Randomized Controlled Trial of the Glycine Transporter 1 Inhibitor PF-03463275 to Enhance Cognitive Training and Neuroplasticity in Schizophrenia

Abstract

N-methyl-D-aspartate glutamate receptor (NMDAR) hypofunction is implicated in the impaired neuroplasticity and cognitive impairments associated with schizophrenia (CIAS). We hypothesized that enhancing NMDAR function by inhibiting the glycine transporter-1 (GLYT1) would improve neuroplasticity and thereby augment benefits of non-pharmacological cognitive training (CT) strategies. This study examined whether co-administration of a GLYT1 inhibitor and computerized CT would have synergistic effects on CIAS. Stable outpatients with schizophrenia participated in this double-blind, placebo-controlled, within-subject, crossover augmentation study. Participants received placebo or GLYT1 inhibitor (PF-03463275), for two 5-week periods separated by 2 weeks of washout. PF-03463275 doses (40 or 60 mg twice daily) were selected to produce high GLYT1 occupancy. To limit pharmacodynamic variability, only cytochrome P450 2D6 extensive metabolizers were included. Medication adherence was confirmed daily. Participants received 4 weeks of CT in each treatment period. Cognitive performance (MATRICS Consensus Cognitive Battery) and psychotic symptoms (Positive and Negative Syndrome Scale) were assessed in each period. 71 participants were randomized. PF-03463275 in combination with CT was feasible, safe, and well-tolerated at the doses prescribed but did not produce greater improvement in CIAS compared to CT alone. PF-03463275 was not associated with improved CT learning parameters. Participation in CT was associated with improvement in MCCB scores.

1. Introduction:

Cognitive impairments associated with schizophrenia (CIAS) encompass deficits in memory, attention, executive functioning, vocabulary, visuospatial skills, and learning (Heinrichs and Zakzanis, 1998), represent a significant decrement from premorbid ability (Keefe et al., 2005) and affect functional status more than schizophrenia’s core disease-defining symptoms (Bowie et al., 2006; Green et al., 2000; Green et al., 2004; Tsang et al., 2010). Antipsychotic drugs that block dopamine (DA) D₂ receptors help manage psychosis (Kapur and Mamo, 2003) but CIAS persist (Buchanan et al., 2007a; Buchanan et al., 2007b; Goldberg et al., 1993; Harvey et al., 2001; Keefe et al., 2007). Non-DA pharmacological targets have been explored for the treatment of CIAS including histaminic (H₃ antagonism), cholinergic (Alpha₇ nicotinic receptor agonism), and N-methyl-D-aspartate receptor (NMDAR agonism via glycine modulatory site) neurotransmitter systems. However, these endeavors have yet to elucidate effective treatments (Girgis et al., 2019), and CIAS remain an important area of unmet medical need for drug development (Yang et al., 2017).

Arguably, the only existing evidence-based treatment for CIAS is cognitive training (CT), or targeted exercises of memory, information processing, attention, and executive functioning, but trials of CT have found modest efficacy (Lejeune et al., 2021; Vita et al., 2021). Well-established deficits in learning and neuroplasticity (Keshavan et al., 2015) could impede the acquisition, or generalization, of trained skills to broader cognitive and functional domains. Indeed, individual response to CT has been associated with baseline differences in behavioral indices of learning potential (Davidson et al., 2016) and in a neurophysiologic metric associated with information processing capacity (Castelluccio et al., 2020).

Pharmacologic and neurostimulation (Jahshan et al., 2017; Jahshan et al., 2020; Krystal et al., 2003) approaches to enhance neuroplasticity have been examined as ways to augment response to CT. Work in this area has focused on increasing NMDAR activity via its high-affinity glycine co-agonist site (glycine_B). While enhancing NMDA signaling is a promising strategy, efforts to do so directly by glycine or D-serine intake may be limited by poor brain bioavailability of these drugs. For example, whereas levels of D-serine have been associated with CT response, adjunctive D-serine does not appear to enhance response to CT (D'Souza et al., 2013; Panizzutti et al., 2019).

An alternative strategy is to boost endogenous glycine levels by administering a glycine transporter-1 inhibitor (GLYT1-I). GLYT1-I’s raise glycine levels in hippocampal slice preparations (Harsing and Matyus, 2013) and in human cerebrospinal fluid (CSF) (Hofmann et al., 2016), and, in turn, increase glycine occupancy at the glycine_B co-agonist site of NMDAR (Bergeron et al., 1998). In a preliminary signal-finding study, our group evaluated dose-related effects of the GLYT1-I PF-03463275 following 1 week of daily administration in healthy adults and patients with schizophrenia (D'Souza et al., 2018). Endpoints included Positron Emission Tomography (PET) measures of receptor occupancy, working memory task-based fMRI, and an electroencephalographic (EEG) biomarker of neuroplasticity based on long-term potentiation (LTP) of visual evoked potentials (VEP). No appreciable effects were observed in healthy participants; patients, showed a dose-related enhancement of LTP with peak effects at 40 mg BID. More recent evidence suggest GLYT1-I may also have therapeutic effects on CIAS (Fleischhacker et al., 2021).

Collectively, these studies suggest that GLYT1 is a potential pharmacologic mechanism for intervention in CIAS. To test this proposal, the current study paired GLYT1 administration with a computer-based CT intervention, providing a standardized curriculum over which task-based learning and generalization to untrained cognitive endpoints could be evaluated. Based on our previous dose-finding study, we chose the most efficacious and highest doses of PF-03463275 (40mg BID and 60mg BID respectively) to achieve optimal GLYT1-I occupancy and augmentation of CT.

2. Methods and Materials:

2.1. Study Design:

(Figure 1) In a double-blind, placebo-controlled, crossover study, stable antipsychotic-treated outpatients with schizophrenia were randomized to either placebo or active PF-03463275 twice daily in addition to their antipsychotic for 2 dosing periods, each lasting approximately 5 weeks. The first week lead-in of each treatment period was to evaluate the neurophysiologic effects of medication alone (i.e., without CT) and to confirm medication adherence and tolerability. After the week-long lead-in period, subjects received CT in addition to placebo/active PF-03463275 for 4 weeks. After completing the first treatment period, there was a 2-week washout period to ensure clearance of the study drug which has a half-life of 3-6 hours (Buchy et al., 2015)(Buchy et al., 2015)(Buchy et al., 2015), before beginning the second treatment period, which was identical to the first, except that subjects who received active drug received placebo and vice-versa (Table S1).

Figure 1:

Study Design Overview. Participants were randomized to one of 3 sequences in a 1:1:1 ratio: A) PF-03463275 60 mg BID in the first treatment period followed by placebo in the second treatment period, B) placebo in the first treatment period followed by either (50%) PF-03463275 40 or (50%) 60 mg in the second treatment period, or C) PF-03463275 40 mg BID first followed by placebo.

All participants received placebo and only 1 of 2 active conditions in the study; thus, each active drug condition had about half the number of participants as the placebo condition.

In the first week of each study period, participants received study drug only and no cognitive training.

Cognitive training began after 1 week of study drug exposure in each treatment period.

MCCB was collected before (pre-drug) and after (post-drug) 5 weeks of treatment with study drug in each period. PANSS was administered before and twice during each 5-week treatment period.

2.2. Regulatory:

This study was conducted with approvals from the Institutional Review Boards of the VA Connecticut Healthcare System and Yale University School of Medicine, under an approved IND, (IND #118,880), and registered on clinicaltrials.gov (NCT01911676).

2.3. Screening / Inclusion/Exclusion criteria (See Supplemental Material):

Adults 20 to 65 years old: 1) with a DSM-IV schizophrenia or schizoaffective disorder; 2) on maintenance antipsychotic (except clozapine); 3) clinically stable (no antipsychotic changes or psychiatric hospitalizations in the preceding 3 months); 4) English speaking; 5) with capacity to provide written consent; and 6) were CYP2D6 extensive metabolizers, were included in the study. PF-03463275 is primarily metabolized by CYP2D6, so, to prevent large differences in CYP2D6 plasma levels, only participants who were extensive metabolizers were included, whereas those with CYP2D6 genotypes for ultra-rapid, intermediate, and poor metabolism were excluded.

2.4. Recruitment:

Participants were recruited by advertisements, flyers, referral from providers, and word of mouth. Research staff explained study details, risks, and procedures before obtaining written informed consent.

2.5. Randomization:

Stratified block randomization, with block size of 6 and stratification by Intelligence Quotient (WTAR estimated FSIQ 71-90 and >91), was used to assign subjects to one of 4 treatment sequences across two 5-week treatment phases that were separated by a 2-week washout phase (Figure 1 and Table S1): PF-03463275 60 mg BID followed by placebo, PF-03463275 40 mg BID followed by placebo, and placebo followed by either PF-03463275 40 or 60 mg BID. The study pharmacist generated the randomization sequence and prepared the study medications; all other study staff and participants were blinded to the condition.

2.6. PF-04363275:

PF-04363275 was provided by Pfizer Inc. (Groton, CT). Study medication adherence was visually confirmed by videoconferencing twice daily on weekdays using cellphone assisted remote observation of medication adherence (CAROMA) (DeWorsop et al., 2016). The dose of PF-03463275 for the clinical trial was selected based on the results of convergent approaches conducted in the earlier phase of the study: 1) receptor occupancy (PET); 2) reversal of ketamine-induced impairments in working memory (fMRI); and 3) augmentation of visually-evoked LTP. We found 60 mg BID produced the highest receptor occupancy, but that target engagement signal was stronger at 40 mg BID (D'Souza et al., 2018).

2.7. Cognitive Training:

All participants attended computer-based cognitive training appointments in a supervised clinic over weeks 2-5 of each treatment period. A training curriculum was developed using the HAPPYneuron program (detailed in Table S2 and Supplemental Methods). HAPPYneuron has been effective in prior studies of schizophrenia (Franck et al., 2013; Vianin et al., 2014). Participants were encouraged to complete 5 sessions of weekly training over a total of 8 weeks and received incentive payments in increasing intervals at the completion of each 5^th session.

2.8. Outcome Measures:

2.8.1. Cognitive Outcome

The primary outcome measure was change in the MATRICS Consensus Cognitive Battery (MCCB) composite score. The MCCB was administered before the start and at the end of each of the two periods (Figure 1). The MCCB composite score and a calculated cognitive composite, which omitted the social cognition test, were calculated to assess the overall effect of treatment, with changes in specific cognitive domains including verbal and visual learning and working memory examined as exploratory analyses.

2.8.2. PANSS:

The Positive and Negative Syndrome Scale (PANSS) (Kay et al., 1989) is a 30-item clinician-rated assessment of symptoms of schizophrenia based on a semi-structured interview with subscales for positive, negative and general symptoms. The same group of raters administered PANSS for all participants thrice in each period: before randomization and during the second and fourth weeks of each period.

2.8.3. Participant Self-Assessments:

Subjective assessment of cognitive and perceptual experience was measured using the Patient’s Assessment of Own Functioning (PAOFI) (Chelune et al., 1986) while self-assessment of attentional processes and perceptual processes was measured using the Sensory Gating Inventory (SGI) (Hetrick et al., 2012).

2.9. Statistical Analyses:

This report is restricted to the cognitive and symptom measures. EEG biomarkers were assessed at Days 1 and 7 of each period, replicating our prior study design (D'Souza et al., 2018), and will be reported elsewhere. Target sample size was 30 per sequence for a total of 90, based on power calculations (Supplementary Methods). All dependent measures were assessed for normality using normal probability plots. Descriptive statistics were calculated prior to analyses. Each dependent variable (MCCB overall and cognitive composite scores, PANSS total and 3 PANSS subscales), was analyzed using a mixed effects model with dose (placebo, 40 mg BID, 60 mg BID), visit, their interaction, and period (first or second) as main effect. The time points were pre- versus post-drug for MCCB, and pre-drug and two post-drug assessments at two and five weeks for PANSS. Interactions involving period were also assessed but dropped from the models for parsimony. Subject was the clustering factor. The best-fitting variance-covariance structure for the repeated measures in each model was selected based on Schwartz’ Bayesian Criterion. Least square means and standard errors were calculated to describe the patterns of means for each outcome. Post-hoc mean comparisons were tested to explain significant effects in the models. For each outcome, we also fit models with drug as a two-level factor (i.e., placebo vs. 40mg and 60mg doses combined). Exploratory analyses (see Supplement for details) included repeated measures ANOVA (rmANOVA) to examine: 1) effects of time, period, and dose on PAOFI and SGI; 2) effects of dose on performance in CT exercises.

3. Results:

71 participants were randomized as shown in Figure 2. Demographics of participants in each of the four randomization assignments are shown in Table 1, and each of the four randomized groups were well-matched for age, education, and IQ.

Figure 2:

Consort Diagram. Of the 71 participants randomized, 6 were terminated by the PI (4 for substance misuse, 1 for abnormal ECG and 1 for wrist drop), 10 withdrew (5 no longer wanted to take study medication, 2 wanted to look for job, 1 had conflict with study activities and other responsibilities, 1 had housing problems) and 1 withdrew and was lost to follow-up. There were 55 completers who did all study activities except follow up.

Table 1:

Participant Demographics:

Abbreviations: SD = standard deviation; WTAR = Wechsler Test of Adult Reading.

Randomization Assignment	Period 1: 40 mg Period 2: Placebo	Period 1: Placebo Period 2: 40 mg	Period 1: Placebo Period 2: 60 mg	Period 1: 60 mg Period 2: Placebo
n	22	12	13	24
Average age in years (SD)	46.1 (11.4)	50.9 (12.4)	40.9 (14.1)	48.8 (11.7)
Number male (%)	20 (90.9%)	7 (58.3%)	9 (69.2%)	21 (87.5%)
Education, years (SD)	12.1 (1.9)	12.3 (2.7)	12.5 (2.8)	12.4 (2.4)
IQ from WTAR (SD)	92.2 (10.5)	96.9 (10.6)	96.0 (10.8)	94.4 (11.0)
Parent education, years (SD)	12.5 (3.1)	11.4 (2.4)	13.4 (3.2)	12.1 (3.6)
African American (%)	16 (72.7%)	5 (41.7%)	6 (46.1%)	15 (62.5%)
Caucasian (%)	6 (27.3%)	7 (58.3%)	5 (38.5%)	7 (29.2%)
Hispanic Ethnicity (%)	2 (9.1%)	3 (25%)	2 (15.4%)	2 (8.3%)

3.1. Safety:

There were no serious adverse events. The rates of non-serious adverse events were low, and there were no differences in adverse event rates between placebo and PF-03463275 (Table S3).

3.2. Cognition:

MCCB MATRICS overall composite and cognitive composite scores, tested either as three independent conditions (PF-04363275 40 mg, PF-04363275 60 mg, and placebo) or with active conditions collapsed in comparison to placebo, increased with time; there were significant main effects of pre- versus post-drug visit (Figures 3 and S1), and of period 1 (before cross over) versus period 2 (after cross over), but no significant study drug condition * visit interaction (Tables S4 and S5). Exploratory analyses of all 10 MCCB subtests revealed no significant visit by study drug condition interactions.

Figure 3:

MCCB overall composite score for all participants collected before (pre-drug) and after (post-drug) 5 weeks of treatment with study drug in each period. All participants were randomized to receive placebo and only 1 of 2 active conditions (PF-03463275 40 or 60 mg) in the study; thus, the active drug conditions each have about half the participants as the placebo condition. X-axis represents dose, and y-axis represents MCCB overall composite score. Error bars represent SEM.

3.3. PANSS

PANSS assessments were performed at three visits in each period, pre-drug, after 2 weeks of study drug, and after 5 weeks of study drug (Figure 1). PANSS total scores and positive, negative, and general subscales decreased over time, with significant main effects of both visit and period, however the visit * study drug condition interactions were not significant for any of the four measures (Figures S2-S5), regardless of whether active doses were treated independently (Tablet S6) or combined (Table S7). A main effect of condition was detected for PANSS total in comparison of combined active conditions to placebo, and for PANSS positive scores tested using independent or combined active conditions (Tables S6 and S7). Condition effects appeared driven by the pre-drug PANSS total and positive scores being higher in active drug than placebo (Figures S2 and S3).

3.4. Exploratory Analyses:

RmANOVA’s revealed no significant main effect of pre- versus post-drug visit, period, or study drug condition on PAOFI or SGI (Supplementary Figures S6, S7), in contrast to other cognitive and clinical measures that improved with time. There was no effect of dose on two different measures of CT performance (Supplement and Figures S8-S9).

4. Discussion:

We tested the hypothesis that pharmacological enhancement of neuroplasticity with the GLYT1-I (PF-03463275) combined with CT would have synergistic effects on CIAS. PF-03463275 treatment, at doses demonstrated to engage the target, was safe and well-tolerated but did not produce a greater improvement in CIAS compared to CT alone. MCCB scores did, however, improve over time regardless of drug condition (active or placebo), as demonstrated by an improvement in period 2 over period 1 and pre- versus post-drug dosing. This improvement in MCCB performance was of effect size (Hedge’s g=0.31) comparable to prior studies of CT in patients with schizophrenia without a combined pharmacologic intervention; Hedge’s g ranging between 0.19-0.33 and Cohen’s d=0.45 (Lejeune et al., 2021; Wykes et al., 2011).

Clinical trials of NMDAR co-agonists (glycine, D-serine, D-cycloserine and D-alanine) and GLYT1-I’s to improve CIAS have produced mixed results. Glycine, low-dose D-serine, and D-cycloserine did not improve cognition in large studies (Buchanan et al., 2007b; Weiser et al., 2012). GLYT1-I’s bitopertin, sarcosine and ORG-25935 did not improve CIAS (Kantrowitz et al., 2017; Lin et al., 2017; Schoemaker et al., 2014). In contrast, in a recent placebo-controlled randomized phase 2 trial, BI 425809 without cognitive remediation improved cognition after 12 weeks of treatment in patients with schizophrenia (Fleischhacker et al., 2021). Two studies of NMDA receptor co-agonists as augmentation to CT have also had conflicting results: D-serine did not augment the effects of CT (D'Souza et al., 2013), and D-cycloserine was associated with improved performance on the trained auditory discrimination task but not with generalized improvements in cognitive outcomes (Cain et al., 2014). Evaluated in the present study, GLYT1-I did not appear to improve learning or peak performance of CT exercises.

PF-03463275 also failed to treat positive or negative symptoms, despite promising improvements in negative symptoms with GLYT1-I’s sarcosine in both chronic and antipsychotic-naïve acutely symptomatic patients (Lane et al., 2005; Lane et al., 2010; Tsai et al., 2004), and bitopertin (Umbricht et al., 2014), the results were not replicated in later studies of sarcosine (Lin et al., 2017; Strzelecki et al., 2015), phase 3 studies of bitopertin (Bugarski-Kirola et al., 2017), or in large studies of GLYT1-I’s AMG 747 and ORG-25935 (Dunayevich et al., 2017; Schoemaker et al., 2014). Though small earlier studies with NMDAR co-agonists suggested improvements in positive and negative symptoms (Heresco-Levy et al., 2004; Heresco-Levy et al., 1996; Heresco-Levy et al., 1999; Javitt et al., 1994; Tsai et al., 2006), large studies of NMDAR co-agonists glycine, D-cycloserine, and low-dose D-serine failed to treat negative symptoms (Buchanan et al., 2007b; Weiser et al., 2012). NMDAR co-agonists might be helpful in combination with behavioral therapy or in specific patients. D-cycloserine enhanced the effects of cognitive behavioral therapy for delusions (Gottlieb et al., 2011). D-serine at higher doses (60 mg/day) improved negative symptoms in those at clinical high risk for psychosis (Kantrowitz et al., 2015) and in schizophrenia in manner that correlated with improvements in the electrophysiological biomarker auditory mismatch negativity (MMN) (Kantrowitz et al., 2018). Glycine treatment in a small study improved MMN and negative symptoms, and there was suggestion of correlation between the two (Greenwood et al., 2018). These studies support target engagement of NMDAR measured with amplified evoked potentials and symptoms.

Reliable biomarkers of NMDAR-mediated neuroplasticity have been difficult to identify. While in our prior study LTP improved with the 40 mg dose of PF-03463275 (D'Souza et al., 2018), D-cycloserine failed to increase LTP (Forsyth et al., 2017). Interpretation of these results is complicated by inconsistent detection of the visual LTP deficit in schizophrenia (Wynn et al., 2019), and recent studies that challenge whether high frequency visual stimulation reliably results in potentiated VEP in healthy people (Dias et al., 2022). Auditory MMN is another proposed EEG measure of NMDAR-mediated synaptic plasticity and has been used to detect target engagement of NMDAR modulating compounds in other studies. However, although the MMN deficit in schizophrenia is robust and found repeatedly (Erickson et al., 2016), NMDAR modulators have had inconsistent effects on MMN (Kantrowitz et al., 2016; Kantrowitz et al., 2017), possibly due to inverted-U dose response curves (Sehatpour et al., 2023); improvements in cognition with BI 425809 were not correlated with MMN changes (Schultheis et al., 2022); and more recent studies in healthy people suggest MMN may not be dependent upon NMDAR functioning (Harms et al., 2021).

There are several possible reasons why we failed to find a synergistic effect of PF-03463275 and CT on cognitive test performance.

First, dose–response relationships for GlyT1-I’s are complicated, with reduced effects observed at higher doses (Javitt, 2009; Umbricht, 2018). This study took several steps to maximize target engagement with PF-03463275. First, the doses of 40 and 60 mg were selected based on in vivo PET imaging data showing ideal occupancy and effects on an electrophysiological index of neuroplasticity (D'Souza et al., 2018). Second, to limit variability in PF-03463275 levels, only CYP2D6 extensive metabolizers were included. Third, daily medication adherence was visually confirmed with CAROMA. Despite these controls, neither dose of the study drug improved MCCB scores in patients receiving CT. Perhaps a wider dose range should have been tested.

Second, the pathophysiology of schizophrenia is heterogenous - perhaps only those individuals with impaired NMDAR functioning benefit from facilitating NMDAR activity. There are no biomarkers at present to identify patients who might benefit from glycine_B modulators. Variations in glycine/NMDAR function may predict different outcomes to treatment targeting NMDAR function. For instance, variations in the gene encoding glycine decarboxylase, the enzyme that catabolizes glycine, may influence response to glycine_B modulators. Triplication of GLDC would be expected to result in low levels of brain glycine and D-serine. In a small proof-of-concept study, glycine improved psychotic and mood symptoms in individuals with psychosis who showed triplication of the gene encoding GLDC (Bodkin et al., 2019).

Third, improvements in cognitive test performance may be more likely to be detected in patients who are earlier in their illness relative to our sample. Participants in this study were, on average, in their mid-40s, so we cannot exclude the possibility that younger participants, earlier in their course of illness, would have had greater benefit of CT with PF-03463275, as older participants seem to benefit less overall from CT (Wykes et al., 2009). In contrast, participants in the BI 425809 study that showed improvements in MCCB scores were younger (mean age ~36 years) and earlier in their illness (Fleischhacker et al., 2021).

Fourth, GLYT1-I’s may have subtly different effects on glycine levels in different individuals. The GLYT1 also works in reverse to increase extracellular glycine (Huang et al., 2004), in addition to the uptake of glycine, to tightly regulate synaptic and extrasynaptic glycine concentrations so that GLYT1 inhibition could decrease extrasynaptic glycine.

Our study design permits specific examination of the impact of GLYT1 inhibition on distinct cognitive and CT outcomes, including learning potential, achievement in CT, cognitive functioning, and self-assessed community functioning. We hypothesized that treatment with PF-03463275 would improve the neurobiological substrate of learning, thus augmenting the impact of CT on cognitive performance. Our results demonstrated no dose-related effect on the CT-related change in MCCB, suggesting that treatment with PF-03463275 did not augment the effects of CT. Further, lack of drug by time interaction on cognitive testing was consistent with a similar lack in improvement on CT. One possible interpretation of the present findings is that PF-03463275 failed to restore/enhance plasticity adequately, and therefore failed to augment the impact of CT. Another possibility is that detecting effects of an augmenting intervention in context of another, especially in a cross-over design, is difficult. A parallel group study comparing drug alone, CT alone, and the combined intervention may have produced clearer results, however, this would not have been feasible given our aim to evaluate PF-03463275 at more than one dose.

Finally, despite a significant improvement in MCCB performance over time in our study, there was no accompanying improvement in the PAOFI or SGI. This is consistent with the recognition that, in contrast to objective functional measures, PAOFI scores are not strongly predicted by neuropsychological test results (Chelune et al., 1986). It is possible that the lack of improvement in participants’ perceived functioning is related to the trial’s modest duration. Nevertheless, it is important to recognize the divergence between this improvement on cognitive testing and the lack of improvement reported by patients, particularly as patient-reported measures are increasingly identified as primary outcomes for clinical trials.

This study’s design has several strengths and limitations. Dosing choices were based on carefully conducted pharmacokinetic/pharmacodynamic studies that employed central target engagement along with functionally relevant outcomes: PET occupancy of GLYT1, LTP, and attenuation of BOLD response associated with ketamine-induced WM deficits. The PET study results were that 10 mg, 20 mg, 40 mg, and 60 mg of PF-03463275 twice daily for 1 week occupied ~44%, 61%, 76%, and 83% of GLYT1 respectively (D'Souza et al., 2018). This led to two doses tested in this study - it is possible that the range of doses tested in this paradigm was too narrow. The optimal occupancy for GLYT1-I’s is unknown, but a large study of bitopertin indicated that low to medium range occupancy was optimal (Umbricht et al., 2014). There may be several targets that a drug may engage, starting at the receptor and many levels downstream (e.g., LTP). For example, in this case, the most proximal target would be engagement of the glycine transporter. That was definitively demonstrated in vivo using PET (D'Souza et al., 2018). In the same study, 40 mg PF-03463275 enhanced LTP. However, further research is necessary to directly and conclusively link GLYT1 occupancy and/or LTP with the primary outcome of interest – improved cognitive test performance, and whether this measure of LTP is a reliable as a biomarker for neuroplasticity. The study was designed to specifically examine whether PF-03463275 could augment the effects of CT on CIAS. All participants in this study received CT, and the expected CT associated improvements occurred. The lack of a PF-03463275 alone arm prevented examination of drug alone on CIAS, as observed by Fleischhacker and colleagues with BI 425809 (Fleischhacker et al., 2021), but this was not the focus of this study. The within-subject cross over design is a more powerful design but one that might be prone to carryover effects; however, analyzing just the first treatment period alone did not reveal any significant effects that could have carried over to the next treatment period beyond the washout period. Likewise, in a within-subject design, practice effects from repeatedly taking the MCCB may have made it harder to detect drug effects. The duration of treatment with PF-03463275 on the MCCB outcome measure was 5 weeks, and perhaps this was too short, given that the Phase 2 12-week trial of BI 425809 where MCCB improvement continued between 6 and 12 weeks. Use of cognitive training to test the study hypothesis necessitated frequent appointments, which challenged retention. Stringent inclusion criteria to limit pharmacokinetic variability in PF-03463275 prolonged screening and restricted recruitment. While it is possible that the study could not detect a statistically significant improvement in MCCB with PF-03463275 due to insufficient sample size or baseline group differences (despite randomization) in race or sex (randomization was stratified by IQ only), there was no robust improvement in cognition with PF-03463275 above the effects of CT in this sample of people with chronic schizophrenia.

While our study failed to reveal positive findings, Phase 3 studies with GLYT1-Is remain in progress, and the positive results of the phase 2 trial of GLYT1-I BI 425809 are encouraging.

Many parameters must be optimized for a medication to augment the small to moderate effects of CT, including the drug itself, drug dose, duration of drug, timing of drug in relation to CT, the duration and intensity of the CT, the choice of CT program, and the study outcomes for people in varying stages of a heterogeneous illness. Nonetheless, improving deficient neuroplasticity in schizophrenia to enhance CT remains an attractive strategy. Robust, in vivo measures of neuroplasticity and their relationship to learning would help optimize the many variables in clinical trials of cognitive enhancers in CIAS.

Abbreviations

BOLD

blood oxygen level dependent imaging

CAROMA

cellphone assisted remote observation of medication adherence

CIAS

cognitive impairments associated with schizophrenia

CSF

cerebrospinal fluid

CT

cognitive training

DA

dopamine

EEG

electroencephalography

fMRI

functional magnetic resonance imaging

FSIQ

full scale IQ

GLYT1

glycine transporter-1

GLYT1-I

glycine transporter-1 inhibitor

LTP

long term potentiation

MCCB

MATRICS Consensus Cognitive Battery

MMN

Mismatch negativity

NMDAR

N-methyl-D-aspartate glutamate receptor

PAOFI

Patient’s Assessment of Own Functioning

PET

positron emission tomography

PANSS

Positive and Negative Syndrome Scale

rmANOVA

repeated measures analysis of variance

SD

standard deviation

SGI

Sensory Gating Inventory

VEP

Visual Evoked Potential

WTAR

Wechsler Test of Adult Reading