Individuals with FXS tend to exhibit increased neural responses reflected by increased ERP amplitudes to auditory stimuli compared to typically developed individuals, likely due to neural hyperexcitability (Ethridge et al. 2016, 2019; Tempio et al. 2023). The original results from the phase 2a clinical trial suggested BPN14770 reduced the absolute N1 ERP component amplitude to the first stimulus in a habituation pair. Given the N1 is a negative-going ERP component, the positive correlation presented in the current study indicates a decrease in negative amplitude with BPN14770, which is an improvement that corresponds to a decreased neural response to the stimulus. Figure 2A demonstrates the expanded negative range of ERP values for the placebo group, highlighting the group-level shift in ERP amplitudes toward smaller values (i.e., range reduction), as well as the relationship between BPN14770 serum concentration and decreases in N1 amplitude. The individual with the largest plasma concentration of BPN14770 shows a nearly zero-amplitude value for N1, suggesting a potential upper limit to effective dosing for this individual for maintaining the N1 amplitude within typical levels. Additionally, it is important to note that, with the exception of two individuals, the majority of individuals in the placebo group showed change from baseline values close to zero (Fig. 2B). Despite the non-significant correlation between N1 amplitude change from baseline and PK values, the baseline adjusted BPN14770 values appear to maintain a similar trend to that noted in Fig. 2A. Baseline ERP values introduce another level of variability, notable in Fig. 2C, which reduces statistical power. Although absolute N1 amplitudes at a group level tends to follow known patterns in FXS (e.g. higher N1 values overall (Ethridge et al. 2019), and reduced N1 values at the group level from baseline and placebo for BPN14770 (Berry-Kravis et al. 2021) absolute amplitude at the individual level and timepoint can be influenced by a number of factors, such as impedances at the scalp, cap placement, and number of clean trials per run. This small increase in potential error variability compounds when adding additional timepoints to compute change values, which can reduce the power of the effect in small sample size studies such as these. Both findings highlight the necessity of a replication with Phase 3 data.

Despite clear limitations, the current results raise confidence in the validity of the original, marginal ERP amplitude group-level result in favor of BPN14770 improvements. Confirmation of prior results lends support to the conclusion that BPN14770 reduces neural hyperexcitability during stimulus processing. Variations in the auditory N1 ERP have been associated with language and communication as well as sensory outcomes in FXS (Ethridge et al. 2016, 2019; Schmitt et al. 2020) the results have implications for language processing and may underlie the noted improvements in clinical outcomes from the original study (Berry-Kravis et al. 2021).

Limitations included the small sample size consistent with Phase 2 studies, expected data loss in EEG data due to movement which further reduced statistical power, and carry-over effects in placebo measures from period 2. While the correlation supports the original findings, and suggests that the effect was present but underpowered, both post-period 1 analyses within the BPN14770 group reflect only 12 individuals. Nevertheless, this study represents the first report of significant correlation between a reliably abnormal EEG marker and serum concentration of a novel pharmaceutical in FXS. The limited sample size may inflate the effect of sources of variability. Specifically, differences in cAMP levels may account for placebo variability but cAMP levels were not assayed for this study as BPN14770 inhibits PDE4D, a predominantly brain specific enzyme (with a goal of avoiding the peripheral side effects of this class of drugs) and, therefore, it was assumed that level in blood after drug initiation would not be a useful marker of activity. Thus-we cannot control for cAMP levels to determine target engagement. The most evident limitation is the statistically non-significant relationship between baseline adjusted N1 values and PK but, again, those differences could reflect natural variability in the baseline ERP values due to non-neural sources. The N1 amplitude will be assessed as part of the ongoing phase 3 trial testing BPN14770 in a larger sample which will provide a sufficiently powered statistical model for assessing N1 amplitude improvement with BPN14770 treatment and the relationship it may share with serum BPN14770 concentrations.