The Post-Apoptotic Fate of RNAs Identified Through
High-Throughput Sequencing of Human Hair
Gloria K. Lefkowitz, Anandaroop Mukhopadhyay, Christopher Cowing-Zitron, Benjamin D. Yu*
Stem Cell Program, Division of Dermatology, Department of Medicine, Institute for Genomic Medicine, University of California San Diego, San Diego, California, United
States of America
The hair of all mammals consists of terminally differentiated cells that undergo a specialized form of apoptosis called
cornification. While DNA is destroyed during cornification, the extent to which RNA is lost is unknown. Here we find that
multiple types of RNA are incompletely degraded after hair shaft formation in both mouse and human. Notably, mRNAs and
short regulatory microRNAs (miRNAs) are stable in the hair as far as 10 cm from the scalp. To better characterize the post-
apoptotic RNAs that escape degradation in the hair, we performed sequencing (RNA-seq) on RNA isolated from hair shafts
pooled from several individuals. This hair shaft RNA library, which encompasses different hair types, genders, and
populations, revealed 7,193 mRNAs, 449 miRNAs and thousands of unannotated transcripts that remain in the post-
apoptotic hair. A comparison of the hair shaft RNA library to that of viable keratinocytes revealed surprisingly similar
patterns of gene coverage and indicates that degradation of RNA is highly inefficient during apoptosis of hair lineages. The
generation of a hair shaft RNA library could be used as months of accumulated transcriptional history useful for
retrospective detection of disease, drug response and environmental exposure.
Citation: Lefkowitz GK, Mukhopadhyay A, Cowing-Zitron C, Yu BD (2011) The Post-Apoptotic Fate of RNAs Identified Through High-Throughput Sequencing of
Human Hair. PLoS ONE 6(11): e27603. doi:10.1371/journal.pone.0027603
Editor: Johanna M. Brandner, University Hospital Hamburg-Eppendorf, Germany
Received August 10, 2011; Accepted October 20, 2011; Published November 16, 2011
Copyright: ? 2011 Lefkowitz et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: This work was supported by grants from the National Institutes of Health AR056667 and the California Institute for Regenerative Medicine RN2-00908,
and institutional funds for B.D.Y. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: email@example.com
Apoptosis is a cellular program utilized by multicellular
organisms to eliminate cells during development or in surveillance
of foreign or abnormal cells altered by viral infection or neoplastic
transformation [1,2]. Cells undergoing apoptosis undergo nuclear,
chromatin, and additional organelle changes induced by a cascade
of molecular signals involving several proteases and deoxyribonu-
cleases . Defects in programmed cell death results in the
abnormal accumulation of cells, altered morphogenesis during
development, and the persistence of transformed cells in disease.
The hair follicle is a model of apoptosis [4,5]. Programmed cell
death in the hair follicle occurs both during its normal growth and
differentiation and during an involutional stage called catagen.
Ultrastructural studies demonstrate that hair follicle cells undergo
a specialized form of apoptosis called cornification during terminal
differentiation . During this process, the nuclear membrane is
lost, and the chromatin becomes less coarse. In addition, evidence
of DNA damage as detected by direct end-labeling of nicked DNA
and indirect double-strand break activity have been identified
during early and late stages of hair differentiation [4,7,8]. Skin-
specific endonucleases such as DNase1L2 target genomic DNA
during cornification [3,9], and in the absence of DNase1L2, nuclear
DNA persists in the hair and causes hair fragility . The
external hair is defined as the hair shaft and consists largely of the
proteinaceous remnants of three cell types . The three hair
shaft cell types, the outer cuticle, cortex and central medulla,
originate from a self-renewing progenitor population called the
matrix . The hair matrix also produces supportive, non-hair
shaft cell types, which form a rigid sheath around the hair shaft
and enable the hair to exit through the skin. Much of the strength
of the hair shaft and sheath come from intermediate filament
proteins called keratins and keratin-associated proteins (KRTAPs)
which become crosslinked by several enzymes during terminal
While DNA degradation is a common hallmark of apoptosis,
the targeting of RNA for degradation during apoptosis is unclear.
The removal of RNA following apoptosis may be of importance as
released endogenous RNA appears to activate inflammation and
the innate immune response . In viable cells, multiple
mechanisms exist to regulate the homeostasis of mRNAs,
including nonsense-mediated decay, targeting of AU-rich ele-
ments, microRNA-mediated destabilization, and others [16,17].
During some forms of apoptosis, RNases with broad specificity are
upregulated such as IFN-gamma induced RNase L . In
addition, specific endonucleolytic activities, which target 28S
ribosomal RNA, have been observed during apoptosis and appear
to trigger independently of DNA degradation [19,20]. More
recently, an endonuclease RNase III, DICER, has been shown to
shift specificities between RNA to DNA during apoptosis .
These several observations indicate that during apoptosis, RNA
and DNA targeting are distinctly regulated but the final fate of
RNAs following apoptosis is unknown.
Here we investigate the extent to which RNAs survive apoptosis
by studying the external hair shaft of mice and humans. We
hypothesized that pools of small RNAs, in particular, regulatory
PLoS ONE | www.plosone.org1November 2011 | Volume 6 | Issue 11 | e27603
microRNAs (miRNAs), might preferentially survive destruction by
nucleases during the process of cornification, because of their
inherent stability . Using RNA sequencing and real-time
quantitative PCR, we instead find that many types of RNAs
survive apoptosis including miRNAs and mRNAs and persist long
after hair shaft formation at several centimeters from the scalp.
The post-apoptotic human hair contains the thousands of mRNAs,
miRNAs, and other small RNAs and accurately reflects tissue of
origin. Finally, we find that the RNA of the post-apoptotic hair
shares similar patterns of intragenic coverage as RNA from viable
keratinocytes. Hence, the removal and destruction of RNA post-
cornification appears to be highly inefficient in the keratinized
structure of the hair and potentially provides an ideal tissue for
studies of genetic and acquired diseases.
Detection of lineage-specific miRNA and mRNA in mouse
In mice, the transcriptional profiles of living hair follicle-specific
miRNAs and mRNAs have been previously examined [23,24].
Based on these expression profiles, we examined whether miRNAs
and mRNAs specifically expressed in the hair follicle could also be
detected in the external hair shaft and tested whether non-hair shaft
transcripts were excluded. As comparison, we utilized whole mouse
skin, which contains all lineages of the hair follicle, epidermis,
dermis and adipose tissue (Fig. 1A). Using quantitative real-time
PCR, miRNAs were readily detected in the hair shaft at 3–4 cycles
higher than from equivalent amounts of whole skin RNA (Fig. 1B).
Relative to control RNAs, e.g. snoRNA-251, we found evidence for
lineage specificity using miRNA expression (Fig. 1C). Significantly,
thelevelofa miRNA (mir-203), whichisconsideredtobeepidermal-
specific , was ten fold less abundant in the mouse hair shaft
compared to whole skin. These findings indicate that miRNAs are
sufficiently intact in the hair shaft for specific detection by
quantitative real-time PCR (qRT-PCR).
As hair keratins are abundantly expressed during hair develop-
ment, we next investigated whether their transcripts can be detected
in the external hair shaft. In addition, hair keratin genes are also
expressed by specific cell types and identify the cell type of origin for
RNAs extracted from external hair (Fig. 1D). Several lineage-
specific keratin  and central medulla genes  could be
detected in the post-cornification hair shaft (Fig. 1D, F). Relative to
whole skin, Keratin 32 (Krt32), Krt34, and Krt35 were detected at 8.6,
16.1 and 1.5-fold higher in the external hair. Sh3d19, a medulla-
specific gene, was also relatively more abundant in the hair shaft.
The enrichment of these RNAs in hair shaft relative to whole skin
occurs as hair follicle transcripts are diluted by other tissues present
in whole skin. As a negative control, genes expressed by tissues that
do not contribute cells to the hair shaft were examined. One such
tissue, called the dermal papilla, is a closely associated mesenchymal
population that supports hair growth but does not itself contribute
cells to the growing hair shaft. We found that two dermal papilla-
specific genes, Sox2  and Bmp6 , were significantly reduced
in expression in the hair shaft, only detectable at 6.8 and 3.7 fold
levels below whole skin, respectively (Fig. 1E, F). In sum, these
studies indicate that miRNAs and mRNAs of at least three cell types
persist in the external hair of mice contains both miRNAs and
mRNAs and that cornification does not sufficiently destroy these
RNAs to prevent their detection.
Generation of A Human Hair RNA-Seq Library
We considered the possibility that residual RNAs in the mouse
hair might reflect its rapid growth and potential persistence of cells
rather than the low ribonuclease activity during and after
cornification. In humans, hair does not contain viable cells 
and undergoes 1–2 months of maturation prior to its emergence
from the scalp, compared to 1 week in mice . While human
hair may be a better model to study post-terminal differentiation
hair, studies of human hair versus mouse hair posed unique
challenges. First, expression data of miRNAs in the living human
hair follicle was limited, and second, gene expression biases may
arise from individual variation in gene expression or detection
biases due to nucleotide polymorphisms. We addressed these
limitations by utilizing parallel sequencing of small RNAs 
isolated from human hair (Fig. 2). To generate a comprehensive
library of RNAs in human hair and to account for some aspects of
human variation, we pooled RNA from the hair shafts of five
individuals, who varied in gender, hair shape, and origin
(demographics detailed in Supplemental Methods). Total RNA
isolated from external human hair shafts by this method was
ligated to adaptors designed for small RNA reads, reverse
transcribed, amplified, size selected and analyzed by Illumina/
Solexa-based small RNA sequencing.
From this approach, 13.5 million high-quality reads and 1.2
million unique sequences were obtained. Subsequently, the reads
were aligned to the human genome (hg18) and to dedicated small
RNA libraries (miRBase and snoBase). A large portion of reads
aligned to more than 10 loci or no loci, due to low complexity of
sequence, highly repetitive targets or the presence of non-human
RNA (Fig. 2A). We also found a high frequency of aligned reads
measuring 22 nucleotides, the size of mature miRNAs (Fig. 2B).
7,193 mRNAs were detected with read coverage of 10 or more,
and 251 mRNAs were found at 200 or more reads. For alignments
that had 200 or more reads, we also identified 449 distinct
miRNAs and 339 snoRNAs in the hair shaft library. Percentages
of reads of different classes of RNA identified from human genome
annotation are shown in Figure 2C.
The vast majority of miRNAs detected in the hair shaft
belonged to the LET-7 family, which plays a role in regulation of
proliferation, differentiation, and stem cell maintenance 
(Fig. 2D, Table S1). The remaining 123 miRNAs account for
21.4% of all miRNAs in the hair shaft. Like the mouse hair, the
human hair shaft contained numerous mRNAs encoding genes of
the keratin associated proteins (KRTAP) or keratins (Fig. 2E,
Table S1). Members of the KRTAP family comprise 58.3% of all
mRNAs in the hair shaft. Like keratins, KRTAP genes function as
support proteins, which become highly crosslinked and participate
in forming the shape of the hair shaft . Many characteristics of
the small RNA library generated from hair allowed us to verify
that the sequences originated from RNA rather than contaminat-
ing genomic DNA. First, sequence data showed significantly
greater coverage of transcriptional units compared to non-
transcribed genomic regions (Fig. 2F). Second, the directional
nature of the small RNA library preparation revealed specific
alignment of sequence reads to the known transcriptional
orientation of individual genes.
Many of the mRNAs found in the hair shaft were not specific to
the hair follicle. Several gene annotation searches, including a Set-
Distiller batch tool , identified strong statistical associations to
phenotypes and pathways affecting other organ systems (Fig. 3A).
377 pathways involved in environmental and pharmacologic
signals and 391 pathways related to specific human disorders were
associated with genes detectable in the hair shaft. Given the large
7,193-gene set used to represent the hair shaft, we attempted to
address the specificity of non-hair dataset associations by
comparing these results to the comparison of an 11,027-gene set
representing normal human epidermal keratinocytes (NHEK).
Fate of Post-Apoptotic RNAs in Human Hair
PLoS ONE | www.plosone.org2November 2011 | Volume 6 | Issue 11 | e27603
The NHEK dataset was chosen, because it still represents
epithelial tissue but is also significantly different from the hair
shaft in origin and grown under different conditions. Eighteen
signatures (p=3.761026to 1.0610211) belonging to exposure,
disease or medication-induced gene expression showed detection
in hair shaft RNA but not NHEK RNA (Fig. 3B–D). These
findings suggest that many associations between hair shaft RNAs
and non-hair related pathways exist irrespective of the large input
gene set. Moreover, these annotation studies suggest that hair shaft
tissue may be of significant medical value in the screening of
transcribed biomarkers shared with epidermal keratinocytes or
uniquely as a source for conditions such as autism, hypertension,
and thyroid diseases. Lastly, many of the transcripts present in the
hair shaft are themselves genetic targets in human disease (Table
S2) or contribute to pathways in involved in response to chemical
compounds and drugs (Table S3). The gene ontology character-
ization of hair shaft RNA profiles provides guidance for the use of
the hair shaft to study organismal responsiveness to medications or
environmental exposures to chemicals or infections.
Relative Similarities in RNA content of Viable
Keratinocytes and External Hair
The surprising stability of RNA in the hair shaft led us to
examine and identify how RNA read coverages differ between
viable and non-viable tissue. We thus compared RNA libraries of
the hair shaft to viable NHEK using the same sequencing
platform. First, we found that lineage-restricted hair keratin genes
were highly enriched in the hair shaft library relative to NHEK
Figure 1. Detection of tissue-specific miRNA and mRNA from mouse hair shafts. (A) Schematic diagram detailing source of tissue used in
mouse studies. Hair shaft (HS) refers to the external portion of hair used for RNA isolation. Whole skin (WS) encompasses the epidermis, dermis, and
the entire hair follicle, which includes cells fated to become the hair shaft (blue), supporting non-hair shaft cell populations (inner root sheath, orange
and dermal papilla, purple). Red circles represent the earliest and latest regions of detected apoptosis. (B) Differential expression of snoRNAs and
miRNAs in whole skin versus external hair shaft from 2–3 week old mice. Detection levels are displayed in cycle numbers normalized for equal input
total RNA. Higher cycle numbers indicate lower levels of detection. (C) Mouse mir-212 and mir-203 levels normalized to snoRNA-251 levels reveal
increased mir-212 and greater than two-fold reduction of mir-203 in mouse hair shaft relative to whole skin. (D) Detection and quantification of
cuticle, cortex and medulla-specific transcripts in hair shaft vs. whole skin reveal patterns of enrichment. (E) Assay for dermal papilla-specific
transcripts in hair shaft vs. whole skin total RNA. (F) Gel electrophoresis of amplification products of hair shaft and non-hair shaft (non-HS) genes in
whole skin and hair shaft.
Fate of Post-Apoptotic RNAs in Human Hair
PLoS ONE | www.plosone.org3 November 2011 | Volume 6 | Issue 11 | e27603
[34,35] (Fig. 4A). Many of the neighboring non-hair shaft keratin
genes were absent from the hair shaft library. Comparing the
expression of six different cell types in the hair follicle, we
concluded that RNA of the human hair shaft library predomi-
nantly originate from transcripts of the three hair committed cell
types (Fig. 4B, C). We also considered whether RNA from a non-
Figure 2. Generation of a human hair shaft small RNA library. Small RNA library constructed from RNA pooled from hair shafts of five
individuals. (A) Uniqueness of RNA sequences. A large portion of reads can be mapped to 1 or more loci in the human genome (hg18). (B)
Distribution of trimmed read lengths demonstrates two peaks at 22 and 29 nucleotides. (C) Types of RNA retained in the human hair shaft.
Percentages of genomic aligned RNAs map to known mRNAs, miRNAs, rRNAs, unannotated RNAs, and others. (D) Complexity of miRNA in chart
reveals high proportion of LET7 family members. Only miRNAs with 200 or more hits are demonstrated in chart. (E) Complexity of mRNAs in human
hair shaft demonstrated in pie chart. (F) Read coverage of hair shaft RNA library of a 60 kb genomic window, representing of a portion of the KRTAP5
gene cluster. Alignment of reads demonstrates enrichment of sequences to transcriptional units and strand preservation of known transcript
orientation of KRTAP5-7, 5-8, 5-9, 5-10, and 5-11 genes. These findings confirm absence of contaminating genomic DNA sequences.
Fate of Post-Apoptotic RNAs in Human Hair
PLoS ONE | www.plosone.org4 November 2011 | Volume 6 | Issue 11 | e27603
viable source might demonstrate different patterns of intragenic
region coverage compared to viable keratinocytes (Fig. 4D). Here,
we found a similar pattern of coverage of intragenic regions, e.g. 59
untranslated regions (UTR), exonic, intronic, and 39 UTR
(Fig. 4E). In both RNA libraries, intronic transcripts, which
include 5S rRNA, were the most abundant. Exonic sequences of
both libraries revealed similar patterns of coverage, with the
exception that the 59 regions of viable keratinocytes (32.5%) were
relatively more enriched than in the hair shaft RNA library
(24.2%). Thus, the abundance of detected transcripts and the high
degree of similarity in the persistent RNA of the non-viable hair
shaft and viable keratinocytes indicate that RNAs are inefficiently
targeted by apoptosis during hair.
miRNAs and mRNAs Are Stable In Distal Hair Segments
During hair growth, cells are continuously incorporated into the
hair shaft such that as the hair elongates, the progressively older
segments of hair are pushed further distally from the scalp. Given
the remarkable stability of RNA following cornification in the hair
shaft inferred from the large number of detected transcripts and
the many similarities to viable keratinocytes, we investigated the
stability of RNA at varying distances, using the the scalp as the
point of origin. Strands of hair from two individuals were
collected, aligned and divided into 2.5 cm segments. We found
that RNA could be consistently isolated even at distant regions of
the hair shaft (Fig. 5A). At increasing distances, higher cycle
numbers were required to detect miRNA and mRNA target genes
by real-time PCR, indicating decreased abundance of specific
RNA transcripts (Fig. 5B). Nevertheless, even at these higher
amplification cycles, specific products were obtained and could be
confirmed by the melting curve (Fig. 5C) and gel electrophoresis
(not shown). Decay of mRNA and miRNA detection was
estimated by examining the increasing cycle numbers over a
relative measure of time, in this case, hair length (Fig. 5D). We find
that a two-fold reduction of mRNAs and miRNAs occurs over the
length of 0.9260.11 cm and 0.8160.16 cm, respectively. These
studies suggest that the conditions that allow RNAs to be stable
during hair formation persist even months after cornification and
provide insights into the natural decay of RNAs in the hair.
In this study, we demonstrate that as a direct remnant of
previously living cells, hair retains a vast amount of transcriptional
data reflective of incomplete degradation of RNA following
apoptosis. In addition to hundreds of miRNAs and snoRNAs,
sequencing of hair RNA identified 7,193 unique mRNAs, or an
equivalent to over a quarter of all genes in the human genome.
Hair RNAs potentially reflect months of expression data
temporally deposited along the length of the hair shaft and may
be useful for investigations of gene expression and biomarker
Figure 3. Gene ontology analysis of hair shaft RNA library. (A) Phenotypic association of genes expressed in the hair shaft library. Shown are
numbers of genes present in the hair shaft that are associated with organ phenotypes identified by mouse mutations. (B) Environmental exposures
pathways associated with genes identified in hair shaft library. Numbers of genes associated with selected exposure pathways are shown. (C)
Disorder pathways associated with genes present in hair shaft library. (D) Compound pathways associated with hair shaft RNA library. Gray shading of
column bars in all panels indicates lower limits of p-values calculated based on input hair shaft library genes vs. random background genes. Red solid
circles denote unique pathway gene sets identified in hair shaft but not keratinocytes.
Fate of Post-Apoptotic RNAs in Human Hair
PLoS ONE | www.plosone.org5 November 2011 | Volume 6 | Issue 11 | e27603
Figure 4. Comparison of hair shaft vs. cultured human keratinocyte RNA sequenced libraries. (A) Cell-type specific detection of human
hair keratins (KRT) in hair shaft versus normal human epidermal keratinocytes (NHEK). Pattern of KRT expression in six different hair shaft (HS) and
non-hair shaft inner root sheath (IRS) are shown in boxes, where filled boxes indicate $50 reads detected. (B) Percent of cell-type specific keratin
genes detected reveals degree of enrichment of each hair shaft and non-hair shaft compartment in hair shaft vs. NHEK RNA libraries. (C) Illustration
demonstrating the spatial distribution of cell types represented by lineage-specific KRT, including three hair shaft cell types (blue) and three IRS cell
types (orange). (D) Genome-wide comparison of intragenic region coverage by hair shaft vs. NHEK RNA libraries reveals similar representation of most
intragenic regions. Exonic coverage was greater in viable NHEK compared to hair, while intronic coverage was more extensive in hair. (E) Distribution
Fate of Post-Apoptotic RNAs in Human Hair
PLoS ONE | www.plosone.org6 November 2011 | Volume 6 | Issue 11 | e27603
As a model for studying differentiation and cell death, the hair
provides insights into the extent to which RNAs are targeted for
degradation during cornification. The numerous RNAs identified
in the hair shaft suggest that RNA degradation is highly inefficient
during cornification. In support of this hypothesis, read coverage
between hair shaft versus viable keratinocyte RNA libraries
exhibited remarkable similarities in coverage of intragenic regions
with small biases toward reduced exonic coverage in hair shaft.
Several RNases were detected in the hair shaft RNA library,
including RNASE4, 7, 12, 13, RNASEN, and RNASET2, indicating
that lack of RNases do not account for the persistence of RNAs in
hair. RNA stability in hair might involve other mechanisms such
Figure 5. Stability of mRNA and miRNA in hair shaft at varying distances from scalp. (A) Total RNA yields from hair shaft at varying
distances from the scalp as detected by A260 absorbance. (B) Detection of specific mRNAs and miRNAs identified from hair shaft library at varying
distances from scalp, indicated in cycle numbers. Higher cycle numbers reflect reduced levels of detection, particularly at 10 cm. All experiments
shown demonstrate at least three replicates and at least two individuals. (C) Melting peaks of KRTAP5-3, MIR-24 and SNO251 as an example of
amplified mRNA, miRNA, and snoRNA products from 2.5 to 10-cm segments, respectively. (D) Calculated average half-lifes of mRNAs vs. miRNAs
based on cycle numbers using non-linear fit one-phase decay modeling. Because growth rates of hair were not directly assessed in these studies, the
distance of hair sampled from scalp was used as a marker of relative time.
of sequence coverage within regions of each domain, e.g. 59 intergenic, 59-UTR, introns, exons, 39-UTR and 39 intergenic regions, are shown as 5%
windows from 59 to 39.
Fate of Post-Apoptotic RNAs in Human Hair
PLoS ONE | www.plosone.org7 November 2011 | Volume 6 | Issue 11 | e27603
as sequestration of RNases, presence of endogenous RNase
inhibitors, protection of RNA by ribonucleoprotein particles or
low water content in hair . Understanding the contribution of
these potential mechanisms may be important in improving RNA
detection and stability in future hair RNA expression studies. The
removal of DNA versus RNA during apoptosis may play distinct
roles in different tissues . In erythrocytes, corneocytes, and lens
fiber cells, where nuclei and DNA are removed during terminal
differentiation, persistence of RNA may be important in
maintaining cell function. Whether incompletely removed RNA
in hair shaft serves a functional purpose is unknown. It is unclear if
mRNAs identified in the hair shaft are available for translation as
the degree of fragmentation and survival of translational
machinery were not examined in this study. In addition, studies
aimed at determining whether RNA persists in other cornified
tissues such as the nail could be of great importance for studying of
diseases that affect localized body sites, such as cancer, infection,
injury and exposures.
The discovery of stable mRNAs and miRNAs distant in the hair
shaft from the scalp overcomes a formidable obstacle in the
application of RNA diagnostics. Currently, RNA studies based on
tissue biopsy and phlebotomy are vulnerable to RNA degradation
. In the current study, we found that older, more distal regions
of hair still contain detectable mRNAs and miRNAs. These
findings have several implications. First, the stability of RNA in
older portions of hair suggests that RNA may be stable in hair over
many months. This attribute differs greatly from other tissues and
blood and potentially resolves a problem in storage and
transportation of RNA, necessary for RNA-based diagnostic or
biomarker studies. Second, temporal changes in gene expression
either physiologic or induced by chemical, drug or disease might
be stored co-linearly along the length of the hair shaft. Use of this
spatiotemporal pattern of RNA deposition might provide a novel
approach to studying the natural course or inciting events of
disease. In addition to these characteristics, the continuous
replacement of hair and its ease of access are advantages to
developing diagnostic approaches based on hair RNA.
While possibly ideal for many types of biomedical studies, the
use of hair RNA for molecular studies and diagnosis currently has
several limitations. At this time, the amount of variation in RNA
expression between different individuals, ages, and genetic
backgrounds are not known. In addition, differential growth rates
of hair in individuals due to differences in genetic background,
age, and other factors obscure accurate measurements of time.
Recent studies indicate that growth of human hair may vary from
1.3 to 2.2 months per centimeter in different individuals . More
accurate measurements might be made possible with the
identification of cyclically expressed genes, which could used to
normalize differential growth rates . An additional limitation is
that since new cells are added to the hair shaft only during active
hair growth, it is not yet known what the effects different stages of
the hair cycle might have on the pattern and stability of RNA in
the hair. In this case, identification of transcripts representative of
the final stages of the hair cycle might be required to determine
whether retained transcripts reflect specific portions of the hair
cycle. Lastly, because the characteristics that contribute to RNA
stability in the hair are unknown, it is possible that RNA stability
varies in individuals. These unknown aspects may bias the results
of RNA detection in the hair.
The use of parallel sequencing of small RNAs provides an atlas of
residual transcripts in humans of different genders, populations, and
hair type. Sequencing technology provides several informative
features valuable to molecular diagnostic studies including tran-
scriptional orientation, multiple means of validation of expression
and quantity of transcripts, and sequence data [39,40]. Sequence
data also provide a significant source for the identification of genetic
polymorphisms, detection of allele-specific expression differences,
and somatic mutations. The extent to which variant alleles are
present in hair shaft RNA was not explored in this study. Current
obstacles to use of this data for allele-specific studies include the
possibility of chemical changes in nucleotides, e.g. cytosine
deamination , and incorporation errors inherent to RNA
polymerases . Further studies are needed to assess the impact of
these factors on the fidelity of hair RNA sequences.
In the past, hair analysis has aided in the diagnosis of inherited
and acquired diseases [14,43,44] and as a phenotypic marker of
medication and chemotherapeutic response . Thus while the
discovery of post-apoptotic RNA stability in hair provides insights
into programmed cell death, these findings also greatly expand the
possible applications of hair in medical discoveries in the clinic and
the lab. The unique properties of the linear growth of the hair
serve as a record of gene activity during organ growth and of the
individual’s own history. This record may be of great value for
studies in organ development, evolution, genetic variation, and
Materials and Methods
Animal and Human Specimens
Mouse lines originated from a CF-1 mixed genetic background
(Charles River). Human samples were unidentified and pooled.
The demographic profiles are shown in Supplemental Methods.
Normal human epidermal keratinocytes were grown in EpiLife
culture media (Cascade Biologics, Portland, OR) containing
0.06 mM calcium and defined growth supplement. All experi-
ments and informed consents were performed and approved
according to the institutional guidelines established by the
University of California, San Diego, Institutional Animal Care
and Use Committee and the University of California, San Diego,
Human Research Protection programs, Protocol ID #091646.
Written informed consents were obtained from each participant of
Hair collection, storage, and RNA extraction
For mouse studies, hair from 3-week old animals was trimmed
and stored in RNAlater (QIAGEN) solution at 280uC. Hair was
visually inspected to verify absence of hair bulbs, which contain
viable cells. For human studies, hair was trimmed at a distance of
5 mm from the scalp. Hair was washed in 70% ethanol and water.
Two methods were used for RNA extraction. During initial studies
in mouse hair and human hair shaft RNA-sequencing, Trizol
reagent (Invitrogen) was utilized in combination with mechanical
disruption with 1.0 mm zirconia beads (BioSpec Products, Inc.).
Subsequently, improved extraction was obtained with the addition
of 0.1 M dithiothreitol (DTT) to Trizol reagent or a urea-based
RNA extraction buffer . The improved yield of RNA
presumably results from the effects of a reducing agent on the
highly disulphide crosslinked hair tissue. A more detailed protocol
is described in Supplemental methods.
Real-time PCR analysis for mRNAs, snoRNAs, and
Primer sequences and miRNA probes are detailed in Supple-
mental Methods. Primers for mRNAs were designed using
Primer3 . Reverse transcription and real time PCR were
performed with Maxima First Strand cDNA Synthesis Kit
(Fermentas) and Maxima SYBR Green (Fermentas), respectively.
For microRNA assays, cDNA was prepared using a RT2miRNA
Fate of Post-Apoptotic RNAs in Human Hair
PLoS ONE | www.plosone.org8 November 2011 | Volume 6 | Issue 11 | e27603
First Strand Kit (QIAGEN) and amplified with stem-loop primers
from SABiosciences/QIAGEN (Table S4) using RT2SYBRH
Green qPCR mastermix (QIAGEN). Real-time PCRs were
performed on an ABI 7300 Real Time PCR system and a Roche
Lightcycler 480. A comparative CT method was used for
quantitative analysis using snoRNA-251 in mouse microRNAs
samples, Gapdh for mouse mRNAs. Half-life of mRNAs and
miRNAs determined using 2‘deltaCT values normalized to levels
at site of origin and Prism (Prism 5, Version 5.0d) non-linear fit
first-order decay modeling.
RNA Sequencing and Analysis
Fifty hair shafts, equivalent to 30 mg, were utilized to prepare
total RNA using the above methods. One microgram of total RNA
isolated from hair was used to generate a small RNA library.
Adapter ligation, reverse transcription, amplification, gel extrac-
tion and sequencing were performed, according to Illumina
protocols for miRNA-Seq (v1.5.0) (Illumina). Sequencing was
carried out on an Illumina Genome Analyzer IIx. Genome
Analyzer Pipeline was used to generate a FASTQ file from raw
RNA-seq data, and quality scores were offset by 64 following
Solexa-1.3+ standards. FASTX-Toolkit package, hg18 genome
alignment with BOWTIE, was used for pre-processing and
alignment. Parameters for alignment are detailed in Supplemental
methods, and a list of human hair shaft transcripts are provided in
Table S5. Computational processes were carried on Triton
Resource at the San Diego Supercomputer Center at UCSD,
using a single node of 32 processors and 512GB of RAM. For
NHEK to hair shaft RNA library comparisons, representative
coverages of intragenic regions in REFSEQ were compared. Each
gene in REFSEQ was segmented into annotated 59-UTR, introns,
exons, and 39-UTR regions; these regions were further broken up
into windows of 5% of the length of the region. The hair shaft
library was aligned to these regions, and the total coverage of all
5% windows for each region type in each gene was plotted,
generating a summary of overall coverage patterns in intragenic
regions. Functional and compound associations were performed
using Set-Distiller batch tool  and Mouse Genome Informatics
. Further methods are described in Methods S1 and
accompanying References S1.
snoRNAs (.200 reads).
Annotation and reads counts of mRNA, miRNA,
transcripts in hair.
Representation of Genetic Association Database
ways in hair.
Representation of Novoseek compound-genetic path-
Primers and reagents used in PCR analysis.
Annotated reads with .10 reads.
We are grateful to Steven Head and the Scripps DNA Core for library
preparation and sequencing, Terry Gaasterland for assisting with use of
San Diego Supercomputer Center.
Conceived and designed the experiments: BDY GKL CCZ AM.
Performed the experiments: GKL CCZ AM. Contributed reagents/
materials/analysis tools: BDY GKL CCZ AM. Wrote the paper: BDY
1. Opferman JT, Korsmeyer SJ (2003) Apoptosis in the development and
maintenance of the immune system. Nat Immunol 4: 410–415.
2. Metzstein MM, Stanfield GM, Horvitz HR (1998) Genetics of programmed cell
death in C. elegans: past, present and future. Trends Genet 14: 410–416.
3. He B, Lu N, Zhou Z (2009) Cellular and nuclear degradation during apoptosis.
Curr Opin Cell Biol 21: 900–912.
4. Magerl M, Tobin DJ, Muller-Rover S, Hagen E, Lindner G, et al. (2001)
Patterns of proliferation and apoptosis during murine hair follicle morphogen-
esis. J Invest Dermatol 116: 947–955.
5. Botchkareva NV, Ahluwalia G, Shander D (2006) Apoptosis in the hair follicle.
J Invest Dermatol 126: 258–264.
6. Morioka K (2005) Hair follicle: differentiation under electron microscope: an
atlas. New York: Springer. 165 p.
7. Soma T, Ogo M, Suzuki J, Takahashi T, Hibino T (1998) Analysis of apoptotic cell
death in human hair follicles in vivo and in vitro. J Invest Dermatol 111: 948–954.
8. Koike M, Mashino M, Sugasawa J, Koike A (2007) Dynamic change of histone
H2AX phosphorylation independent of ATM and DNA-PK in mouse skin in
situ. Biochem Biophys Res Commun 363: 1009–1012.
9. Kawane K, Fukuyama H, Yoshida H, Nagase H, Ohsawa Y, et al. (2003)
Impaired thymic development in mouse embryos deficient in apoptotic DNA
degradation. Nat Immunol 4: 138–144.
10. Fischer H, Szabo S, Scherz J, Jaeger K, Rossiter H, et al. (2011) Essential role of
the keratinocyte-specific endonuclease DNase1L2 in the removal of nuclear
DNA from hair and nails. J Invest Dermatol 131: 1208–15.
11. Finch C (2009) Degeneration of nuclei and mitochondria in human hairs.
J Forensic Sci 54: 346–349.
12. Muller-Rover S, Handjiski B, van der Veen C, Eichmuller S, Foitzik K, et al.
(2001) A comprehensive guide for the accurate classification of murine hair
follicles in distinct hair cycle stages. J Invest Dermatol 117: 3–15.
14. Shimomura Y, Christiano AM (2010) Biology and genetics of hair. Annu Rev
Genomics Hum Genet 11: 109–132.
15. Kariko K, Buckstein M, Ni H, Weissman D (2005) Suppression of RNA
recognition by Toll-like receptors: the impact of nucleoside modification and the
evolutionary origin of RNA. Immunity 23: 165–175.
16. Houseley J, Tollervey D (2009) The many pathways of RNA degradation. Cell
17. Lykke-Andersen J (2001) mRNA quality control: Marking the message for life or
death. Curr Biol 11: R88–91.
18. Castelli JC, Hassel BA, Maran A, Paranjape J, Hewitt JA, et al. (1998) The role
of 29-59 oligoadenylate-activated ribonuclease L in apoptosis. Cell Death Differ
19. Samali A, Gilje B, Døskeland SO, Cotter TG, Houge G (1997) The ability to
cleave 28S ribosomal RNA during apoptosis is a cell-type dependent trait
unrelated to DNA fragmentation. Cell Death Differ 4: 289–293.
20. Houge G, Robaye B, Eikhom TS, Golstein J, Mellgren G, et al. (1995) Fine
mapping of 28S rRNA sites specifically cleaved in cells undergoing apoptosis.
Mol Cell Biol 15: 2051–2062.
21. Nakagawa A, Shi Y, Kage-Nakadai E, Mitani S, Xue D (2010) Caspase-
dependent conversion of Dicer ribonuclease into a death-promoting deoxyribo-
nuclease. Science 328: 327–334.
22. Lau NC (2004) Molecular characterization of animal microRNAs: sequence,
expression, and stability Ph.D. thesis, Massachusetts Institute of Technology.
23. Rendl M, Lewis L, Fuchs E (2005) Molecular dissection of mesenchymal-
epithelial interactions in the hair follicle. PLoS Biol 3: 1910–1924.
24. Yi R, Pasolli HA, Landthaler M, Hafner M, Ojo T, et al. (2009) DGCR8-
dependent microRNA biogenesis is essential for skin development. Proc Natl
Acad Sci U S A 106: 498–502.
25. Yi R, Poy MN, Stoffel M, Fuchs E (2008) A skin microRNA promotes
differentiation by repressing ‘stemness’. Nature 452: 225–229.
26. Shimomura Y, Aoki N, Ito K, Ito M (2003) Gene expression of Sh3d19, a novel
adaptor protein with five Src homology 3 domains, in anagen mouse hair
follicles. J Dermatol Sci 31: 43–51.
Fate of Post-Apoptotic RNAs in Human Hair
PLoS ONE | www.plosone.org9November 2011 | Volume 6 | Issue 11 | e27603
27. Driskell RR, Giangreco A, Jensen KB, Mulder KW, Watt FM (2009) Sox2- Download full-text
positive dermal papilla cells specify hair follicle type in mammalian epidermis.
Development 136: 2815–2823.
28. Rendl M, Polak L, Fuchs E (2008) BMP signaling in dermal papilla cells is
required for their hair follicle-inductive properties. Genes Dev 22: 543–557.
29. Lebeau MA, Montgomery MA, Brewer JD (2011) The role of variations in
growth rate and sample collection on interpreting results of segmental analyses of
hair. Forensic Sci Int 210: 110–116.
30. Hafner M, Landgraf P, Ludwig J, Rice A, Ojo T, et al. (2008) Identification of
microRNAs and other small regulatory RNAs using cDNA library sequencing.
Methods 44: 3–12.
31. Roush S, Slack FJ (2008) The let-7 family of microRNAs. Trends Cell Biol 18:
32. Rogers MA, Schweizer J (2005) Human KAP genes, only the half of it?
Extensive size polymorphisms in hair keratin-associated protein genes. J Invest
Dermatol 124: vii–ix.
33. Stelzer G, Inger A, Olender T, Iny-Stein T, Dalah I, et al. (2009) GeneDecks:
paralog hunting and gene-set distillation with GeneCards annotation. OMICS
34. Langbein L, Rogers MA, Praetzel-Wunder S, Helmke B, Schirmacher P, et al.
(2006) K25 (K25irs1), K26 (K25irs2), K27 (K25irs3), and K28 (K25irs4) represent
the type I inner root sheath keratins of the human hair follicle. J Invest Dermatol
35. Schweizer J, Bowden PE, Coulombe PA, Langbein L, Lane EB, et al. (2006)
New consensus nomenclature for mammalian keratins. J Cell Biol 174: 169–174.
36. Leveque JL, Garson JC, Pissis P, Boudouris G (1981) Free water in hair keratin?
A depolarization thermal-current study. Biopolymers 20: 2649–2656.
37. Lo YM, Chiu RW (2004) The biology and diagnostic applications of plasma
RNA. Ann N Y Acad Sci 1022: 135–139.
38. Lin KK, Kumar V, Geyfman M, Chudova D, Ihler AT, et al. (2009) Circadian
clock genes contribute to the regulation of hair follicle cycling. PLoS Genet 5:
39. Trapnell C, Williams BA, Pertea G, Mortazavi A, Kwan G, et al. Transcript
assembly and quantification by RNA-Seq reveals unannotated transcripts and
isoform switching during cell differentiation. Nat Biotechnol 28: 511–515.
40. Pickrell JK, Marioni JC, Pai AA, Degner JF, Engelhardt BE, et al.
Understanding mechanisms underlying human gene expression variation with
RNA sequencing. Nature 464: 768–772.
41. Hofreiter M, Jaenicke V, Serre D, Haeseler Av A, Pa ¨a ¨bo S (2001) DNA
sequences from multiple amplifications reveal artifacts induced by cytosine
deamination in ancient DNA. Nucleic Acids Res 29: 4793–4799.
42. Li M, Wang IX, Li Y, Bruzel A, Richards AL, et al. (2011) Widespread RNA
and DNA Sequence Differences in the Human Transcriptome. Science 333:
43. Cheng AS, Bayliss SJ (2008) The genetics of hair shaft disorders. J Am Acad
Dermatol 59: 1–22.
44. Budowle B, Allard MW, Wilson MR, Chakraborty R (2003) Forensics and
mitochondrial DNA: applications, debates, and foundations. Ann Rev Genomics
Hum Genet 4: 119–141.
45. Whiting DA, Dy LC (2006) Office diagnosis of hair shaft defects. Semin Cutan
Med Surg 25: 24–34.
46. Gough NM (1988) Rapid and quantitative preparation of cytoplasmic RNA
from small numbers of cells. Anal Biochem 173: 93–95.
47. Rozen S, Skaletsky H (2000) Primer3 on the WWW for general users and for
biologist programmers. Methods Mol Biol 132: 365–386.
48. Blake JA, Bult CJ, Kadin JA, Richardson JE, Eppig JT (2011) The Mouse
Genome Database (MGD): premier model organism resource for mammalian
genomics and genetics. Nucleic Acids Res 39: D842–848.
Fate of Post-Apoptotic RNAs in Human Hair
PLoS ONE | www.plosone.org10 November 2011 | Volume 6 | Issue 11 | e27603