to only being able to determine which MCI subjects will develop dementia in a very short time-frame (4 years).  For true utility as a biomarker, use of additional groups would have been extremely beneficial (Con-Con-Con and possibly even Con-Con-MCI). Without comparison to a group that did not develop MCI or dementia, the potential clinical value of this study may be limited. 

There were also a number of areas where additional clarity would strengthen the paper:

The n used for many of the analysis are unclear. In the initial discovery experiments n=10 were used for each group.  However, the validation qRT-PCR experiments (e.g. F1) do not state the numbers employed.  Every figure in the manuscript or the accompanying text should clearly indicate the sample sized employed for each experiment.    The MCI discovery cohort started with n=30, and then 11 subjects were lost to follow-up. However, only 17 are reported in table 2.  Statistical Analysis: Each 3-group analysis (K-wallis test) should document the p-value of the overall test, then the p-value for the appropriately adjusted post-hoc test.  Additionally, the methods state that outliers were removed.  There are always concerns with removing datapoints from a small sample. A better approach would be to at least show the outliers in the figures and describe that difference in results following the removal of outliers.  On a similar subject, in Fig 1D four points for each mIR were outside the IQR: were these the same samples for each miR, and does this point to some experimental error or concern with that datapoint/ plasma sample?

Author Response

We would like to thank the reviewer for the overall positive and constructive comments. Please find below our detailed point-by-point response to the concerns raised.

 

Main concerns: 

 

1) Firstly the AD samples (n=25) were obtained from a biobank, while the control and MCI sample from the Vallecas project.  It is possible that differences in plasma storage conditions and the timeframe of storage could impact miR levels, resulting in the observed differences.  This could be tested via measuring e.g. microRNA-206 in additional control and MCI samples from the biobank. At the very least the storage conditions etc., should be described and discussed. 

 

The reviewer is correct that AD samples are from a biobank, while the control and MCI samples are from the Vallecas project cohort and we appreciate that we were not sufficiently clear describing the collection and storage of the samples used in our study in the Material and Methods section. However, both biobank and Vallecas project samples are collected at the same Center (CIEN Foundation), by the same group of technicians and using the same protocol (see below), thus assuring that sample processing is comparable. All samples are stored under the same conditions following identical protocols. Timeframes of storage from controls have been selected to match the timeframes of AD samples.

We have elaborated upon sample collection and storage to remove any confusion or concern over this issue (See Methods section 2.1 and 2.2). Please also see attached protocol.

Line 115-117: “Collections of biological samples and clinical data were all performed at CIEN Foundation (Madrid, Spain). Retrospective samples from AD subjects (n = 25) were obtained from the onsite BT-CIEN biobank located at the same Centre [22].”

Line 129-133: “All blood samples (Control, MCI and AD) were collected at the same Center (CIEN Foundation), by the same group of technicians and using the same protocol [21], thus assuring that sample processing is comparable. All samples are stored under the same conditions following identical protocols. Timeframes of storage from controls have been selected to match the timeframes of AD samples.

 

2) The longitudinal timeframe may have been insufficient.  The most promising results are derived from the experiment measuring baseline miR-206 and using this as a predictor for those likely to develop dementia, but at this point we do not know if those subjects really did develop dementia or not.

We agree with the reviewer that our study is limited by the relatively short timeframe of sample analysis. We cannot rule-out that some stable controls will develop MCI or dementia within the following years and it is true we do not have sufficient number of subjects who have progressed to dementia to make any definitive statement on whether miR-206 is a good predictor for dementia. It did though enable us to identify miR-206 relationship with cognitive decline which we then applied within our longitudinal samples which gave strong results to indicate that miR-206 becomes elevated prior to the onset of dementia.

This has been highlighted in our new discussion and, in line with this, we have kept the general conclusions of the paper reserved, discussing miR-206 much greater focus on its relationship to cognitive decline in MCI than the development of dementia.

 

Line 437-438: “In addition, only a limited number of subjects in the discovery cohort progressed to dementia from MCI so far over the 4 years (n = 3), representing a potential limitation for a definite diagnosis of dementia and Alzheimer’s disease. In the absence of this, neuropsychological indicators collected over the duration of the study were used in each case-series to assess the likelihood of progression to dementia and identify potential pre-symptomatic AD. ΔMMSE over the 4 years was employed to evaluate the cognitive decline symptomatic of prodromal AD [43-45]. Indicating that higher levels of miR-206 were predictive for cognitive decline and onset of dementia in MCI subjects, miR-206 levels correlated strongly to declining MMSE scores over time. Age adjusted FCSRT was employed as an evaluation of episodic and working memory (earliest areas of cognition damaged by AD [46]) and an indicator for likelihood of dementia onset [25]. MiR-206 was also elevated in MCI subjects below the FCSRT cut-off indicating that subjects with increased miR-206 were at high risk of developing dementia.”

 

 

 

3) Additionally, there was a concern that because the baseline assay of miR-206 (Fig 4b) did not differ between the first two groups, the biomarker assay seems limited to only being able to determine which MCI subjects will develop dementia in a very short time-frame (4 years).  For true utility as a biomarker, use of additional groups would have been extremely beneficial (Con-Con-Con and possibly even Con-Con-MCI). Without comparison to a group that did not develop MCI or dementia, the potential clinical value of this study may be limited. 

 

We believe the investigation of these groups will have limited value as all results have shown miR-206 to not be predictive for the progression to the MCI state from a control state even with an eventual progression to dementia with no difference between any of the groups regardless of disease progression.

 

 

4) The n used for many of the analysis are unclear. In the initial discovery experiments n=10 were used for each group.  However, the validation qRT-PCR experiments (e.g. F1) do not state the numbers employed.  Every figure in the manuscript or the accompanying text should clearly indicate the sample sized employed for each experiment.   The MCI discovery cohort started with n=30, and then 11 subjects were lost to follow-up. However, only 17 are reported in table 2.  Statistical Analysis: Each 3-group analysis (K-wallis test) should document the p-value of the overall test, then the p-value for the appropriately adjusted post-hoc test. 

 

We have now included n numbers of subjects with all statistical tests in the Figure Legends to clarify the subjects tested in each group (see Figure legends).

 

Regarding the discrepancy of sample number reported in the table, this was due to the addtional 2 subjects which were removed as outliers as per Grubb’s test on the discovery cohort along with subject removed due to follow-up.

 

Regarding the Kruskal-Wallis, no post-hoc test was used in the study. For all the datasets tested by Kruskal-Wallis (Table 1) the statistical question being asked was whether there was significant difference between the three groups to highlight covariants which could influence the result and not to identify a specific significant differences between two groups.

 

 

5) Additionally, the methods state that outliers were removed.  There are always concerns with removing datapoints from a small sample. A better approach would be to at least show the outliers in the figures and describe that difference in results following the removal of outliers. 

We acknowledge the issues with removing theses outliers and we have included a Supplementary Figure with outliers included in the dataset and referenced within the Results sections (Supplementary Material Figure 1B). As can be clearly seen, the values identified as outliers by Grubb’s Test heavily skew the results in all 3 of the microRNAs.

 

 

6) On a similar subject, in Fig 1D four points for each mIR were outside the IQR: were these the same samples for each miR, and does this point to some experimental error or concern with that datapoint/ plasma sample?

Only one sample is consistenly outside the IQR in all three miRNAs in MCI (sample#: VK-599). The other points identifed in Fig 1D by the reviewer were not the same samples for each miRNA. We have no concerns with regards experimental error or sample quality with any of the other datapoints. It would be tempting to remove the VK-599 from the dataset but there is not a sufficient evidence that it’s upregualtion across 3 miRNAs is a result of experimental error and not a biological coincidence (In addition all significant variances between groups remain after removal of VK-599).  

Author Response File: Author Response.pdf

Reviewer 2 Report

This is a nicely written manuscript correlating serum levels of the miRNA miRNA-206 is persons with MCI to correlations with development of Alzheimer's. It is well-written and easy to read and comprehend. The authors find a correlation and demonstrate a non-invasive and easily implementable method for administering this test. It would be nice to see a follow-up on a larger cohort, however, these findings are of strong interest to the AD field and are a nice biomarker study. It is worthy of publication.

Author Response

We would like to thank the reviewer for the positive comments.