Impaired visuospatial but not verbal working memory in adult patients with neurofibromatosis type 1

Hanlu Tang (  ttsw1994@163.com ) Beijing Tiantan Hospital https://orcid.org/0000-0003-4766-6136 Qiong Wu Beijing Key Laboratory of Learning and Cognition, School of Psychology, Capital Normal University, Beijing, China. Shiwei Li Beijing Tiantan Hospital Yehong Fang Beijing Tiantan Hospital Zhijun Yang Beijing Tiantan Hospital Bo Wang Beijing Tiantan Hospital XingChao Wang Beijing Tiantan Hospital Pinan Liu Beijing Tiantan Hospital


Introduction
Neuro bromatosis type 1 (NF1) is an autosomal-dominant disorder with an average global prevalence of approximately 1/3000 [1] . The disease not only causes neuro bromas, café-au-lait macules, optic pathway gliomas and malignant peripheral nerve sheath tumors but also leads to structural changes in the brain [1,2] as well as various cognitive dysfunctions [3] . Compared with unaffected peers, individuals with NF1 usually exhibit de cits in many important cognitive domains, such as IQ [4] , visual perception [5] , language [6] , reading [3] , calculation [7] , attention de cit [8,9] and executive function [10] . In addition, working memory (WM), an important advanced cognitive function, has received increasing attention from clinicians.
WM is an important part of cognitive processing, providing temporary storage and manipulating essential information for complex cognitive activities such as learning, reasoning and language comprehension [11,12] . Based on the most in uential model of WM provided by Baddeley et al [13] , WM can be subdivided into verbal WM and visuospatial WM, involving the temporary maintenance and manipulation of verbal and visuospatial information, respectively [14,15] . Verbal WM supports language such as syntactic and semantic operations, as well as vocabulary acquisition during development [16] , and visuospatial WM supports perception, attention, actions to guide thought and higher-level cognition [17,18] . Their de cits will have an important effect on education-and work-related activities and consequently affect the quality of life. Moreover, WM dysfunction is heritable and is associated with susceptibility to schizophrenia [19] .
To date, previous NF1 studies involving WM have mainly focused on children [20][21][22] ,with few studies involving adult populations. Furthermore, the sample size of previous studies on WM involving adult patients was insu cient [23,24] , and one previous study recruited unmatched-IQ healthy controls (HCs) [23] .
The paucity and de ciency of WM data on adults with NF1 have limited our further understanding of NF1 disease from the perspective of the population characteristics. Moreover, most previous NF1 studies paid only one-sided attention to the general estimation of WM or focused on spatial WM [20,21] , verbal WM [25,26] , auditory WM [24] , or phonological short-term WM [27] separately. Few studies have conducted an integral analysis of WM in individuals with NF1 by isolating different domains of WM. Furthermore, previous methods measuring WM in NF1 varied without a uniform standard. It is debatable whether these methods can accurately assess WM. In addition, previous investigations have claimed that verbal and visuospatial WM have different domain-speci c cortical networks in the human brain [28][29][30] . However, few studies have offered evidence for this view.
To address the previous problems and biases, the current study recruited adults with NF1 and matched HCs to comprehensively analyze the features of WM in individuals with NF1 in both verbal and visuospatial WM domains based on Baddeley's WM model [13] using the N-back task [31] , which allows the verbal and visuospatial domains of WM to be precisely and simultaneously examined. The aims of our study were as follows: (1) to determine whether adults with NF1 exhibit de cits in WM (including verbal and visuospatial WM) and, if so, to clarify the detailed information of the de cits; (2) to supplement and re ne the existing data on WM in NF1 disease to provide a theoretical basis for clinical drug therapy and psychological intervention; and (3) to provide behavioral evidence for the previous view that verbal and spatial WM are independent of each other functionally and anatomically.

Participants
We recruited thirty-three adult patients with NF1 from the Neuro bromatosis Outpatient Department of Beijing Tiantan Hospital between 2019 and 2020. Thirty-six HCs were recruited from the community. Two patients were excluded for not completing the task, and two healthy controls were excluded due to misunderstanding the introduction. The nal sample included 31 adults with NF1 and 34 HCs. All individuals with NF1 ful lled the diagnostic criteria established by the National Institutes of Health (NIH) Consensus Development Conference [32] , and most of them had café-au-lait macules or small benign subcutaneous nodules on the body, but these skin lesions did not affect their daily activities. All patients were clinically stable, and none of them presented abnormalities on general neurological examination or limitations in daily life. Seven patients complained of mild memory decline, and three patients complained of attention de cits. All participants were right-handed and had normal or corrected-tonormal vision.
All participants were required to complete the short form of the Beck Depression Inventory (BDI-SF) [33] and the Mini-Mental State Examination (MMSE) [34] . The BDI-SF was used to quantify the general emotional state of the participants to avoid interference from emotional factors. Scores below or equal to 4 points indicate no or minimal depressive symptoms. Scores above this threshold indicate mild (5-7 points), moderate (8-15 points) or severe (≥ 16 points) depressive symptoms. The MMSE was used to assess the general cognitive status of the participants to ensure their ability to understand and cooperate in the further tasks of advanced cognitive function. Any score greater than or equal to 25 points (out of 30) indicates normal cognition. Scores below this threshold indicate mild (21-24 points), moderate (10-20 points) or severe (≤ 9 points) cognitive de cits.
The admission criteria of the individuals with NF1 and the HCs were (1) 18 ≤ age ≤ 60 y; (2) an MMSE score ≥ 24; (3) no severe NF1 symptoms, such as plexiform neuro bromas or malignant peripheral nerve sheath tumors; (4) no intracranial surgery history; (5) no history of serious chronic disease; (6) no mental illness or family history of mental illness; (7) no recent use of any medications that could affect cognitive abilities; and (8) voluntary participation as given by a signed consent document. All participants experienced the same experimental procedures in the same quiet room. All tests were administered by the same tester. This study was approved by the Medical Ethics Committee of Beijing Tiantan Hospital, The nal sample included 31 adults with NF1 (12 males and 19 females) and 34 HCs (10 males and 24 females). The ages of the individuals with NF1 and the HCs were 30.4 ± 7.7 and 31.1 ± 9.8 y, respectively.
The years of education of the individuals with NF1 and the HCs were 12.8 ± 3.1 and 13.0 ± 3.3 y, respectively. No individuals with NF1 or HCs presented with cognitive de cits as measured by the MMSE (28.8 ± 1.2 and 29.2 ± 0.9, respectively). The BDI-SF scores of the individuals with NF1 indicated moderate depressive mood, and the BDI-SF scores of the HCs indicated no or minimal depressive mood (8.9 ± 7.5 and 2.6 ± 3.5, respectively). The individuals with NF1 and HCs were matched for age (t (63) = -0.33, p = 0.741), gender (χ 2 (63) = 0.63, p = 0.429), educational attainment (t (63) = -0.24, p = 0.810) and MMSE score (t' (63) = -1.65, p = 0.105). The BDI-SF score (t' (63) = 4.28, p = 0.000) was not matched and was removed as a covariate in the data analysis (Table 1). The N-back Task The N-back task was used to measure the participants' accuracy and reaction times (RTs) to the stimulus under various memory load levels, including the visuospatial N-back task and the verbal N-back task, which examined the abilities of visuospatial WM and verbal WM, respectively [31] . The participants were required to determine whether the presented stimulus was the same as the Nth stimulus before [35] . The Nback program runs on the E-prime™ (Psychology Software Tools, Pittsburgh, PA, USA).
In the visuospatial N-back task, four gray boxes (up, down, left, right) were located in the center of the screen (Fig. 1a). In each trial, a random box turned yellowed for 500 ms followed by an interstimulus interval of 1000 ms. There were three conditions: 0-back, 1-back, and 2-back. The participants were instructed to indicate the location of the previous Nth box that turned yellow using the left, right, up, and down direction keys. In the 0-back condition, the participants pressed the corresponding key as the box turned yellow. The participants' RTs and accuracy were recorded. The task consisted of 18 blocks with six blocks in each condition and was presented in random order. Each block contained 20 trials and lasted approximately 30 s. The total time of the visuospatial N-back task was approximately 9 minutes. The participants practiced the task for 3 blocks (0-back, 1-back, 2-back each) before the formal test began.
In the verbal N-back task, a series of letters were presented one by one for 500 ms, followed by a 3500 ms blank screen (Fig. 1b). The four blocks (0-back, 1-back, 2-back, 3-back) were presented once in a xed order, with 18 letters in each block. In each trial, the participants were asked to press the left button if the current letter was the same as the Nth letter before; otherwise, they pressed the right button. For the 0back condition, the participants were required to indicate whether the current letter was "X". The participants' accuracy and RTs were recorded. The total time for the verbal N-back task was approximately 5 minutes.

Data analysis
The data from visuospatial and verbal N-back tasks were analyzed separately because there were different levels of N in each task. The accuracy and RTs were entered into repeated-measure ANOVAs, with the task di culty level (3 for the visuospatial N-back task and 4 for the verbal N-back task) as the within-subject factor and the group (NF1 and HCs) as the between-subject factor. Since the individuals with NF1 showed a slightly depressed mood, as indicated by higher BDI-SF scores than those of the HCs, we controlled this potential confounding factor by regressing out the BDI scores as the covariate.

Results
In the visuospatial N-back task, the overall mean accuracies of the NF1 and HCs groups were 62.84% ± 12.67% and 72.94% ± 13.41%, respectively, and their RTs were 598 ± 145 ms and 491 ± 158 ms, respectively (see Table 2 for details).  In the verbal N-back task, the overall mean accuracies of the NF1 and HCs groups were 87.16% ± 8.31% and 92.16% ± 4.71%, respectively, and their RTs were 753 ± 157 ms and 793 ± 193 ms, respectively (see Table 2 for details). Together, the results from the visuospatial and verbal N-back tasks suggest that the participants with NF1 have selective de cits in visuospatial WM but intact capability in verbal WM.

Discussion
In the current study, we analyzed WM in 31 adults with NF1 and 34 HCs by using the N-back task, and the results showed that adults with NF1 presented selective de cits in different subgroups of WM.

Visuospatial WM de cits in adult patients with NF1
In the visuospatial N-back task, the results showed that the accuracy decreased as task di culty levels increased in both NF1 patients and HCs. The mean accuracy of NF1 was signi cantly lower than that of HCs. In the 0-back task, adults with NF1 performed no difference from HCs. However, as the task di culty levels increased, the gap between the NF1 patients and HCs gradually became more obvious, especially in the 2-back task. The result of RTs revealed a similar phenomenon. These results suggest that adults with NF1 have signi cant de cits in visuospatial WM compared with HCs, and the de cits became more obvious as memory load increased. Our nding was consistent with the results of the Huijbregts S et al [36] study, in which WM de cits became apparent in children with NF1 as task di culty increased.
A visuospatial WM de cit was also reported in children and adolescents with NF1 (see S1), which suggested that this de cit could occur early in the life of individuals with NF1 and affect the development and academic achievement of school-age patients. Although a study recruiting 5 elderly individuals with NF1 (age>60, mean age 65) found that elderly NF1 patients presented spatial WM impairments compared with HCs [26] , few previous studies noticed visuospatial WM de cits in adult populations. Shilyansky C. et al [23] assessed spatial WM in adults with NF1 via two spatial delayed response tasks, and adults with NF1 presented signi cantly lower accuracy than HCs, with an apparent decline as memory load increased. However, his study used non-IQ-matched HCs for comparisons. In our study, we demonstrated visuospatial WM de cits in NF1 adults compared to HCs without general cognitive status bias by using the MMSE. In addition, combined with our reviewed previous literature involving WM in individuals with NF1 (see S1), we found visuospatial WM dysfunction to be a typical feature of NF1 patients across the lifespan (from childhood to advanced adulthood). Therefore, we should pay more attention to WM dysfunction in individuals with NF1 in the clinic.

No verbal WM de cits in adult patients with NF1
The present study showed no difference in verbal WM between adults with NF1 and HCs, which supported two previous studies on children patients. Hyman et al [25] found that there was no noticeable difference between NF1 patients and HCs in verbal WM, as assessed by both the digit span backwards test and the digit span forward minus digit span backwards test. Chaix Y et al [27] reported no auditoryverbal WM or phonological short-term WM impairment in children with NF1 via the ''WM index'' of the Wechsler Intelligence Scale for Children-Fourth Edition and the pseudoword repetition task, respectively. The descriptions and de nitions of verbal WM in previous studies varied, lacking universal terminology. Different researchers de ne verbal WM differently based on their educational background and professional a liation, which makes it di cult for us to interpret and compare the results of these studies.
However, several previous studies put forward different opinions. Descheemaeker M J et al [24] reported auditory WM de cits in adults with NF1 through the auditory verbal learning test (Dutch version). Costa Dde S et al [26] found that elderly individuals with NF1 presented verbal WM impairments compared with HCs through the digit span backwards and digit span forward tests. A possible reason for these con icting results may be that the methods (such as batteries) used to assess verbal WM varied (see S1).
For example, previous studies have assessed verbal WM via the ''WM index'' of the Wechsler Adult Intelligence Scale, which is biased toward a composite score and can obscure or skew the measurement of verbal WM. In addition, several previous studies have assessed verbal WM via the digit span forward and backward test, which essentially measures attention and the executive component of WM, respectively [37] . Therefore, the appropriateness of these measures is debatable. In the current study, we used the N-back task, which can better examine verbal and visuospatial WM, with the advantage of manipulating memory load by controlling the number of stimuli between the current stimulus and the target stimulus to increase memory load while eliminating other interfering factors [35] . Moreover, most previous studies used non-IQ-matched HCs and did not consider the effects of depressive symptoms on WM [38,39] . These factors affect the accuracy of the research results and cause controversy. In our study, we used general cognitive status-controlled HCs and eliminated the effect of depressive symptoms as a covariate.

Functional-neuroanatomical dissociation between verbal and visuospatial WM in NF1
Jonides J et al [28] found that verbal and visuospatial WM were implemented by different neural structures via positron emission tomography studies. Then, with the application of functional MRI, Gruber O et al [30] further demonstrated that verbal and spatial WM were mediated by distinct neural networks with some overlapping and some distinct neural circuitry. The present study showed that adults with NF1 presented de cits in visuospatial WM but not verbal WM, which demonstrates that verbal and visuospatial WM exhibit functional dissociation. Our results offer new behavioral evidence for this established view.
Dissociation has also been found in patients with autism [40] . Moreover, dissociation could account for why individuals with NF1 complain of no di culties in daily life but perform poorer than their peers in academic or advanced brain function activities because visuospatial WM involves higher cognitive functions than verbal WM.
Potential neural mechanisms underlying WM dysfunction in NF1 As a monogenic disease, NF1 provides a unique genetic model to explore and mechanistically dissect the molecular mechanism underlying WM. NF1 gene mutation is the core cause of WM impairment in NF1 patients. Reduced NF1 gene expression in neurons leads to decreased neuro bromin production, resulting in downstream molecular and signaling pathway abnormalities related to memory, such as reduced cAMP levels [41,42] , increased GABA release [23] and reduced long-term potentiation [41] . Moreover, decreased neuro bromin production will cause low intracranial DA levels in individuals with NF1 [41] , which is associated with WM impairments [43] . Notably, prefrontal DA levels are related to the activity of catechol-O-methyltransferase (COMT), which can degrade cortical DA levels [44] . However, a variant of the COMT gene (Val-to-Met) has reduced COMT activity, leading to higher prefrontal DA levels, which will ameliorate verbal WM de cits [45,46] . Individuals with Val/Val genotypes, Val/Met genotypes, and Met/Met genotypes had lower, moderate, and higher performance in the N-back task, respectively [19] . COMT polymorphisms (Val/Val, Val/Met, Met/Met) may account for the inconsistent results of previous studies involving verbal WM in individuals with NF1.
Additionally, the hypoactivation of key components of WM circuitry (the right parietal cortex and the left dorsolateral prefrontal cortex) and aberrant functional connectivity in individuals with NF1 may underlie their visuospatial WM di culties [47] . Furthermore, individuals with NF1 show a more diffuse pattern of increased brain activation than HCs during high-vs low-memory-load tasks, which may re ect a less e cient pattern of brain activity [47] . This could explain why, in our study, the participants with NF1 performed worse than HCs during high-visuospatial-memory-load tasks (1-back and 2-back tasks) but not during low-memory-load tasks (0-back task). Further investigations of the neural mechanisms underlying WM dysfunction in NF1 will be conducted in the future.

Conclusion
Compared with HCs, adults with NF1 have selective de cits in WM (impaired visuospatial but not verbal WM), and visuospatial WM dysfunction becomes more obvious as memory load increases. Visuospatial WM dysfunction is a typical feature of NF1 across the lifespan (from childhood to advanced adulthood).
Decreased NF1 gene expression and its downstream molecular and signaling pathway abnormalities as well as local brain neuronal activity abnormalities may be the potential neural mechanisms underlying WM dysfunction in NF1. Our results supplement and re ne WM data in NF1 and demonstrate functional separation between visuospatial and verbal WM. letter before, otherwise pressed the right button. For the 0-back condition, participants were required to indicate whether the current letter was "X".

Figure 2
Results on accuracy (a) and reaction times (b) in the visuospatial N-back task. Adult NF1 patients showed de cits in visuospatial WM compared with HCs. Error bar was 95% con dence interval. HCs, healthy controls; NF1, adults with neuro bromatosis type 1.

Figure 3
Results on accuracy (a) and reaction times (b) in the verbal N-back task. Adult NF1 patients showed no de cits in verbal WM compared with HCs. Error bar was 95% con dence interval. HCs, healthy controls; NF1, adults with neuro bromatosis type 1.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. S1.docx