Towards DSM 10: A Bio-Classification of Developmental Schizophrenia?

Our traditional diagnostic systems developed starting in the 19th century, with the works of Emil Kraepelin and Eugen Bleuler, and later that of Kurt Schneider. Thus, current clinical features have incorporated, to a greater or lesser extent, Kraepelin’s chronicity and Bleuler’s negative symptoms within the construct of schizophrenia. The most recent iterations of the DSM and ICD provide clinicians with dimensional assessments based on key symptom domains covering positive, negative, affective and cognitive symptoms (note however, that the DSM-5 in contrast to its predecessor, DSM-IV, does not give credence to Schneiderian first rank symptoms). The patient will thus receive a diagnosis of schizophrenia and an indication of the severity of symptoms across each dimension.

However, the dominant etiological model for schizophrenia posits that the illness is the end state of abnormal neurodevelopmental processes that started years before the illness onset (Rapoport et al., 2012). Presently, the diagnosis of schizophrenia does not incorporate any etiology-based components nor neurodevelopmental markers. This raises the question “if it talks like a neurodevelopmental disorders and walks like a neurodevelopmental disorder why don’t we diagnose it as a neurodevelopmental disorder?”

Longitudinal epidemiology and the development of schizophrenia

As a potential way forward that may even resolve the aforementioned ‘dilemma’, longitudinal developmental cohort studies are powerful tools for investigation and provide valuable insight into disease etiology, while overcoming the ‘noise’ that frequently dominates cross-sectional and clinical studies. Such studies show, for example, that positive psychotic symptoms are very common in the general population, presenting in over 10% of adults (Mollon et al., 2016) while being transient manifestations in the large majority of subjects (Calkins et al., 2017). Furthermore, psychotic symptoms were not specific to a diagnosis of schizophrenia in adulthood and instead, could serve as a broad marker of adult mental health problems (Fisher et al., 2013).

However, longitudinal developmental studies of cognition and schizophrenia have shown that only future schizophrenia involves dynamic developmental processes, affecting cognition throughout the first 2 decades of life and leading to increasing dysfunction. Cognitive impairment is ubiquitous in patients with schizophrenia and childhood cognitive deficits precede the appearance of adult schizophrenia by a decade at least. Critically, these developmental processes do not manifest in other psychiatric disorders, including not in psychosis with depression, or depression without psychosis (Mollon et al., 2018), thereby refuting the view that early disorders of cognitive performance are too common to be effective in predicting schizophrenia (David et al., 1997). These key observations from the longitudinal approach then align well with the neurodevelopmental model of schizophrenia that postulates subtle behavioral, motor, and cognitive deviations already be apparent in childhood, years before overt clinical symptoms of adult schizophrenia manifest.

Another example illustrating the potential promise of the longitudinal approach, though less frequently highlighted in the field, is the developmental deficit in social functioning. Schizophrenia is associated with severe deficits in social functioning. Importantly, similar deficits in social functioning are present prior to psychosis onset, in childhood and adolescence. Furthermore, poor social functioning does differentiate pre-schizophrenia children and adolescents from their peers and can be a sensitive and potentially specific predictor of schizophrenia, not just psychopathology in general (Reichenberg et al., 2002; Tarbox and Pogue-Geile, 2008).

Developmental or (epi)genetic Bio Signatures?

We believe that the evidence reviewed above is compelling in terms of implementation into diagnostic decision making, or at least offers very strong rationale for public investment into ongoing and future ‘population-scale’ longitudinal, decades-long studies with subjects recruited in childhood years.

However, it remains an important question whether the approach could have an even stronger impact with the integration of neurobiological markers. A comprehensive discussion on the plethora of markers and ‘endophenotypes’ implicated in the developmental neurobiology of psychiatric disease would be far beyond the scope of this commentary. We would like to mention, however, two informative examples. One of these includes longitudinally conducted functional and structural neuroimaging. These studies, while enriching the field with evidence-based hypotheses such as ‘accelerated aging in schizophrenia’ (Schnack et al., 2016) are for practical reasons, perhaps with the exception of youth clinically at high risk for developing the disease (Collin et al., 2020), difficult to implement in the years and decades prior to the emergence of symptoms.

Another biomarker example, that on a population-scale would be much easier to implement, at least when compared to neuroimaging, would be DNA methylation profiling in peripheral blood samples obtained by venipuncture. Genome-scale, sequence-specific methylation profiling of DNA cytosine residues from blood samples is nowadays a widely used approach in all fields of genomic medicine, with the underlying idea that stress and other environmental, or internal and genetic factors, potentially operating in concert, alter the epigenetic landscape of the subject’s DNA. Blood-based DNA methylation certainly is amenable to long-term longitudinal measurements starting at the time of birth. Furthermore, there are attractive concepts currently infiltrating the field, such as the ‘accelerated epigenetic clock’ and its interrelation with increased mortality risk in schizophrenia (Higgins-Chen et al., 2020). But, early evidence suggests that neuroimaging-based acceleration of brain aging, and DNA methylation as an aging marker in blood, may reflect distinct biological processes and, by combined integrative analyses, do not appear informative in the context of schizophrenia (Teeuw et al., 2021). Eventually, most of the current studies of biomarkers face issues related to being practical to implement in longitudinal larger scale childhood cohorts (e.g., neuroimaging), or of questionable informativeness about brain function (e.g., blood DNA methylation).

As a final point in this discussion we would like to mention risk-associated genetic variation, including common variants mostly defined by single nucleotide polymorphisms each contributing a minimal but tractable contribution to heritable risk and in toto often computed into ‘polygenic risk score’ (PRS), and rare mutations of high penetrance. Many of such types of risk-associated DNA sequences are considered to regulate genes of pivotal importance for orderly brain development and, moreover, neurodevelopmental risk sequences are over-represented among the hundreds of epigenetically altered loci in schizophrenia postmortem prefrontal cortex (Girdhar et al., 2021). Interestingly, the schizophrenia PRS, when studied in some larger-scale cohort of +8000 adolescents followed over the course of multiple years, have shown an association with anxiety and negative symptoms - but not psychosis (Jones et al., 2016). This finding aligns well with the aforementioned psychological and behavioral domains associated with premorbid level of functioning of children and adolescent that developed schizophrenia in their later years. From this perspective, it will be exciting to see whether or not future, population-scale longitudinal studies monitoring the risk for schizophrenia, or the emergence of disease, could successfully integrate the PRS score in their analysis and prediction models. However, we would like to end this commentary with a word of caution, given that the PRS, at least in a larger study with 8,541 densely phenotyped adult subjects with schizophrenia, failed to improve the performance of prediction models for symptom progression, poor prognosis or other disease outcomes (Landi et al., 2021).

In conclusion then, we consider ‘population-scale’ longitudinal neuropsychological, behavioral and cognitive profiling, starting in childhood and ideally continuing throughout early- and mid-adulthood, as an extremely promising approach to advance the clinical conceptualization of schizophrenia as a neurodevelopmental disorder, while (pending some unforeseen dramatic technological advances) remaining skeptical about the informative power of present day biomarker- or genetics-based approaches in such type of clinical context.

We are aware that formidable challenges remain for such type of endeavor. For example, since there is no gold standard for the diagnosis of schizophrenia, longitudinal profiling in itself is unlikely to shift the timing of the diagnosis towards an earlier age for many of the affected individuals. Further complicating is the fact that not all patients diagnosed with schizophrenia exhibit pre-morbid developmental antecedents. Finally, practical hurdles in terms of resources and subject participation could limit such types of approach to all but those with familial high-risk for the disease. Given these challenges, in combination with the arduous and very lengthy process of testing and validating the aforementioned concepts in the field, the authors would predict that the term ‘developmental schizophrenia’, as a disease specifier, will find its way into diagnostic manuals not before another few decades have passed, metaphorically referred to as ‘DSM-10’ in the title of this commentary.