Supporting Data. Analyses by short and long telomere length (TL) groups. As another way of examining the relationship between stress and telomere length categorically, telomere length was categorized as long (n = 25) versus short (n = 32), with a mean split of 1.22 for the relative telomere length (T/S) ratio. Analyses of covariance (ANCOVAs) were performed, controlling for age, on each outcome.

Women in the short TL group were caregivers for greater number of years [mean (M) = 6.70 years; SE = 0.64) compared with the long TL group (M = 4.77, SE = 0.52; P < 0.02). When perceived stress was examined as a continuous measure across the entire sample (n = 58), stress was marginally significantly higher in the short TL group (M = 18.2, SE = 1.4) compared with the long telomere length group (M = 15.5, SE = 1.2; P = 0.08). When extreme high- and low-stress groups (upper and lower quartiles, n = 28) were examined, 79% of low-stress subjects fell into the long TL group (versus 21% in the low TL category), whereas 64% of high-stress subjects fell into the short TL group (versus 36% in the high TL group). This finding was significant according to a c ² test [X²(1) = 5.2, P < 0.03].

Furthermore, we tested for differences in measures of telomerase and oxidative stress in the high versus low TL groups, controlling for age and by using ANCOVAs. Telomerase was significantly lower in the short TL group (M = 0.059, SE = 0.013) than the long TL group [M = 0.101, SE = 0.011; F(1,55) = 5.9; P = 0.01]. Similarly, the oxidative stress index was significantly higher in the short TL group (M = 0.085, SE = 0.010) versus the long TL group [M = 0.060, SE = 0.010; F(1,40) = 3.49; P = 0.035].

Testing for the potential confound of illness. To rule out the possible confound of current infectious illness on responsive markers of aging, such as telomerase activity, participants were asked about whether they experienced any illness symptoms the day before they were studied. On the day of study, their temperature was taken to make sure core body temperature was in the normal range. In addition, participants were monitored for 3 days during a 1-week period after the laboratory session, and any daily symptoms were noted by reporting them on a daily diary log. If a participant currently had any symptoms of a flu or cold, their sessions were postponed for a month, to their next menstrual cycle, early follicular stage. One participant reported "not feeling well" during her session. However, she reported no symptoms on the following 3 days. Eight participants, including six caregivers (13% of caregivers) and two controls (7% of controls) reported at least one infectious symptom (e.g., cough, congestion, or fever) at some point during the 3 days of monitoring after the session. These symptoms could indicate that they had an unknown infection at the time of the session. Having a subsequent symptom was not related to telomerase levels during the session (M = 0.08 ± 0.01 in the well group; M = 0.10 ± 0.03 in the possibly ill group; P = 0.58). Although there were no meaningful differences between groups, telomerase tended to be higher, not lower, in the possibly ill group, which is consistent with the known effects of antigens on increasing telomerase expression (1, 2). Thus, it is not likely that the lower telomerase levels observed in the high-stress group is due to greater current infection, which should theoretically lead to higher telomerase levels.

Testing immune cell subpopulations in high- versus low-stress groups. Loss of naïve T cells is partly responsible for the shorter TL in peripheral blood mononuclear cells (PBMCs) in aged monkeys (3). Similarly, the chronically stressed group may show an aged profile in terms of having a greater number of infections over time, which would have a cumulative effect on one’s T cell proliferation, which would result in having fewer naïve cells and greater numbers of memory cells, which have shorter TL (4).

To test for any such differences in cell subpopulation numbers, we measured cell subpopulations from the same blood sample used for all other measurements. The lymphocytes and monocytes from this same sample pool were measured with flow cytometry, and were not different in terms of their absolute numbers in the high- versus low-stress groups (see Table 2). Both high- and low-stress groups had similar absolute numbers of cell subpopulations, including similar numbers of naïve and memory T cells. This finding suggests that the possible cascade of "more infections, more proliferation, and, thus, more memory T cells" cannot account for the observed results, but it does suggest that the high-stress group’s telomeres were shortened more directly.

Detailed Methods. Sample description and recruitment. Participants were recruited by their child’s health care professional in several clinics in San Francisco Bay-area hospitals or by public postings. The study was described over the telephone, and the subjects were screened for eligibility. If eligible and interested, the participants came to the Oakland Children’s Hospital Pediatric Clinical Research Center and provided written informed consent. Participants then had a morning fasting blood draw as well as other health assessments, and they completed psychosocial questionnaires. All sessions took place during the first 7 days of the follicular stage of the menstrual cycle, when sex hormones are lowest.

The participants were mothers of either healthy children (n = 19) or children with a chronic condition (n = 39). The average number of children the participants had was two, the range was from one to four. There was no difference in the control group (M = 2.1) and caregiver group (M = 1.8) in average number of children.

Participants were free of any chronic health conditions except controlled hypertension with Beta blockers or ACE Inhibitors (n = 2) and controlled hypothyroidism with Synthroid supplementation (n = 1). Results were similar when excluding these three participants. Other medications included oral contraceptives (30%; equal use in each group) and antidepressant use (all SSRIs) among caregivers (7%). Nine percent currently smoked, but refrained from smoking the morning of the blood draw. Oral contraceptive use, smoking, and antidepressant use were not related to TL. Participants did not consume >7 alcoholic beverages per week and refrained from alcohol the day of the study. Although all women were still menstruating and were studied during the follicular stage (unless they were taking oral contraceptives), some could have been in the perimenopausal stage, which was not determined with hormone tests or by tracking the regularity of menses.

Measurement of Stress. Objective stress. Objective stress was measured by caregiver group status (i.e., being a mother of a child with a chronic health condition), which is considered to be a stressful ongoing situation by most people. They had to identify as the primary caregiver, i.e., responsible for most of the child’s care. The children’s conditions required special care and included a range of illnesses, including a pervasive developmental delay, such as autism, neurological disorders, such as cerebral palsy, and gastrointestinal disorders, such as short gut. The control group of mothers had to have at least one biological child, and none of their children could have a chronic condition.

A second measure of objective stress was duration of caregiving. This measure was only relevant to the caregiving mothers. Duration of caregiving was operationalized as the number of years, rounded to the nearest half year, that the child had the chronic condition. In the majority of cases, this was determined by the date the child received a diagnosis. Caregivers had children who had a chronic condition for an average of 5.9 years (SD = 3.3), and the range was from 1 to 12 years. Because the mother’s age can be a confound for the chronicity of the child’s condition (older mothers might have been caregivers for a longer period), age was controlled for in all analyses.

Perceived stress. Perceived stress was measured by using the 10-item version of the Perceived Stress Scale (PSS) (5, 6), which is the most commonly used and extensively validated scale for assessment of perceived stress. The items assess aspects of stress, such as feeling overwhelmed, stressed, or not having control over important things, rated from never (0) to very often (4) over the last month. High scores on this scale have been related to several objective indices of health, such as infectious illness in a community sample (7). Normal values for adults >20 years old from a 1983 Harris poll of a representative U.S. sample are available (6). The average stress score for a typical sample of adult women is 13.7 ± 6.6. The average stress score for the current sample was 16.7 ± 6.9. As expected, caregivers were higher on perceived stress than the control group of women (with mean scores of 18.3 versus 13.1, P < 0.01).

Categorization into extreme high- and low-stress groups. The primary analyses were undertaken on the highest and lowest quartiles of stress scores. The mean for the highest quartile stress group was 21.5, which can be considered "high stress," whereas the mean for the lowest quartile stress group was 10.5, which can be considered "low stress," based on the published normal values described above (6).

Thirteen of 14 participants who fell into the high-stress group were caregivers, which was expected. There were both control mothers (57%) and caregiving mothers (43%) in the low-stress group. The overrepresentation of caregivers in the high perceived stress group was significant (c ² = 8.0, P < 0.005).

The caregiving mothers in the high perceived stress group tended to have a longer duration of caregiving (M = 6.4 years, SE = 0.79) compared with the caregivers in the low-stress group (M = 4.6 years, SE = 1.1), although this was not a significant difference (P = 0.21).

Further detail on select assay protocols. Cell preparation for TL and telomerase. Blood was drawn from the dominant arm with a butterfly tube. Thirteen milliliters of whole blood were drawn into a lavender top tube. Blood was kept at room temperature and processed within 1 hour of collection by using sterile technique. PBMCs (lymphocytes and monocytes) were isolated by using density-gradient centrifugation (with Ficoll- Paque PLUS). Cells were washed twice and put in freezing medium (10% DMSO/20% FBS/RPMI medium 1640) and immediately frozen at –80.

TL. DNA was extracted from PBMCs by the University of California, San Francisco DNA bank. Genomic DNA isolation was performed by using a standardized and quality-controlled PureGene DNA isolation system (Gentra Systems, Minneapolis). The quantity and quality of the genomic DNA isolate was determined by 260/280 UV spectrophotometery. At regular intervals, the integrity of isolated DNA was evaluated by agarose gel electrophoresis performed on randomly selected isolates.

Measurement of relative TLs (T/S ratios) by quantitative PCR was performed as described in ref. 8 with the following modifications. The primers for the telomere PCR were tel1b [5'-CGGTTT(GTTTGG)₅GTT-3'] and tel2b [5'-GGCTTG(CCTTAC)₅CCT-3'], each used at a final concentration of 900 nM. The primers for the single-copy gene (retinoic X receptor b gene) PCR were rxrbu5 (5'-CCAGAGTCTTTCTCTCATGGGCTTCCTCGTGCTCAGCTAATCC-3') and rxrbd7 (5'-GCGGCCCAAGACATGATCCCTGGCTGAGAGT-3'), each used at a final concentration of 300 nM. A heat-activated Stoffel fragment of TaqDNA polymerase (Titanium Taq) and accompanying 10´ buffer were purchased from Becton Dickinson Biosciences. The final reaction mix contained 40 mM Tricine·KOH, pH 8.0, 18 mM KCl, 3.5 mM MgCl₂, 0.2 mM of each dNTP, 3.75 m g of BSA per m l, 1% DMSO, 2.5 mM DTT, 0.4´ Sybr Green 1 (Molecular Probes), and » 8 ng of the research subject’s DNA, in a final volume of 20 m l. Tubes containing 40, 20, 10, 5, and 2.5 ng of a reference DNA were included in each PCR run so that the quantity of targeted templates in each research sample could be determined relative to the reference DNA sample by the standard curve method. The same reference DNA was used for all PCR runs.

All PCRs were carried out on a Rotor Gene 3000 (Corbett Research, Mortlake, Australia) real-time PCR machine with the 72-tube capacity carousel, 0.1-ml reaction tubes, and version 5 of the QPCR analysis software. The telomere thermal cycling profile consisted of 1 min at 95°C, followed by 20 cycles of 95°C for 15 sec, 56°C for 30 sec, and 68°C for 30 sec, with signal acquisition. The single-copy gene (rxrb) thermal cycling profile consisted of 1 min at 95°C, followed by 32 cycles of 95°C for 20 sec and 68°C for 20 sec with signal acquisition.

The above method was also used to determine the T/S ratios, relative to the same reference DNA, for a separate set of 84 DNA samples with known mean terminal restriction fragment (TRF) lengths (8). The slope of the plot of mean TRF length versus T/S for these samples served as the conversion factor for calculation of approximate TLs in bp for each T/S ratio in this study.

Telomerase activity. Telomerase activity was determined by the telomerase repeat amplification protocol (TRAP) (9) by using a commercial kit (Trapeze, Chemicon). PBMCs stored at –70°C were thawed at 37°C, washed three times with 10 ml cold PBS, and resuspended in 1 ml of cold PBS. Cells were counted with a hemocytometer, and between 5 ´ 10⁵ to 1 ´ 10⁶ cells were pelleted and lysed with 1´ 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS) buffer as directed by the manual for the Trapeze kit. An extract corresponding to 4,000 cells per m l was made, and three concentrations (4,000, 10,000, and 20,000 cells) were used in TRAP reactions to ensure that the assay was in the linear range. The reaction was carried out according to the manual for Trapeze kit and run on a 10% polyacrylamide–8M urea sequencing gel. The signals were quantified by a PhosphorImager (Molecular Dynamics) using 0.02 amol (1.2 × 10⁵ molecules) of the TSR8 quantification template control provided in the kit. Telomerase activity was expressed as total product generated (TPG) units by using the following formula: TPG(units) = [(x – x₀)/c]/[(r – r₀)/c_R], where x is the signal from the products for the sample, x₀ is signal from negative control, c is signal from internal control for that sample, r is signal from the products for TSR8 quantification control, r₀ is signal from negative control for the TSR8, and c_R is signal from internal control for TSR8.

Leukocyte populations. Four milliliters of whole blood were drawn into a sodium heparin tube (green top) (Vacutainer, Becton Dickinson) and maintained at room temperature. Blood was shipped in a well-insulated Styrofoam container to maintain room temperature during shipment by overnight mail to the Dhabhar Laboratory (Ohio State University, Columbus) for flow cytometric analyses. Total leukocyte counts were obtained on a hematology analyzer (F800, Sysmex, McGraw Park, IL). Specific leukocyte subtypes were identified by fluorescently conjugated antibody-staining coupled with flow cytometry (FACSCalibur, Becton Dickinson). Whole blood was stained by using mouse anti-human FITC-conjugated-CD45RA (clone HI100), PE-conjugated-CD45RO (clone UCHL1) and CyChrome-conjugated-CD3 (clone UCHT1) monoclonal antibodies (Pharmingen). Briefly, cell suspensions were incubated with antibody for 20 min at room temperature, lysed and fixed with FACSLysing solution (Becton Dickinson), washed with PBS, and read on the FACSCalibur, with 3,000–5,000 events being acquired from each preparation. Matched antibody isotype controls were run for every sample and used to set negative staining criteria. Lymphocyte, monocyte, and neutrophil populations were identified by using forward- versus side-scatter characteristics. T cell-related data were identified and analyzed within the lymphocyte gate by using CELLQUEST PRO analysis software (Becton Dickinson), and the percentage of positive cells were calculated. Absolute numbers of each specific leukocyte population were calculated taking into account the total leukocyte count for each sample and the relative percentage of the specific leukocyte subtype within that sample. CD3⁺CD45RA⁺ cells were identified as naïve T cells and CD3⁺CD45RO⁺ cells were identified as memory T cells.

1. Maini, M. K., Soares, M. V., Zilch, C. F., Akbar, A. N. & Beverley, P. C. (1999) J. Immunol. 162, 4521–4526.

2. Plunkett, F. J., Soares, M. V., Annels, N., Hislop, A., Ivory, K., Lowdell, M., Salmon, M., Rickinson, A. & Akbar, A. N. (2001) Blood 97, 700–707.

3. Lee, W. W., Nam, K. H., Terao, K. & Yoshikawa, Y. (2002) Immunology 105, 458–465.

4. Hodes, R. J., Hathcock, K. S. & Weng, N. P. (2002) Nat. Rev. Immunol. 2, 699–706.

5. Cohen, S., Kamarck, T. & Mermelstein, R. (1983) J. Health Soc. Behav. 24, 385–396.

6. Cohen, S. & Williamson, G. (1988) in The Social Psychology of Health: Claremont Symposium on Applied Social Psychology, ed. Oskamp, S. S. S. (Sage, Newbury Park, CA).

7. Cohen, S., Tyrrell, D. A. & Smith, A. P. (1993) J. Pers. Soc. Psychol. 64, 131–140.

8. Cawthon, R. M. (2002) Nucleic Acids Res. 30, e47.

9. Kim, N. W. & Wu, F. (1997) Nucleic Acids Res. 25, 2595–2597.