In vivo response of the human epigenome to vitamin D: A Proof-of-principle study

Accepted Manuscript Title: In vivo response of the human epigenome to vitamin D: A proof-of-principle study Authors: Carsten Carlberg, Sabine Seuter, Tarja Nurmi, Tomi-Pekka Tuomainen, Jyrki K. Virtanen, Antonio Neme PII: DOI: Reference:

S0960-0760(18)30003-7 https://doi.org/10.1016/j.jsbmb.2018.01.002 SBMB 5098

To appear in:

Journal of Steroid Biochemistry & Molecular Biology

Received date: Revised date: Accepted date:

2-11-2017 28-12-2017 4-1-2018

Please cite this article as: Carlberg C, Seuter S, Nurmi T, Tuomainen T-P, Virtanen JK, Neme A, In vivo response of the human epigenome to vitamin D: A proofof-principle study, Journal of Steroid Biochemistry and Molecular Biology (2010), https://doi.org/10.1016/j.jsbmb.2018.01.002 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Carlberg et al.: In vivo response of the human epigenome to vitamin D

1

In vivo response of the human epigenome to vitamin D: A proof-of-principle study

Carsten Carlberg1,#, Sabine Seuter1, Tarja Nurmi2, Tomi-Pekka Tuomainen2,

School of Medicine, Institute of Biomedicine, University of Eastern Finland, Kuopio,

SC R

1

IP T

Jyrki K. Virtanen2 and Antonio Neme1

Finland 2

Institute of Public Health and Clinical Nutrition, University of Eastern Finland, Kuopio,

N A

Corresponding author:

Prof. Carsten Carlberg

M

#

U

Finland

School of Medicine

ED

Institute of Biomedicine

A

CC E

PT

University of Eastern Finland POB 1627 FI-70211 Kuopio Tel.: +358-40-355-3062 E-mail: [email protected]

Carlberg et al.: In vivo response of the human epigenome to vitamin D

2

Highlights One individual was exposed three times every 28 days to an oral bolus (2,000 µg) of vitamin D3. At nine time points epigenome-wide chromatin accessibility was assessed by FAIRE-seq. 853 most prominent genomic loci were classified into early, delayed and non-responding

SC R

IP T

A rather minor rise in 25(OH)D3 serum levels results in significant changes of the epigenome.

ABSTRACT

In vitro cell culture studies showed that the hormonal form of vitamin D3, 1α,25-

U

dihydroxyvitamin D3, significantly (p < 0.05) affects the human epigenome at thousands of

N

genomic loci. Phase II of the VitDbol vitamin D intervention trial (NCT02063334) involved a

A

proof-of-principle study of one individual, who was exposed three times every 28 days to an

M

oral bolus (2,000 µg) of vitamin D3. Blood samples were taken directly before each supplementation as well as one and two days after, chromatin was isolated from peripheral

epigenome-wide

ED

blood mononuclear cells without any further in vitro culture and at all nine time points chromatin

accessibility

was

assessed

by

applying

FAIRE-seq

PT

(formaldehyde-assisted isolation of regulatory elements sequencing). The vitamin D3 bolus resulted in an average raise in 25-hydroxyvitamin D3 (25(OH)D3) serum concentration of 11.9

CC E

and 19.4 nM within one and two days, respectively. Consistently accessible chromatin was detected at 5,205 genomic loci, the 853 most prominent of which a self-organizing map algorithm classified into early, delayed and non-responding genomic regions: 70 loci showed

A

already after one day and 361 sites after two days significant (p < 0.0001) chromatin opening or closing. Interestingly, more than half of these genomic regions overlap with transcription start sites, but the change of chromatin accessibility at these sites has no direct effect on the transcriptome. Some of the vitamin D responsive chromatin sites cluster at specific loci within the human genome, the most prominent of which is the human leukocyte antigen region in chromosome 6. In conclusion, this study demonstrates that under in vivo conditions

Carlberg et al.: In vivo response of the human epigenome to vitamin D

3

a rather minor rise in 25(OH)D3 serum levels is sufficient to result in significant changes at hundreds of sites within the epigenome of human leukocytes.

Keywords: Vitamin D3; epigenome; vitamin D3 bolus supplementation; accessible chromatin;

A

CC E

PT

ED

M

A

N

U

SC R

IP T

PBMCs; FAIRE-seq; SOM; HLA cluster.

Carlberg et al.: In vivo response of the human epigenome to vitamin D

4

ABBREVIATIONS 1α,25-dihydroxyvitamin D3

25(OH)D3

25-hydroxyvitamin D3

APOO

apolipoprotein O

CLIP4

CAP-Gly domain containing linker protein family member 4

EHMT2

euchromatic histone lysine methyltransferase 2

FAIRE-seq

formaldehyde-assisted isolation of regulatory elements sequencing

FBF1

Fas binding factor 1

FDR

false discovery rate

FE

fold enrichment

FKBPL

FK506 binding protein like

GEO

Gene Expression Omnibus

HLA

human leukocyte antigen

IGV

Integrative Genomics Viewer

MAPK8

mitogen-activated protein kinase 8

SC R U

major histocompatibility complex peripheral blood mononuclear cell

SOAT1

sterol O-acyltransferase 1

SOM

self-organizing map

A

N

A

M

ED

CC E

PBMC

PT

MHC

IP T

1,25(OH)2D3

SUN1

Sad1 and UNC84 domain containing 1

TBCCD1

TBCC domain containing 1

TNF

tumor necrosis factor

TSS

transcription start site

Carlberg et al.: In vivo response of the human epigenome to vitamin D

5

vitamin D receptor

WRAP73

WD repeat containing, antisense to TP73

A

CC E

PT

ED

M

A

N

U

SC R

IP T

VDR

Carlberg et al.: In vivo response of the human epigenome to vitamin D

6

INTRODUCTION Vitamin D3 is a pre-hormone that has via its metabolite 1α,25-dihydroxyvitamin D3 (1,25(OH)2D3), which is binding with high affinity to the transcription factor vitamin D receptor (VDR), direct effects on gene expression in most human tissues and cell types [1, 2]. The natural way of producing vitamin D3 is its synthesis from 7-dehydrocholesterol in UV-B

IP T

exposed skin [3], but because of the increased indoor activities, skin protection by clothing or sunscreen use or life at latitudes above 40°, most individuals have to take up the molecule via

SC R

diet or direct supplementation with pills [4].

Serum 25(OH)D3 concentrations serve as a biomarker of the vitamin D status [5] and a level below 50 nM (20 ng/ml) is defined as vitamin D insufficiency [6]. Vitamin D has a number of

U

physiological functions, such as maintaining the balance of calcium and phosphorus

N

homeostasis, controlling innate and adaptive immunity and modulating cellular growth [7-9].

A

A low vitamin D status can have musculoskeletal consequences, such as rickets in children and osteomalacia and fractures in adults [10]. Moreover, vitamin D insufficiency is linked to

M

multiple sclerosis, type 2 diabetes, cardiovascular disease and cancers of the breast, prostate

ED

and colon [11-15].

Chromatin is a complex of nucleosomes around which genomic DNA is wrapped [16]. The

PT

default state of chromatin is densely packed heterochromatin that protects genes from uncontrolled activation [17]. Thus, chromatin is only accessible at selected regions, such as

CC E

enhancers and transcription start sites (TSSs) of those genes that need to be expressed in a given tissue or cell type [18]. Accessible chromatin loci can be detected genome-wide by the method FAIRE-seq [19]. For example, in THP-1 human monocytes 1,25(OH)2D3 changes

A

significantly (p < 0.05) chromatin accessibility at nearly 9,000 genomic loci [20]. Maximal chromatin opening was observed after 24 h, while after 48 h most sites are again closed. This indicates that the primary effect of activating VDR by 1,25(OH)2D3 is a transient opening or closing of chromatin at specific enhancer and TSS regions, which eventually results in activation or repression of vitamin D target genes [21]. There are a few hundred primary target genes of VDR and 1,25(OH)2D3 [2], i.e. these genes change their expression already

Carlberg et al.: In vivo response of the human epigenome to vitamin D

7

within a few hours after VDR activation. However, for a most comprehensive insight on the physiological actions of vitamin D also long-term, secondary effects on gene expression in the time frame of 1-2 days have to be taken into account. In order to study vitamin D-dependent gene regulation under in vivo conditions in human, we designed the VitDbol vitamin D intervention trial (NCT02063334, ClinicalTrials.gov). In

IP T

phase I of the study, 35 healthy young adults were exposed once to a vitamin D3 bolus (2,000 µg) and the change of chromatin accessibility at selected genomic regions [22] and the

SC R

induction of a few primary vitamin D target genes [23] was measured from peripheral blood mononuclear cells (PBMCs) at days 0, 1 and 2. This demonstrated that a single vitamin D3 bolus is sufficient to activate genes. This result is in line with the study of Hossein-nezhad et

U

al. [24], which indicated that two months of daily vitamin D3 supplementation with either 10 or 50 µg resulted in the significant (p < 0.01) regulation of 291 genes in PBMCs of in total

N

eight study participants. Similarly, a weekly supplementation of 47 subjects with 500 µg

A

vitamin D3 over 3-5 years resulted in the significant (p < 0.05) regulation of 99 genes in whole

M

blood compared to 47 placebo treated individuals [25].

ED

In the study presented here, we show results from phase II of VitDbol, in which one individual was exposed three times in a row every 28 days with a vitamin D3 bolus. Epigenome-wide chromatin accessibility within PBMCs was measured at days 0, 1 and 2.

PT

From 5,205 consistently detected sites of accessible chromatin 70 loci showed already after

CC E

one day and 361 sites after two days a significant (p < 0.0001) response to vitamin D supplementation. This proof-of-principle study demonstrates that under in vivo conditions a rise in 25(OH)D3 serum levels results in significant changes at hundreds of sites within the

A

epigenome of human PBMCs.

MATERIAL AND METHODS Sample collection

Carlberg et al.: In vivo response of the human epigenome to vitamin D

8

In phase II of VitDbol one individual (male, 50 years, body mass index 24.8) was exposed three times every 28 days to a vitamin D3 bolus (2,000 µg). The study took place between February and April 2014, i.e. during a period of little or no UVB exposure from the sunlight in Finland. The research ethics committee of the Northern Savo Hospital District had approved the study protocol (#9/2014).

IP T

Blood samples for serum and PBMC isolation were collected after overnight (12 h) fasting of the study participant at days 0, 1, 2, 28, 29, 30, 56, 57 and 58. Serum 25(OH)D3

electrode array as described previously [26].

U

FAIRE-seq and its analysis

SC R

concentrations were measured using a high performance liquid chromatography/coulometric

N

PBMCs were isolated within one hour after blood draw from 8 ml of peripheral blood in a

A

Vacutainer CPT Cell Preparation Tube with sodium citrate (Becton Dickinson) according to

M

the manufacturer’s instructions. FAIRE was performed according to the protocol published by Giresi et al. [27] with some modifications [20]. In short, 7.5 x 106 PBMCs were cross-linked

ED

for 10 min and stopped with glycine. The washed cell pellets were resuspended and incubated sequentially in 750 µl of buffer L1, 750 µl of buffer L2 and 300 µl of buffer L3. The lysates

PT

were sonicated in a Bioruptor Plus (Diagenode) to result in DNA fragments of 200 to 500 bp and cellular debris was removed by centrifugation. Aliquots (40 µl) of the sonicated lysates

CC E

were reverse cross-linked and proteinase K-digested overnight at 65 °C, in order to prepare input samples. FAIRE samples were subjected to two sequential phenol/chloroform/isoamyl alcohol (25/24/1) extractions, resuspended in 10 mM Tris-HCl (pH 7.4) and treated with 1 µl

A

of RNase A (10 mg/ml) for 1 h at 37 °C. The samples were then reverse cross-linked and proteinase K-digested for 2 h at 65 °C. Genomic DNA was purified from input and FAIRE samples using the ChIP DNA Clean&Concentrator Kit (Zymo Research). Purified genomic DNA samples were sequenced at 50 bp read length using standard manufacturer protocols at the Gene Core at the EMBL (Heidelberg, Germany). For FAIRE-

Carlberg et al.: In vivo response of the human epigenome to vitamin D

9

seq data analysis statistically significant peaks were identified using the Zinba program package version 2.02 by setting the mean fragment length at 200 bp and using other settings as

recommended

for

FAIRE-seq

in

the

Zinba

website

(http://code.google.com/p/zinba/wiki/UsingZINBA) including peak refinement [28]. In order to enable comparisons across all nine time points, common peak borders were identified by

IP T

detecting the local maxima of original peak border density and filtering out all resulting intervals that were > 2 kb. The peak statistics (except the false discovery rate (FDR)) were

recalculated as done in MACS2 version 2.0.9 and all intervals, where the maximal fold

SC R

enrichment (FE) over the entire data set was < 3, were filtered out. FAIRE-seq raw data are

U

available at Gene Expression Omnibus (GEO; www.ncbi.nlm.nih.gov/geo) at GSE106162.

N

Functional characterization of FAIRE peak loci

A

5,205 FAIRE-seq peaks were identified at all nine time points (Table S1). In order to obtain

M

higher accuracy in data integration, persistent peaks fulfilling more stringent conditions were selected by the threshold of a likelihood function, which takes into account i) a p-value

ED

< 0.05, ii) the FE with respect to input, iii) the number of sequencing reads and iv) the shape of the peak: 1 - k x 1/normalized FE – r/R, where k and r are normalization factors, and R is

PT

the proportion of reads that are piled up very close to the peak’s summit giving the peak the shape of a rectangle as opposed to the expected Gaussian or Poisson-like shape. The

CC E

Integrative Genomics Viewer [29] was used to visualize FAIRE-seq data at example genomic regions.

For a better characterization of the dynamics of the resulting 853 FAIRE peaks, self-

A

organizing maps (SOM) were applied. SOM [30] is an unsupervised method that allows the identification of representative profiles of the inspected high-dimensional data. The method was performed with the peak FE at all nine time points, i.e. the SOM vector represented a nine-dimensional input space. SOM segregated the 853 most prominent FAIRE peaks into one of nine possible classes in a 3x3 lattice. This allows a visual inspection of the similarities

Carlberg et al.: In vivo response of the human epigenome to vitamin D

10

of chromatin accessibility as a function of time and provides grounds for more elaborated statistical interrogations. The nine SOM classes differed along several lines. Firstly, the number of peaks in each SOM class was not homogeneous, varying from 13 to 360. This large variation is a sign of nonrandomness in the data, since the clustering of a random collection of peaks would have

IP T

produced groups of similar size. Secondly, the profiles were different not only in terms of the peak FE, but also in terms of the dynamics along the nine measured days, which is shown as

SC R

the different shapes depicted by the bars in the SOM classes. Finally, while some of the profiles of the SOM classes showed a cyclic behavior with clear statistical differences

compared to the chromatin state at day 0, others could be described by a less variable profile

N

U

as a function of time.

A

Statistics

M

The segregation of the 853 most prominent accessible chromatin regions into nine different classes allowed hypothesis testing. In order to test for significant changes between day 0 and

ED

the subsequent days for a given profile, a Wilcoxon-Mann-Whitney test (U-test) was applied. The test is less restricted than the Student’s T-test, since among other features, it allows data

PT

to be described by any distribution, thus being non-parametric. To test for statistical differences between a given day and day 0, a U-test was applied for the peak FE at a given

CC E

day and the FE at day 0 for all peaks in the class. Once the U-test and its corresponding pvalue were obtained, the ratio between the peak FE at the given day and the peak FE at the corresponding day 0 for all peaks in the class was calculated. If the average of this ratio was

A

higher than 1.5 and the p-value < 0.01, a statistically significant difference was declared.

Carlberg et al.: In vivo response of the human epigenome to vitamin D

11

RESULTS VitDbol phase II study The focus of phase II of the vitamin D intervention study VitDbol was the genome-wide analysis of RNA and chromatin samples that were obtained from PBMCs of healthy adults one or two days after they had been exposed to a vitamin D3 bolus (2,000 µg once). An

IP T

essential part of phase II is the here presented proof-of-principle study, in which in a

consecutive experiment one individual was exposed at days 0, 28 and 56 with a vitamin D3

SC R

bolus. PBMCs were isolated immediately after blood samples were drawn at days 0, 1, 2, 28,

29, 30, 56, 57 and 58. 25(OH)D3 serum levels increased on average by 11.9 ± 2.8 nM after one day and 19.4 ± 5.4 nM after two days of vitamin D3 bolus supplementation (Fig. 1).

U

These rises were slightly lower than the averages of 17.4 and 21.1 nM that were determined

N

after one and two days, respectively, for the 27 vitamin D3 bolus supplemented participants of

M

A

phase I of VitDbol (the 8 remaining participants of VitDbol received placebo) [22].

In vivo responsiveness of the human epigenome to vitamin D

ED

FAIRE-seq analysis was performed with chromatin samples that were isolated from PBMCs obtained from the triplicate vitamin D3 bolus supplementation experiment (Fig. 1). In total

PT

5,205 genomic regions showed at all nine time points a significant (FDR > 0.05) FE in reference to input reference, i.e. these sites showed consistently accessible chromatin

CC E

throughout the whole in vivo experiment (Table S1). A peak likelihood value of more than 0.7 turned out to be a meaningful threshold, since at this value a prominent change in the proportion of affected sites was observed (Fig. S1). The proportion of peaks is described by

A

an exponential decrease for likelihoods lower than 0.7, but at that value, an abrupt change occurs, giving indication that above this specific point, the nature of the peaks is different to that of the peaks in the previous range. This led to a list of 853 most prominent sites of accessible chromatin in PBMCs (Table S1). Then SOM with a 3x3 lattice was applied, in order to segregate the 853 genomic regions into nine classes of most similar properties

Carlberg et al.: In vivo response of the human epigenome to vitamin D

12

concerning their chromatin accessibility over time (Fig. 2). The significance of the change in chromatin accessibility was determined for the average of the members of each SOM class at each time point in relation to the respective status at day 0. In almost all classes, it was observed an increase in the subsequent days after the bolus ingestion. However, the statistically significant effects are less common. The 14 members of

IP T

SOM class 1 showed a significant change exclusively at day 29, i.e. at day 1 of the second intervention series, while the 56 members of class 3 reproducibly displayed increased

SC R

chromatin accessibility at days 1, 29 and 57, i.e. at the first day of each of the three series (Fig. 2). We interpret the 70 genomic regions in classes 1 and 3 as early responding regions of the human epigenome to vitamin D stimulation. In contrast, the 13, 39 and 83 members of

U

classes 4, 5 and 7, respectively, showed a significant and reproducible change in chromatin access at days 2, 30 and 58, i.e. at the second day of all three series. Moreover, the

N

226 members of class 8 responded only at days 2 and 58, i.e. at the second day of series 1 and

A

3, and displayed a rather low average level in chromatin accessibility. We consider the

M

361 genomic loci of these four classes as delayed vitamin D responding regions of the epigenome. Finally, the 30, 32 and 360 members of classes 2, 6 and 9 did not show any

ED

significant change in their chromatin accessibility and are referred to as chromatin loci that are non-responsive to vitamin D.

PT

Interestingly, we observed in none of the nine SOM classes a significant difference in the

CC E

chromatin accessibility when comparing day 0 with days 28 and 56. This means that even at the early and delayed responsive regions the chromatin status had returned within 28 days back to their ground state, i.e. the chromatin opening effects of the vitamin D3 bolus on the

A

human epigenome were primarily transient.

TSS overlap of accessible chromatin regions From the complete set of 5,205 consistently accessible genomic sites 1,688 (32.4%) overlapped with a TSS region (Table S1). This rate significantly increased to 58.7% (502 of

Carlberg et al.: In vivo response of the human epigenome to vitamin D

13

853) for most prominent accessible chromatin loci (2 = 141, p-value = 1e-32). Representative examples of early responding chromatin regions are those on the TSS of the FBF1 (Fas binding factor 1) gene (Fig. 3A), close to the TSS of the SUN1 (Sad1 and UNC84 domain containing 1) gene (Fig. 3B) and on the TSS of the WRAP73 (WD repeat containing, antisense to TP73) gene (Fig. 3C). In contrast, examples of the delayed responding chromatin

IP T

loci were those on the TSS regions of the genes MAPK8 (mitogen-activated protein kinase 8, Fig. 3D), TBCCD1 (TBCC domain containing 1, Fig. 3E) and FKBPL (FK506 binding

SC R

protein like, Fig. 3F).

In total 229 of the 431 vitamin D responsive chromatin sites overlapped with a TSS region. With 53.1% this rate is high, but still slightly lower than that of all prominent sites of

U

accessible chromatin in PBMCs. Moreover, only a minority of the genes related to the overlapping TSS regions are known as vitamin D target genes, such as SOAT1 (sterol O-

N

acyltransferase 1), CLIP4 (CAP-Gly domain containing linker protein family member 4) and

A

APOO (apolipoprotein O) in THP-1 cells [31]. Thus, in most cases the overlap of a TSS with

M

a vitamin D responsive chromatin region seems not to have any functional consequence on gene expression, i.e. in this in vivo experiment epigenome changes were not directly coupled

PT

ED

with alterations of the transcriptome.

Genome-wide distribution of vitamin D responsive chromatin regions

CC E

The genome-wide display of the average FE of the 853 most prominent sites of chromatin accessibility over the nine time points of the triplicate vitamin D3 bolus experiment visualized that the 70 early vitamin D responsive regions were on average more prominent than the

A

delayed responsive loci (Fig. 4). The latter sites showed in turn a higher average FE than the non-responsive regions. In particular, when normalized by TSS density, which is highest at chromosome 19, there was no striking clustering of the sites of chromatin accessibility. Nevertheless, the genomic region around the human leukocyte antigen (HLA) cluster in chromosome 6 showed the highest normalized density of accessible chromatin. This region

Carlberg et al.: In vivo response of the human epigenome to vitamin D

14

contains not only the genes for the different subtypes of the major histocompatibility complex (MHC), but also those for key cytokines, such as TNF (tumor necrosis factor), and other important signaling molecules, such as the lysine methyltransferase EHMT2 (euchromatic histone lysine methyltransferase 2). For example, a 1 Mb sub-region of the HLA cluster contains six prominent sites of chromatin accessibility, three of which were significantly

IP T

vitamin D responsive (Fig. 5).

SC R

DISCUSSION

This study refers to phase II of the VitDbol vitamin D3 intervention trial. In a proof-ofprinciple it demonstrates that a vitamin D3 bolus can reproducibly induce significant

U

(p < 0.0001) changes at hundreds of sites within the human epigenome. Importantly, the

N

results were obtained within 1-2 days under natural conditions and with minimal harm to the

A

study subject, i.e. the design of the VitDbol trial allows safe human in vivo experiments.

M

The results of this study confirm our observations from in vitro cell culture experiments with THP-1 cells, where the biologically active vitamin D3 metabolite 1,25(OH)2D3 changed

ED

chromatin accessibility within the same time frame (48 h) than in this study [20]. However, in THP-1 cells nearly 9,000 genomic sites responded significantly (p < 0.05) to VDR ligand

PT

treatment. Observing a far higher number of vitamin D-responsive genomic regions in vitro than in vivo can have a number of different reasons: The reference point of the in vitro experiment are vitamin D-depleted cells, while the in

CC E

i)

vivo experiment refers to the vitamin D status of the individual before vitamin D3 bolus supplementation, i.e. 25(OH)D3 serum levels of 37-78 nM.

A

ii) THP-1 cells were treated with a high concentration (100 nM) of the VDR ligand 1,25(OH)2D3 in the culture medium, while in vivo vitamin D3 needs to be first absorbed and then enzymatically converted to 1,25(OH)2D3. iii) Comparison of transcriptome profiles indicated that monocytes, such as THP-1 cells, have a higher number of vitamin D target genes [20], i.e. they seem to be more

Carlberg et al.: In vivo response of the human epigenome to vitamin D

15

responsive to vitamin D, than other components of PBMCs, such as T and B cells [32]. Thus, investigating PBMCs, in which monocytes only represent some 10%, dilutes their contribution. The rather modest rise in 25(OH)D3 serum levels of less than 20 nM per bolus supplementation may be due to incomplete conversion of vitamin D3 into 25(OH)D3 within

IP T

2 days in contrast to adaptions to vitamin D supplementation that often take 6 months [33]. Nevertheless, many immune cells have the capability to metabolize 25(OH)D3 into

SC R

1,25(OH)2D3 [34, 35]. This implies that the vitamin D status, as measured via 25(OH)D3 serum levels, may not the best parameter for monitoring changes in the vitamin D signaling system of PBMCs.

U

Cell culture conditions are always constant concerning environmental challenges, such as

N

nutrient availability and temperature, while in vivo individuals are exposed to a large number

A

of variant challenges. This also includes nutritional measures, such as variations in phases of fasting and meal intake. Accordingly, full reproducibility is impossible to achieve with human

M

in vivo experiments. Therefore, it is remarkable that 85.9% of the prominent sites of

ED

chromatin accessibility in PBMCs are also found in THP-1 cells [20]. Moreover, this reemphasizes the impact of monocytes within the epigenomic profile of PBMCs.

PT

Already in THP-1 cells we observed that the response of the human epigenome to a challenge with the vitamin D metabolite 1,25(OH)2D3 is largely transient and has its maximum 24 h

CC E

after onset of VDR ligand stimulation [20]. Since the absorption, transport and enzymatic conversion of vitamin D3 to the active VDR ligand takes some time, in the in vivo experiment the maximal vitamin D response of the PBMC epigenome occurs rather after 2 days.

A

Nevertheless, also in vivo the effect of vitamin D on the epigenome seems to be transient. In this context it should be reminded that under conditions of a natural life a fast increase of the vitamin D status is rare and may occur only in light skin individuals after excessive sun bathing. Thus, future long-term vitamin D3 intervention studies should indicate, whether a daily lower dose vitamin D3 supplementation or a limited daily sun exposure during summer result in more persistent effects of vitamin D on the human epigenome.

Carlberg et al.: In vivo response of the human epigenome to vitamin D

16

Together with cell-specific enhancer loci, TSS regions belong to the most critical sites within the human genome that need to be controlled by chromatin accessibility. Therefore, it is not surprising that more than half of the most prominent sites of chromatin accessibility in PBMCs overlap with a TSS region. Since 51.9% of all prominent sites are responsive to vitamin D, many of these loci also overlap with a TSS region. In a first view this may suggest

IP T

that the respective genes are direct target of vitamin D. However, also in THP-1 cells some 20% of the 1,25(OH)2D3-responsive chromatin sites overlap with a TSS region, but only 8.3% of the latter pointed to a vitamin D target gene [20]. This suggests that more complex gene

SC R

regulatory scenarios apply, such as vitamin D responsiveness of topologically associated domains [20, 31, 36]. So far it can only be concluded that the vitamin D response of chromatin accessibility represents a kind of memory function of the epigenome that needs

U

further investigations.

N

Basically all cellular components of PBMCs belong to the innate and adaptive immune

A

system. Therefore, it is not surprising that the immunologically most important region of the

M

human genome, the HLA cluster, also highlights as a “hotspot” in the epigenome of PBMCs. However, it is remarkable that the HLA cluster is also a focused region of the vitamin D

ED

responsiveness of the epigenome. This observation provides a strong link to the impact of vitamin D on the control of the immune system [37].

PT

In conclusion, in this proof-of-principle study we demonstrated that under in vivo conditions a

CC E

rather minor rise in 25(OH)D3 serum levels results in significant changes at hundreds of sites within the epigenome of human leukocytes.

A

ACKNOWLEDGEMENTS

This work was supported by the Academy of Finland (grant No. 267067) and the Juselius Foundation (both to C.C.). We acknowledge the help of Noora Saksa in PBMC isolation. Kind thanks to the Gene Core Facility at the EMBL in Heidelberg, Germany, for massively parallel sequencing services.

Carlberg et al.: In vivo response of the human epigenome to vitamin D

17

AUTHOR CONTRIBUTION A.N. did the computational analysis, S.S. and T.N. performed the experiments, A.N. and C.C. analyzed the data, C.C., J.V. and T.-P.T. designed and performed the VitDbol study and C.C. wrote the manuscript. All authors gave an intellectual contribution to the study and took part

IP T

in editing the manuscript.

SC R

REFERENCES

A

CC E

PT

ED

M

A

N

U

[1] C. Carlberg, Genome-wide (over)view on the actions of vitamin D, Front Physiol, 5 (2014) 167. [2] M.J. Campbell, Vitamin D and the RNA transcriptome: more than mRNA regulation, Front Physiol, 5 (2014) 181. [3] M.F. Holick, J.E. Frommer, S.C. McNeill, N.M. Richtand, J.W. Henley, J.T. Potts, Jr., Photometabolism of 7-dehydrocholesterol to previtamin D3 in skin, Biochem Biophys Res Commun, 76 (1977) 107-114. [4] M.F. Holick, Vitamin D deficiency in 2010: health benefits of vitamin D and sunlight: a D-bate, Nat Rev Endocrinol, 7 (2011) 73-75. [5] B.W. Hollis, Circulating 25-hydroxyvitamin D levels indicative of vitamin D sufficiency: implications for establishing a new effective dietary intake recommendation for vitamin D, J Nutr, 135 (2005) 317-322. [6] Institute-of-Medicine, Dietary reference intakes for calcium and vitamin D, Washington, DC: National Academies Press, (2011). [7] A. Verstuyf, G. Carmeliet, R. Bouillon, C. Mathieu, Vitamin D: a pleiotropic hormone, Kidney Int, 78 (2010) 140-145. [8] D. Feldman, A.V. Krishnan, S. Swami, E. Giovannucci, B.J. Feldman, The role of vitamin D in reducing cancer risk and progression, Nat Rev Cancer, 14 (2014) 342-357. [9] C. Carlberg, The physiology of vitamin D-far more than calcium and bone, Front Physiol, 5 (2014) 335. [10] R. Bouillon, G. Carmeliet, L. Verlinden, E. van Etten, A. Verstuyf, H.F. Luderer, L. Lieben, C. Mathieu, M. Demay, Vitamin D and human health: lessons from vitamin D receptor null mice, Endocr Rev, 29 (2008) 726-776. [11] E. Hypponen, E. Laara, A. Reunanen, M.R. Jarvelin, S.M. Virtanen, Intake of vitamin D and risk of type 1 diabetes: a birth-cohort study, Lancet, 358 (2001) 1500-1503. [12] T.J. Wang, M.J. Pencina, S.L. Booth, P.F. Jacques, E. Ingelsson, K. Lanier, E.J. Benjamin, R.B. D'Agostino, M. Wolf, R.S. Vasan, Vitamin D deficiency and risk of cardiovascular disease, Circulation, 117 (2008) 503-511. [13] E. Giovannucci, Y. Liu, B.W. Hollis, E.B. Rimm, 25-hydroxyvitamin D and risk of myocardial infarction in men: a prospective study, Arch Int Med, 168 (2008) 1174-1180. [14] C.J. Rosen, J.S. Adams, D.D. Bikle, D.M. Black, M.B. Demay, J.E. Manson, M.H. Murad, C.S. Kovacs, The nonskeletal effects of vitamin D: an Endocrine Society scientific statement, Endocr Rev, 33 (2012) 456-492.

Carlberg et al.: In vivo response of the human epigenome to vitamin D

18

A

CC E

PT

ED

M

A

N

U

SC R

IP T

[15] C.F. Garland, E.D. Gorham, S.B. Mohr, F.C. Garland, Vitamin D for cancer prevention: global perspective, Ann Epidemiol, 19 (2009) 468-483. [16] K. Helin, D. Dhanak, Chromatin proteins and modifications as drug targets, Nature, 502 (2013) 480-488. [17] Z.D. Smith, A. Meissner, DNA methylation: roles in mammalian development, Nat Rev Genet, 14 (2013) 204-220. [18] A.I. Badeaux, Y. Shi, Emerging roles for chromatin as a signal integration and storage platform, Nat Rev Mol Cell Biol, 14 (2013) 211-224. [19] P.G. Giresi, J. Kim, R.M. McDaniell, V.R. Iyer, J.D. Lieb, FAIRE (FormaldehydeAssisted Isolation of Regulatory Elements) isolates active regulatory elements from human chromatin, Genome Res, 17 (2007) 877-885. [20] S. Seuter, A. Neme, C. Carlberg, Epigenome-wide effects of vitamin D and their impact on the transcriptome of human monocytes involve CTCF, Nucleic Acids Res, 44 (2016) 4090-4104. [21] C. Carlberg, F. Molnár, Vitamin D receptor signaling and its therapeutic implications: Genome-wide and structural view, Canadian journal of physiology and pharmacology, 93 (2015) 311-318. [22] S. Seuter, J.K. Virtanen, T. Nurmi, J. Pihlajamäki, J. Mursu, S. Voutilainen, T.P. Tuomainen, A. Neme, C. Carlberg, Molecular evaluation of vitamin D responsiveness of healthy young adults, J Steroid Biochem Mol Biol, 174 (2017) 314-321. [23] M. Vukic, A. Neme, S. Seuter, N. Saksa, V.D. de Mello, T. Nurmi, M. Uusitupa, T.P. Tuomainen, J.K. Virtanen, C. Carlberg, Relevance of vitamin D receptor target genes for monitoring the vitamin D responsiveness of primary human cells, PLoS One, 10 (2015) e0124339. [24] A. Hossein-nezhad, A. Spira, M.F. Holick, Influence of vitamin D status and vitamin D3 supplementation on genome wide expression of white blood cells: a randomized double-blind clinical trial, PLoS One, 8 (2013) e58725. [25] Y. Pasing, C.G. Fenton, R. Jorde, R.H. Paulssen, Changes in the human transcriptome upon vitamin D supplementation, J Steroid Biochem Mol Biol, 173 (2017) 93-99. [26] T. Nurmi, T.P. Tuomainen, J. Virtanen, J. Mursu, S. Voutilainen, High-performance liquid chromatography and coulometric electrode array detector in serum 25-hydroxyvitamin D3 and 25-hydroxyvitamin D2 analyses, Anal Biochem, 435 (2013) 1-9. [27] P.G. Giresi, J.D. Lieb, Isolation of active regulatory elements from eukaryotic chromatin using FAIRE (Formaldehyde Assisted Isolation of Regulatory Elements), Methods, 48 (2009) 233-239. [28] N.U. Rashid, P. Giresi, J.G. Ibrahim, W. Sun, J.D. Lieb, ZINBA integrates local covariates with DNA-seq data to identify broad and narrow regions of enrichment, even within amplified genomic regions, Genome Biol, 12 (2011) R67. [29] H. Thorvaldsdottir, J.T. Robinson, J.P. Mesirov, Integrative Genomics Viewer (IGV): high-performance genomics data visualization and exploration, Brief Bioinf, 14 (2013) 178192. [30] T. Kohonen, Self-organizing maps, Series Inform Sci, 30 (1995). [31] A. Neme, S. Seuter, C. Carlberg, Selective regulation of biological processes by vitamin D based on the spatio-temporal cistrome of its receptor, Biochim Biophys Acta, 1860 (2017) 952-961. [32] S.V. Ramagopalan, A. Heger, A.J. Berlanga, N.J. Maugeri, M.R. Lincoln, A. Burrell, L. Handunnetthi, A.E. Handel, G. Disanto, S.M. Orton, C.T. Watson, J.M. Morahan, G. Giovannoni, C.P. Ponting, G.C. Ebers, J.C. Knight, A ChIP-seq defined genome-wide map of vitamin D receptor binding: associations with disease and evolution, Genome Res, 20 (2010) 1352-1360.

Carlberg et al.: In vivo response of the human epigenome to vitamin D

19

A

CC E

PT

ED

M

A

N

U

SC R

IP T

[33] M. Holick, Vitamin D the underappreciated D-lightful hormone that is important for skeletal and cellular health, Curr Opin Endocrinol Diabetes, 9 (2002) 87-98. [34] J.S. Adams, F.R. Singer, M.A. Gacad, O.P. Sharma, M.J. Hayes, P. Vouros, M.F. Holick, Isolation and structural identification of 1,25-dihydroxyvitamin D3 produced by cultured alveolar macrophages in sarcoidosis, J Clin Endocrinol Metabol, 60 (1985) 960-966. [35] M. Hewison, L. Freeman, S.V. Hughes, K.N. Evans, R. Bland, A.G. Eliopoulos, M.D. Kilby, P.A.H. Moss, R. Chakraverty, Differential regulation of vitamin D receptor and its ligand in human monocyte-derived dendritic cells, J Immunol, 170 (2003) 5382-5390. [36] A. Neme, S. Seuter, C. Carlberg, Vitamin D-dependent chromatin association of CTCF in human monocytes, Biochim Biophys Acta, 1859 (2016) 1380-1388. [37] C. Aranow, Vitamin D and the immune system, J Invest Med, 59 (2011) 881-886.

Carlberg et al.: In vivo response of the human epigenome to vitamin D

20

FIGURE LEGENDS

Fig. 1: Triplicate vitamin D3 bolus supplementation experiment. In a consecutive experiment one individual was exposed at days 0, 28 and 56 with a vitamin D3 bolus of 2,000 µg (blue arrows). The rise in 25(OH)D3 serum levels at days 0, 1, 2, 28, 29, 30, 56, 57

IP T

and 58 is displayed.

Fig. 2: SOM analysis of the most prominent sites of accessible chromatin. The 853 most

SC R

prominent sites of accessible chromatin in PBMCs (Table S1) were segregated by SOM into

nine classes of most similar FE over time. A Wilcoxon-Mann-Whitney U-test was applied, in order to determine the significance of the change in chromatin accessibility for the average of

U

the members of each SOM class at each day of the intervention in relation to the respective

N

status at day 0 (* p < 0.01; ** p < 0.001; *** p < 0.0001). An additional essential condition

A

was a fold change of more than 1.5 in relation to day 0. The green horizontal line indicates the

M

average level of chromatin accessibility in each class.

ED

Fig. 3: Time-resolved chromatin opening of vitamin D responsive chromatin regions. The IGV browser was used to display normalized FAIRE-seq data from PBMCs isolated at

PT

the indicated days of the triplicate vitamin D3 bolus supplementation experiment.

CC E

Representative examples of early (A-C) and delayed (D-F) vitamin D responsive chromatin regions are shown in an identical 22 kb genomic window. The respective SOM classes (Table S1 and Fig. 4) are indicated. Gene structures are shown in blue and significantly vitamin D responsive genomic regions are shaded in grey. Please note the adaption of the

A

peak range, in order to filter background signals.

Fig. 4: Genome-wide view on prominent sites of chromatin accessibility. Genome-wide display of the average FE of the 853 prominent sites of chromatin accessibility, color-coded for the nine SOM classes (Table S1). The positions of the six example regions shown in

Carlberg et al.: In vivo response of the human epigenome to vitamin D

21

Fig. 3 are marked and the HLA cluster (Fig. 5) is shaded in grey. Horizontal lines indicate the average of all 70 early responding regions (red), 361 delayed responding loci (blue) and 422 non-responding sites (grey).

Fig. 5: Genomic view on chromatin accessibility at the HLA cluster. The IGV browser

IP T

was used to display normalized FAIRE-seq data from PBMCs isolated at the indicated days

of the triplicate vitamin D3 bolus supplementation experiment. A 1 Mb region of the HLA

SC R

cluster is shown and prominent sites of chromatin accessibility are shaded in grey. The respective SOM classes (Table S1 and Fig. 4) are indicated and gene structures are shown in

A

CC E

PT

ED

M

A

N

U

blue. Please note the adaption of the peak range, in order to filter background signals.

N U SC

Figure 1

M

A

Fig. 1

A

CC E

PT

ED

2,000 µg

2,000 µg

2,000 µg

Figure 2

Fig. 2 Class 2 (n = 30)

Class 1 (n = 14)

Class 3 (n = 56)

***

15 12 9

3

2 28 29 30 56 57 58

0

1

Class 4 (n = 13) 15

***

12

**

2 28 29 30 56 57 58

Class 5 (n = 39) * ***

***

6

2 28 29 30 56 57 58

Class 6 (n = 32)

***

1

2 28 29 30 56 57 58

3 0

1

1

2 28 29 30 56 57 58

Class 9 (n = 360)

***

2 28 29 30 56 57 58

CC

0

***

EP

***

6

0

TE

9

2 28 29 30 56 57 58

Class 8 (n = 226)

15 12

1

D

Class 7 (n = 83)

0

M

0

A

3 0

1

N

9

0

SC R

1

***

U

0

A

FAIRE peak fold enrichment

0

***

IP T

***

6

*** 0

1

***

2 28 29 30 56 57 58

Day of vitamin D intervention

0

1

2 28 29 30 56 57 58

Figure 3

Fig. 3 A

B Chr 17:

73,930 kb

73,940 kb

Chr 7:

early (SOM1)

[0.2-1.2]

day 0

850 kb

early (SOM1)

[0.2-1.2]

[0.2-1.2]

[0.2-1.2]

[0.2-1.2]

[0.2-1.2]

[0.2-1.2]

[0.2-1.2]

[0.2-1.2]

[0.2-1.2]

[0.2-1.2]

[0.2-1.2]

[0.2-1.2]

[0.2-1.2]

[0.2-1.2]

[0.2-1.2]

[0.2-1.2]

[0.2-1.2]

860 kb

day 1 day 2

day 57 day 58

FBF1

[25(OH)D]

ACOX1

C

D 3,560 kb

3,570 kb early (SOM3)

[0.2-1.2]

day 0

[0.2-1.2]

day 28

[0.2-1.2]

[0.2-1.2]

day 58 [25(OH)D]

WRAP73

CC

E

EP

[0.2-1.2]

A

Chr 3:

186,280 kb

FAIRE-seq from PBMCs

[0.2-1.2]

[0.2-1.2] [0.2-1.2] [0.2-1.2]

TP73

MAPK8

F 186,290 kb delayed (SOM4)

[0.2-2.5]

day 0

[0.2-1.2]

[0.2-1.2]

TE

[0.2-1.2]

day 57

49,520 kb delayed (SOM4)

[0.2-1.2]

D

[0.2-1.2]

day 56

49,510 kb

[0.2-1.2]

M

[0.2-1.2]

day 2

day 30

Chr 6:

32,090 kb

delayed (SOM8)

[0.2-1.2]

[0.2-2.5]

[0.2-1.2]

[0.2-2.5]

[0.2-1.2]

[0.2-2.5]

[0.2-1.2]

[0.2-2.5]

[0.2-1.2]

[0.2-2.5]

[0.2-1.2]

[0.2-2.5]

[0.2-1.2]

[0.2-2.5]

[0.2-1.2]

[0.2-2.5]

[0.2-1.2]

32,100 kb

day 1 day 2 day 28 day 29 day 30 day 56 day 57 day 58 [25(OH)D]

SUN1

[0.2-1.2]

[0.2-1.2]

day 1

day 29

Chr 10:

A

Chr 1:

SC R

day 56

U

day 30

N

day 29

IP T

day 28

TBCCD1

DNAJB11

ATF6B

FKBPL

M

A

N

U

SC R

IP T

Figure 4

Fig. 4

D

TE

FBF1

SUN1

EP CC

10

1 2 3 4 5 6 7 8 9

MAPK8

TBCCD1 Average early

A

Average enrichment

SOM class

HLA cluster

15

WRAP73

Average delayed

FKBPL

5

Average non responsive

0 1

2

3

4

5

6

7

8

9

10

11

Chromosome number

12

13

14 15 16 17 18 19 20 21 22

X

D

M

A

N

U

SC R

IP T

Figure 5

[0.2-1.2]

day 0

SOM9

SOM8 SOM8

31,600 kb SOM9

31,800 kb SOM9

32,000 kb SOM8

CC

[0.2-1.2]

31,400 kb

EP

Chr 6: 31,200 kb

TE

Fig. 5

day 1

[0.2-1.2]

day 2

A

[0.2-1.2]

day 28

[0.2-1.2]

day 29 [0.2-1.2]

day 30 [0.2-1.2]

day 56 [0.2-1.2]

day 57 [0.2-1.2]

day 58 [25(OH)D]

HLA-C

HLA-B MICA

HCP5 MCCD1 LST1 BAG6 ABHD16A NEU1 CFB TNXB ATF6B NOTCH4 MICB NFKBIL1 CSNK2B MSH5 HSPA1B C2 C4A FKBPL PRRT1 LTA TNF ABHD16A LSM2 EHMT2 DOM3Z CYP21A2 AGPAT1