Higher CSF/serum free-T4 ratio is associated with improvement of quality of life during treatment with L-thyroxine

Up to 20% of individuals with primary hypothyroidism treated with L-thyroxine still suffer from severe symptoms. These are supposedly brain derived and involve both cognitive and emotional domains. Previously, no consistent relationship has been found between thyroid hormones (TH) or TSH levels in blood and quality of life (QoL). Recently, we reported an association between cerebrospinal fluid (CSF)/serum free-thyroxine (f-T4) ratio and QoL, in juvenile hypothyroid patients. Here, we investigated if CSF/serum f-T4 ratio and QoL estimates correlate also during L-thyroxine treatment. Moreover, the CSF biomarker neurogranin (Ng) was used as a biomarker for synaptic function and integrity in clinical research. Ng is partially controlled by TH and therefore we investigated the relationship between QoL parameters and Ng levels. Patients diagnosed with primary

hypothyroidism were investigated using vital parameters, serum and CSF analyses of TH, TSH, Ng and QoL questionnaires.Similar procedures were performed after 6 months of treatment.The most marked associations with QoL were found for CSF/serum f-T4 ratio, which was strongly related to several QoL parameters such as the mental subscore of SF-36 (r = 0.83, p < .0005).Ng, which did not differ from that in our healthy controls, was lower in some patients during treatment and higher in others.However, the change in Ng during treatment was significantly correlated with QoL parameters including the mental subscore of SF-36 (r = À0.86,p < .0001).In addition, the CSF/serum f-T4 ratio correlated with the change in Ng (r = À0.75, p = .001).Our results suggest that the ratio between CSF and serum f-T4 is an important biomarker for QoL during treatment of patients with primary hypothyroidism, so far in research, but in the future maybe also in clinical settings.Moreover, this ratio also correlates with the changes in Ng levels during L-thyroxine treatment, further supporting the impact of the TH balance between serum and CSF on QoL.

| INTRODUCTION
4][5][6][7] PH is associated with a broad array of clinical symptoms as thyroid hormones affect virtually all the cells in our body.The symptoms include muscle pain, decreased sweating, peripheral edema and CNS symptoms such as fatigue, depressive state, and cognitive deficits. 8,9Most of these symptoms will recede following levothyroxine (L-thyroxine) treatment.[14][15] Thyroid-stimulating hormone (TSH) is increased in untreated PH due to lack of sufficient feedback inhibition from T4 and T3 and the measurement of TSH in peripheral blood is used as a biomarker to assess the disease status.However the reference interval for TSH in healthy individuals, which is used to adjust the dose of L-thyroxine in patients with PH, is to a large extent based on clinical practice rather than on randomized controlled trials. 16Furthermore, it has been reported that the level of TSH is not related to the QoL as experienced by patients receiving L-thyroxine treatment. 4,17Substitution with L-thyroxine and the subsequent lowering of TSH decreases the activity of the type II iodothyronine deiodinase-enzyme (DIO2) by downregulation of the synthesis. 18This could possibly explain why serum(s) T3/T4 ratio, both in terms of total and free fractions, of the hormones are lower during treatment as compared with healthy controls, since DIO2 activity is downregulated by the excess T4 concentration. 19,20These lower T3/T4 ratios have been inversely associated with physical performance 6 and mood. 21We recently reported that low cerebrospinal fluid (CSF)/serum f-T4 ratio is correlated with QoL parameters in patients with untreated PH, but not in healthy controls, indicating that also the ratio of TH over the blood-brain-barrier (BBB) is of importance. 22urogranin (Ng) has been proposed as a potential biomarker of synaptic function and integrity.Its transcription is regulated by thyroid hormones, 23,24 and knockout of the Ng-encoding NGRN gene in mice results in inability to produce Ng which entails impaired spatial memory, which can at least in part be reversed by the administration of Lthyroxine. 25 However, there is, as yet, no data for patients with PH showing that Ng could potentially be useful as a biomarker of central nervous functions and synaptic dysfunctions. 26 summary, many patients suffer from residual symptoms during Lthyroxine treatment despite normalized serum TH levels.It is therefore important to further study the mechanisms underlying the residual symptoms in PH.The main aim of the present study was to evaluate whether the CSF/serum f-T4 ratio is associated with QoL during treatment with Lthyroxine.A second aim was to investigate whether Ng levels are associated with the CSF/serum f-T4 ratio and QoL parameters.The results showed an association between CSF/serum fT4 ratio and QoL, and both these variables were in turn related to changes in Ng.

| Study population
Patients with PH (n = 15, 11 women), mean (M) age: 52 ± 12.2 were recruited at the Unit of Endocrinology at the Department of Medicine, Halland Central Hospital, Sweden (Table 1).The patients had been invited by their primary physician to participate in the study.Out of screened patients there were only two who did not want to participate, for which we do not have any characteristics.All patients displayed at least two pathologically increased TSH (>4.0 mIE/L; reference interval: 0.40-4.0mIE/L) prior to admission.In addition, all patients had a f-T4 level in the lower-half of the reference interval (which was 11-22 pmol/L) with a mean of 11.3 ± 2.6 pmol/L.Furthermore, they showed more than five points according to the Zulewski scale 27 to ensure that they were clinically hypothyroid.The second visit occurred 182 (119-623) days after the first visit.The basal patient parameters are presented in Table 1.
All 25 hypothyroid patients described in our previous study 22 were invited to a second visit.Ten patients did not attend the second visit as three had moved out of our region, four had experienced post-puncture headache or discomfort after the initial lumbar puncture, one was pregnant, one was severely ill, and one was unreachable.The patients had a regular contact with the principal investigator of the study to make sure of compliance.Regular blood sampling was also done during the study period to make sure that the TSH levels did decrease.
Exclusion criteria were treatment with antithyroid substances or thyroid hormones, and recent use of iodine contrast or pregnancy.In all 15 patients, recent onset of fatigue, hypersomnia or lethargy was the main reason for consulting their primary physician.Apart from PH, one subject was previously diagnosed with type-1 diabetes, four patients were treated with estrogen hormone replacement (oral estrogen or patch), and one subject received glucocorticoid treatment (i.e., budesonid inhalation).
None of the 15 patients fulfilled any criteria of a depressive diagnosis based on medical record review and clinical assessment at the visit.All patients were of Caucasian origin.
Since there are no normal reference values for TH in CSF, we present data for all our healthy controls.In the study they are referred as a reference group.They were studied at one visit.The reference group (n = 50, 26 women) were recruited by posters in the staff rooms at the medical clinic, ambulance, local senior citizen association, and the University of Halmstad.All controls were Caucasians, had TH and TSH levels within the reference intervals, and a Zulewski score 27 of 0-4 points.One person in the reference group had to be excluded due to a very high neurogranin value.

| Ethical considerations
The study was approved by the Regional Ethical Committee in Lund, Sweden (diary number 253/2008, 585/2009 and 491/2010).Oral and written informed consent was obtained from all participants.

| Bodyweight, height and area
Bodyweight was measured in the morning to the nearest 0.1 kg and body height was determined barefoot to the nearest 0.01 m.
Body surface area (BSA) was calculated according to the Du Bois formula.

| Sampling of blood and CSF
All sampling of blood and CSF was performed after a minimum of 8 h fasting.Venous blood samples were obtained by venipuncture, centrifuged at 2400 g for 10 min between 1 and 3 h after sampling and serum were aliquoted in polypropylene tubes and stored at À20 C pending biochemical analyses of TSH, f-T3, f-T4, and thyroid peroxidase antibodies (TPO-Ab).
T A B L E 1 Clinical characteristics in hypothyroid patients before and during treatment, as well as in healthy controls.

| Biochemical procedures
All serum and CSF samples were analyzed in the same assay run to minimize the analytical interassay variation.TSH, f-T3 and f-T4 were analyzed by dissociation-enhanced lanthanide fluoroimmunoassays 28 (Auto DELFIA).The intra-assay coefficient of variation (CV%) for the TSH assay was <4.4% for all levels down to 0.1 mIE/l, and the intraassay CVs for the f-T4 and f-T3 assays were <6.1% for all levels down to five and two pmol/L, respectively.We chose the free component of TH and not the total as in the clinical setting it is most relevant to use serum f-T4 and not serum T4.TPO-Ab were analyzed by chemiluminescent microparticle immunoassay (CMIA) (Architect i2000, Abbott).The intra-assay CV for the TPO antibody assay was <10% for all levels down to 5 IU/mL; >15 IU/mL was considered as a positive result.
CSF neurogranin (Ng) was measured using an inhouse enzymelinked immunosorbent assay (ELISA), based on the Ng2 and Ng22 monoclonal antibodies, as previously described. 29

| Quality of life
After lumbar puncture and venous blood sampling had been performed, all patients completed the self-assessment of general health using a Likert scale (GHLS) 30 as well as the Montgomery-Åsberg Depression Rating Scale (MADRS) questionnaire to evaluate symptoms of depression. 31SF-36 32 was used to calculate the sum score for physical health (PCSc) and for mental health (MCSc) as previously described. 33l QoL assessments were performed both before and after in mean 6 months of L-thyroxine treatment and we could therefore calculate the change between them (duringbefore).

| Statistical methods
We used the R 4.0.5 software for statistical analyses.The biomarker and test results are presented as the mean ± SD.
We used the Shapiro test and QQ-plots to confirm that the data was normally distributed.The delta values, with both positive and negative values, were transformed with the power of 10 transformation.Spearman's rank correlation coefficient was sought for correlation analysis.A paired t-test was used to calculate withingroup effects (comparing before and during treatment).A twotailed p-value <0.05 was considered statistically significant.pvalues are marked in tables with * according to the international standard for significance level.Since we have looked at several parameters of TH we adjusted the p-values using the Holm method with seven TH values.

| Quality of life
The patients improved in most QoL parameters during treatment (Table 2 and Figure 1).MADRS score was reduced, GHLS and MCSc were increased, and PCSc tended T A B L E 2 Mean ± SD for thyroid hormones, neurogranin and quality of life (QoL).to increase (n.s.).However, during treatment, QoL was still lower in the patients compared with the healthy controls.In Table 2 data are also presented for 51 healthy controls.There was also a significant decrease in both serum and CSF f-T3/f-T4 ratio, from 0.51 to 0.37 and 0.33 to 0.16 respectively (Table 2).

| Correlations between CSF/serum f-T4 ratio and QoL
The CSF/serum f-T4 ratio correlated strongest with QoL parameters during treatment, (Table 3 and Figure 2).

| Correlations between CSF/serum f-T3 ratio and QoL
There were no significant correlations between CSF/serum f-T3 ratio and QoL.

| Correlations between serum f-T3/f-T4 and QoL
The Spearman correlation test did not show any significant correlation between f-T3/f-T4 ratio in serum and QoL parameters.

| Correlations between CSF f-T3/f-T4 and QoL
The CSF f-T3/f-T4 ratio did not correlate significantly with any QoL parameters.Since treatment with L-thyroxine lowers the ratio between f-T3/f-T4 in both serum and CSF we also tested for correlations between ΔCSF f-T3/f-T4 and QoL parameters, but no significant correlation was found.

| Multiple comparison effect
The most important measurement we have for QoL is MCSc.The adjusted p-values gave a significance for serum f-T3 during treatment (adjusted p = .032)and for the CSF/serum f-T4 ratio during treatment (adjusted p = .0024)compared with MCSc during treatment.Other TH values and ratios were not significant to MCSc when adjusted.

| Correlations between neurogranin and CSF/ serum f-T4 ratio
There was no significant correlation between Ng level and CSF/serum f-T4 ratio during treatment, but there was a highly significant correlation between ΔNg (the change in Ng level) and the ratio (r = À0.75, p = .001).There was no significant correlation between ΔNg and the ratio of CSF/serum fT4 before treatment.

| Correlation between neurogranin and QoL
ΔNg correlated with all QoL parameters during treatment (Table 3).recently been reported that these hormone levels do not correlate with QoL measurements including mood parameters. 4,17Earlier, only a few studies have investigated the relationship between serum and CSF levels of thyroid hormones, [34][35][36] but only to the extent of dynamics between the two compartments.We recently reported 22 that the CSF/serum f-T4 ratio, but not the serum or CSF concentrations of f-T4 or f-T3, was associated with QoL parameters, in untreated PH.The present study shows that also during L-thyroxine treatment, the CSF/serum f-T4 ratio is correlated with QoL estimates.No correlation was found for f-T3 or the CSF f-T3/f-T4 ratio, and not even for the ΔCSF f-T3/f-T4 ratio, which is regularly changed in response to treatment. 37Furthermore, we believe that a very low QoL before treatment has a variety of different reasons.After L-thyroxine treatment, when our PH patients had a better general QoL, we observed that QoL measures were more markedly correlated with the CSF/serum f-T4 ratio despite the relatively small number of patients (n = 15).

| Thyroid hormones in CNS
All TH are produced in the thyrocytes of the thyroid gland.Under physiological normal pH conditions the main part of TH produced is T4. 38There are some post synthetic conversions to T3 in the thyroid gland 39 but mainly T4 is the secreted hormone transported to the effector organs, for instance through the BBB.If T4 in CNS has a unique role for functions like mood is not fully clear although there are in vitro studies suggesting this. 40Furthermore, mutations affecting any of the two most important proteins for the transport of T4 over the BBB results in severe mental retardation. 41,42These two proteins, organic anion trans-porter1 C1 (OATP1c1) and monocarboxylate transporter 8 (MCT-8), have a higher affinity for T4 than for T3. 43,44Therefore, the TH entering CNS is primarily T4, together with a minor amount of T3 passing via the MCT-8. 43,44In addition to the entrance via these mechanisms there is evidence for a protein-binding transport from the choroid plexus to the CSF with, in the CNS existing protein, transthyretin (TTR). 45,46The initial entrance to the choroid plexus is also via the TH transport proteins. 47After the entrance, there is a local conversion inside the brain from T4 to T3.The activity of the DIOs, for this conversion, is regulated by TH meaning that when THs are abundant, the DIO2 activity is consequently downregulated.The DIO2 and the DIO3 are the clinically active DIOs in the adult brain where the DIO3 exists as a scavenger mechanism in the neurons. 47,48e next step for the TH is to bind to the thyroid hormone receptors (TRs).Then, specific effects are induced in the CNS such as promoting axonal myelinization, dendritic spine growth, astrocyte and oligodendrocyte activation and microglial development. 49Probably T3 mostly bind to TRs although T4 may also have some effect per se. 40us, it is most likely that T3 is the main effector for the changes in gene transcription and other effects.It may therefore be surprising that we did not detect any correlation between QoL and the absolute concentration of f-T3 in CSF, or with the CSF/serum fT3 or CSF fT3/CSF fT4 ratios.Instead, we found a strong correlation with the CSF/serum f-T4 ratio.
It could be assumed from our results that low QoL, and other CNS symptoms in PH, are linked to deficiency of "T4-substrate" inside the brain.Therefore, the primary aim of treatment should be to activate or induce more abundant TH transport proteins.Another interesting theory regarding the "TH-substrate" for the brain is the TH-binding protein in the CNS, TTR.We have not measured the amount of this protein and, to our knowledge, no one has in hypothyroid patients.There are mouse studies with a knockout of TTR, and consequently lower amount of T4 in the CSF, and with developmental failures but no evidence that TTR plays a major role in grown up animals. 45A recent review article for CSF biomarkers in depression 50 depict two studies, a total of 39 depressive patients, where TTR is lowered, compared to healthy controls.Whether this has any implication for the TH in the brain remains unknown.It seems most likely that TTR-status in humans is correlated with inflammation and malnutrition, and not with hypothyroidism. 51 important objective to the theory of elevated T4 as a primary goal is that in the present study CSF f-T4 concentration was increased by the L-thyroxine treatment, but there was no significant correlation between CSF f-T4 and the QoL measurements.
Therefore, it is probably not that simple-that a higher concentration of f-T4 in the brain is the most important.Instead, the truth probably lies in a more complex "TH-world."Therefore, the CSF/serum fT4 ratio could be a "marker" rather than an actual exerting mechanism.Although this "marker" of a better QoL in PH is undoubtedly controlled by the TH-transporters, the mechanisms underlying the effects on the TH-transporters could be of greater importance.It seems reasonable to believe that the main effector of the CNS symptoms in PH, that exerts an effect on great emotional and cognitive domains, is TRH.TRH is shown in several studies to be a widely CNS controlling hormone [52][53][54][55] and has even been proposed to be considered a neurotransmitter instead of a hormone. 56It exerts effects on almost all hypothalamic systems 48,[52][53][54]56,57 and is therefore not solely a hormone controlling the pituitary hormone TSH.

| Neurogranin
QoL parameters did not only correlate with the CSF/serum f-T4 ratio, but also with the ΔNg level, probably reflecting the degree of synaptic plasticity. 24,25Ng is a protein kinase C substrate that binds calmodulin in its nonphosphorylated state.It is of great importance to transfer the signal in the postsynaptic neuron that begins with glutamate binding to the NMDA receptor.The building up of synapses from one neurons axon, or cell body, to other neurons dendrites is one of the most plastic processes in our body.If the brain is active in all its cognitive and emotional domains, we can grow several thousand synapses per day.In a more redundant stage, many of these synapses actually break down again, and some of the materials will be recycled elsewhere.
When synapses degrade or remodel, there is leakage or secretion of, the whole or especially the C-terminal domain of the Ng protein.This protein can be analyzed in CSF and is believed to reflect postsynaptic dendritic plasticity.Xiang et al. 26 have also suggested that Ng can be used as a biomarker in a variety of diseases affecting the brain, such as neurodegenerative, neurological and psychiatric diseases.In experimental studies, knockout of the Ng-gene in mice caused disability to build this protein, leading to impaired spatial memory. 23,24Interestingly, administration of L-thyroxine can, in part, reverse this impairment. 24The transcription of the Ng-gene is controlled by a variety of factors, including TH.The Ng-gene was the first gene in the brain that was found to be regulated by THs. 24 has been proposed that Ng is a valid marker of the effects of THs in the rodent CNS. 49As Ng is important for neuronal plasticity, we hypothesized that Ng would be decreased during L-thyroxine treatment in PH, but in the present study the CSF Ng level nonsignificantly tended to increase during treatment (Figure 1).Thus, these findings do not support our hypothesis that Ng is a marker of brain insufficiency in PH.However, ΔNg (the change in Ng level) correlated significantly with all QoL parameters during treatment, with lower QoL after L-thyroxine treatment in patients who had marked increases in Ng concentrations (Figure 2).Ng-levels in CSF might be a unique biomarker of synaptic plasticity that could be used in further research on diseases affecting cognition, even in mild cases.The changes in Ng-levels must always be looked at on an individual, not on a group level.
An elevation of Ng on the other hand, from within a reference interval, may reflect a disease that primarily degrade neurons, that is, a neurodegenerative disease.A recent review of biomarkers in depression 50 did not find any changes in absolute concentration of Ng, in two studies comparing Ng in depression from healthy controls.
Because Ng levels in PH patients were in close proximity as to the levels in the healthy controls, and the changes in CSF Ng levels were variable in individual PH patients, we cannot state that PH is a neurodegenerative disease affecting the degradation of neurons.Since there was an increase in Ng in about half of the patients and the other half decreased we cannot state that TH in the brain is directly involved in Ng dynamics (meaning transcription and degradation of Ng).Instead, as PH clearly does change cognitive processing, we believe that this change is seen in the Ng measurements of CSF.
Therefore, being a marker for synaptic plasticity, 25,58 the Ng concentration in CSF would presumably be changed in any disease affecting cognitive processes, and not only because of concentrations of TH in the CNS.

F I G U R E 1
During treatment, serum-TSH decreased from 8.5 to 0.8 mIU/L, and both serum and CSF f-T4 increased from 11.3 to 17.1 pmol/L and Results presented as box plots with median, interquartile range (IQR) and outliers.(A) Shows the Likert scale for general health (GHLS, points).(B) Illustrates the Montgomery-Åsberg Depression Rating Scale (MADRS) score (points) in PH before and during treatment.Since no data on MADRS in the healthy controls were collected in the present study, this data is based on a review article by Zimmerman et al. and values for 569 patients.59(C) Represents the mental subscore of SF-36 (MCSc, points).(D) Depicts the physical subscore of SF-36 (PCSc, points).(E) Shows the ratio between f-T4 in CSF and serum.(F) jPresents the ratio between f-T3 and f-T4 in CSF.Finally, box plot (G) illustrates neurogranin (pg/mL) in hypothyroid patients before treatment, hypothyroid patients during treatment and in healthy controls.from 8.5 to 13.1, respectively.Serum f-T3 increased from 5.4 to 6.3, but in CSF there were no significant changes.

F
I G U R E 2 (A) Shows the correlation between the change in neurogranin (during-before, pg/mL) and the CSF/s f-T4 ratio during treatment (r = À0.75, p = .001).(B) Illustrates the correlation between the mental subscore of SF-36 (MCSc, points) during treatment and the change in neurogranin (during-before, pg/mL) (r = À0.87,p < .0001).(C) Depicts the correlation between the ratio between the free component of thyroxine (fT4) in CSF and serum during treatment and the mental subscore of SF-36 (MCSc, points) during treatment (r = 0.80, p < .0005).