**Cell Senescence and the Human Intervertebral Disc
H.E. Gruber, PhD; E.N. Hanley, Jr., MD; H.J. Norton, PhD
Cell senescence has been recognized as a potential cause of disc degeneration, but little basic research has addressed this topic. Senescent cells are viable and metabolically active but they cannot divide. Senescent cells have altered patterns of gene expression, altered responsiveness to external stimuli, and altered interaction with the extracellular microenvironment. Senescence in the disc is important since it may reduce the disc’s ability to generate new cells to replace existing ones lost to necrosis or apoptosis. It is also possible that the products of gene expression of senescent cells may effect disc physiology. The following paragraphs summarize our findings to date on in vivo and in vitro experiments.
1) Increased cell senescence is present in vivo in the human disc in association with decreased cell proliferation. The objective of this study was to determine the relationship in vivo between disc degeneration, the proportion of senescent cells, and the proportion of proliferating cells in the human annulus. Studies were approved by our human subjects Institutional Review Board. Annulus tissue from 26 discs from 24 subjects was studied. 22 were surgical patients and 2 were control donors. There were 24 lumbar specimens and 2 cervical specimens. Distribution according to Thompson grades was: 1 grade I disc, 4 grade II discs, 9 grade III discs, and 12 grade IV discs. Mean subject age was 42.3 years + 14.3 (24) (mean + SD(n)). Immunohistochemical localizations were performed to detect the Ki-67 antigen which marks proliferating cells (Fig. 1A), and senescence-associated beta-galactosidase which identifies senescent cells at pH 6.0 (Fig. 1B). (Arrows mark cells negative for the immunocytochemical localizations.) A significant negative correlation was present between the % senescent cells vs the % proliferating cells (Fig. 1C), r = -0.013, p = 0.013. In addition, more degenerated grades III and IV discs contained significantly greater percentages of senescent annulus cells than did the healthier grades I and II discs (44.4% + 20.0 (21) vs 18.8% + 11.0 (5), respectively; p = 0.011). There was no significant difference for the % proliferating cells for these two groups (4.7% + 1.6 (21) for grades III and IV vs 5.3% + 1.9 (5) for grades I and II). No correlation was present between age and the % senescent cells or age and the % proliferating cells.
2) Exposure of human annulus cells to hydrogen peroxide (H2O2) in vitro is a reliable, reproducible model of stress-induced senescence: Oxidative stress has been used previously to induce senescence in other cell types, but to date there have been no studies to see if stress-induced premature senescence could be induced in human annulus cells by H2O2 exposure. As shown in Figure 2, exposure to 50 µM. H2O2 for 2 hours results in a significantly elevated incidence of senescence 4 days post-treatment, p <0.0001. An example from a culture of H2O2-treated annulus cells with senescence-associated beta-galactosidase localization in senescent cells is shown in Figure 3A (Fig. 3B shows the relevant negative control).
In summary, results identified to date: 1) Help us better understand the quantitative presence of senescent cells and proliferating cells within the human annulus. A significant negative correlation was present between the proportion of senescent and proliferating cells, such that increased cell senescence is present in association with decreased cell proliferation in vivo; 2) Verify that stress-induced premature senescence induced in human annulus cells by H2O2 exposure is a reliable and reproducible experimental model; studies are now underway during the last portion of this grant to test the ability of selected growth factors to rescue disc cells from senescence in vitro.
**Local Delivery of Small Inhibitory RNA (siRNA) and Radiation Therapy to Treat Metastatic Spine Tumors
D.M. Sciubba, MD; Z.L. Gokaslan, MD, FACS; T.L. DeWeese, MD
Objective: When eukaryotic cells are unable to repair DNA damage, such as double-stranded breaks that are sustained commonly with ionizing radiation, apoptosis results. DNA protein kinase (DNA-PK) is an endogenously expressed protein that facilitates the repair of DNA double-stranded breaks. By using short-inhibitory RNA (siRNA) to DNA-PK in NMU breast cancer cells, we have sought to create a possible molecular radiation sensitizer for metastatic breast cancer to the spine.
Methods: NMU breast cancer cells were cultured in vitro. Cells were transfected with either siRNA to DNA-PK (via a 5 Kb plasmid provided by T.L. DeWeese), green fluorescent protein (GFP; to test the efficacy of transfection), or not transfected at all. Forty-eight hours after transfection, GFP transfected cells were observed for level of transfection, siRNA transfected cells were re-suspended and plated for radiation (5Gy dose using Gammacell 40 irradiator); nine to ten days after radiation, cells were stained with cresyl violet and the colonies were counted and analyzed. The remaining cells were lysed and proteins were run on a Western blot assay. Protein concentration was compared between transfected and non-transfected cell lines, using the endogenously expressed beta-actin as a baseline control for protein expression.
Results: Transfected NMU breast cancer cells showed a 20% decrease in the expression of DNA-PK using Western blot analysis; following in vitro radiation. In addition, cell death in cancer cells was increased following successful transfection of siRNA to DNA-PK.
Conclusion: NMU breast cancer cells transfected with a plasmid-based siRNA to DNA-PK yielded a modest decrease in endogenous expression of DNA-PK, with concomitant increased cell-kill due to ionizing radiation. Because DNA-PK is a key protein involved in the repair of DNA double stranded breaks, these results suggest that siRNA-mediated silencing of this molecule may permit the increased sensitization of metastatic breast cancer to the spine with ionizing radiation.
Figure 1. GFP Transfection at 40% Confluency
Figure 2. Western blot analysis demonstrating 20% downregulation of DNA-PK and graphical representation of reduced DNA-PK rxpression when compared to Beta-Actin baseline.
**The Regulation of Gene Expression to Compressive Stress in the Nucleus Pulposus
G. Sowa, MD, PhD; J. Kang, MD; P. Smolinski, PhD
BACKGROUND: Despite the high prevalence of degenerative disc disease, the early molecular events involved in initiating disc degeneration are not well understood. The intervertebral disc is the major compression carrying component of the spine, with the nucleus experiencing compressive stress and the annulus experiencing tensile stress. In addition to potential destructive effects of mechanical forces, it is clear that beneficial effects of motion and load exist as well. In addition, it is clear that both inflammatory mediators and mechanical forces are involved in this process and interact to lead to disc degeneration. However, the signaling pathways initiated by mechanical stress in the intervertebral disc have not been elucidated. The goal of this work was to investigate the gene expression response of nucleus pulposus cells to compressive stress and explore the existence of a threshold effect, in which low magnitudes of compressive pressure may stimulate a reparative or anti-inflammatory effect and high magnitudes of compressive pressure may stimulate a traumatic or pro-inflammatory effect. We also proposed to determine if this threshold differs between tissue from healthy and degenerative intervertebral discs.
METHODS: Healthy nucleus pulposus cells from New Zealand white rabbits or human surgical samples (degenerative grade 3) were cultured in three dimensions in alginate beads. The cells were suspended in dialysis tubing bathed in tissue culture media (F12, 1% penicillin, streptomycin, 10% fetal bovine serum) within a standard tissue culture incubator (5% CO2, 37° C). Cells were exposed to a compressive stress of 0.7, 2, and 4 MPa (100, 300, 600 psi)for four hours in a custom designed compression chamber. Gene expression of markers of inflammation (inducible nitric oxide synthase; iNOS), of matrix catabolism (matrix metalloproteases-3; MMP-3) and anti-catabolic metabolism (tissue inhibitor of metalloproteases; TIMP-1) was examined using real time PCR compared to unstressed controls. Expression of GAPDH was used as the housekeeping gene and relative gene expression levels calculated compared to control.
RESULTS: Healthy rabbit cells demonstrated a threshold response to compressive stress, with decreased expression of catabolic genes (iNOS and MMP-3) at 100 and 300 psi, and increased expression at 600 psi. Anti-catabolic gene TIMP-1 showed a decrease in response to 100 psi, minimal change in response to 300 psi, and increase in response to 600 psi. Human degenerative cells demonstrated a lower threshold to conversion to traumatic response to compressive stress, with decreased iNOS and increased TIMP-1 at 100 psi, decreased iNOS and minimal change in TIMP at 300 psi, and increase in both iNOS and TIMP at 600 psi.
CONCLUSIONS: In normal rabbit cells, while catabolic mediators iNOS and MMP-3 were decreased in response to low levels of compressive stress (100 psi), anti-catabolic mediator TIMP-1 was also decreased. An inverse pattern was noted with high magnitudes (600 psi)of compression, where catabolic and anti-catabolic mediators were both increased. However, for moderate levels of stress (300 psi), catabolic mediators were decreased and anti-catabolic mediators increased, suggesting that this was the most beneficial level of stress for healthy rabbit cells. For degenerative human cells, however, 100 psi, which demonstrated decreased catabolic and increased anti-catabolic gene expression, was the most beneficial magnitude of stress. It is not clear if this difference is related to difference in species or degenerative phenotype. Future work examining the gene expression changes in degenerative rabbit cells and normal human cells will facilitate direct comparisons between normal and degenerative cells. In addition, future work investigating additional catabolic, anti-catabolic and anabolic genes will provide important information regarding the overall metabolic balance within the cell in response to each level of compressive stress.
**The Minimal Clinically Important Difference (MCID) for Spinal Disorders: Finding the Threshold of Clinically Significant Change
V. Deviren, MD; S. Berven, MD
STATUS OF RESEARCH PROGRESS: Currently, we have collected data from 115 consecutive lumbar patients and 43 consecutive cervical patients who had undergone operative treatment for degenerative spinal disorders. We are continuing enrollment at this time.
BACKGROUND CONTEXT: Patient-reported outcomes instruments, including the Medical Outcomes Short Form 36 (SF-36) and the Oswestry Disability Index (ODI), are commonly used to evaluate the effectiveness of spinal surgery. As we begin to utilize patient-based health status outcomes as a measure of utility for a given intervention, it becomes important to differentiate between statistically and clinically significant changes in health status. The minimum clinically important difference (MCID) represents the smallest improvement considered worthwhile by a patient. The concept of the MCID is offered as a new standard for determining the effectiveness of a given treatment and describing patient satisfaction in reference to that treatment.
PURPOSE:
The primary aim of this study is to establish the threshold at which the MCID occurs for patients with variety of degenerative disorders on the SF-36 and ODI. The maximum clinically important difference (MaCID), representing a major improvement in health status as recognized by the patient, is also be examined.
METHODS: This is a prospective, observational study. The primary means of determining the MCID for patients are divided into anchor-based and distribution-based methods. Outcomes data SF-36, Oswestry Disability Index, General Self-Assessment (14-item questionnaire), and Visual Analog Scale (VAS) were analyzed using both methods.
Changes in outcomes scores over time (baseline, 6 weeks, 3 months, 6 months, 1 year, 2 years) were compared. The MCID for function, as measured by the ODI, and separate MCIDs for general well being, pain, mood, and relationships, as measured by the SF-36 total score, the Bodily Pain Domain of the SF-36, the Mental Health Component Score of the SF-36, and the Social Function Domain of the SF-36, respectively, were established for patients with degenerative spinal disorders. Two different techniques (anchor-based and distribution-based) were used to establish the MCID and these results were compared.
In this study we used 15-item Likert scale questionnaire (See Figure 1) to evaluate treatment effect as global assessments of change and to provide an anchor for domains of general health, pain, mood (mental health), relationships (social function), and physical function. We developed a visual analog scale (VAS) (See Figure 2) for each domain in order to obtain a more subjective anchor to differentiate the MCID.
Figure 1 Figure 2
Using the anchor-based technique for determining the MCID, we compared observed changes in the SF-36 and ODI scores from pre-treatment to post-treatment with patient self report change assessments. For the Likert scale we defined a positive minimal clinically significant difference as a change that is recognized as “a little better”, “somewhat better”, and “moderately better” by the respondent and a negative minimal clinically significant difference as a change that is recognized as “a little worse”, “somewhat worse”, and “moderately worse” by the respondent.
To validate the use of retrospective self-assessment questionnaires (GAQ, VAS) in assessing change in health status, we compared the mean changes in health status for those self-assessing improvement, no change, and worsening of health. Rank correlation analysis (Spearman’s rho) and linear regression was performed separately for each MCID domain (well-being, pain, function, mood, and relationships) on the GAQ and VAS in order to validate the use of each domain.
Using two distribution based methods (SEM, 95 % confidence interval, and standard deviation), we measured the MCID. Using the 95% confidence interval method, the MCID for improvement and deterioration was defined as a range, with the lower confidence interval of the group with minimum change being the lowest estimate, and the highest confidence interval of the group reporting no change being the highest estimate.
We utilized the standard error of measurement (SEM) as a second distribution-based method to assess the MCID for our patient groups. The SEM is calculated using the formula:
SEM= σx√ 1-ru
where σx is the standard deviation of a HRQoL measure and ru is the reliability coefficient. The advantage of using the SEM is that it, unlike effect size or the ratio of difference to variance, is not reliant on the sample, i.e. it is consistent across the entire range of a measurement. [16] Wyrwich et al. has found that 1 SEM criterion corresponds to external anchors for MCIDs for various HRQoL questionnaires for a variety of different diseases. [16]
PILOT RESULTS: Outcomes data was prospectively collected from 115 consecutive lumbar patients and 43 consecutive cervical patients who had undergone operative treatment for degenerative spinal disorders. See Table 1a and 1b. The average age of the lumbar cohort was 55 (range 17-82), and was comprised of 61 females and 54 males. The average age of the cervical cohort was 52 (range 32-87), and was comprised of 29 females and 14 males. Datapoints were collected at baseline, 6 weeks (n, cervical=14; n, lumbar=40) 3 months post-op (n, cervical=15; n, lumbar=38), 6 months post-op (n, cervical=14; n, lumbar=36), 1 year post-op (n, cervical=6; n, lumbar=31), 2 year post-op (n, cervical=2; n, lumbar=3).
To assess the validity of our VAS questionnaire as a potential anchor, we ran Spearman two-tailed correlation analysis for the VAS questionnaire and the 15-item Likert scale for overall well-being, pain, function, mood (mental health), relationships (social function). We found a significant correlation between the two scales in each of the domains (Likert Well-Being/VAS Well-Being r=0.60, Likert Pain/VAS Pain r=0.73, Likert Function/VAS Function r=0.77, Likert Mood/VAS Mood r=0.74, Likert Relationships/VAS Relationships r=0.58).
In the lumbar cohort, at 114 datapoints patients reported experiencing an overall improvement from their pre-operative health status, while 10 patients reported experiencing no change and 20 reported overall worsening (Table 3a and 3b). In the cervical cohort, at 32 datapoints patients reported experiencing an overall improvement from their pre-operative health status, while 9 patients reported experiencing no change and 10 reported overall worsening (Table 3c and 3d). We ran Spearman two-tailed correlation analysis and reported only the moderate or greater r values for the correlations between the SF-36 domain scores/ODI/NDI scores and the GAQ and VAS self-assessment ratings. listed in Table 2.
Using the patient self-report 15-item questionnaire as an anchor, the MCID for general well-being, pain, function, mood, and relationships were measured. Patients reporting a minimum improvement (≤4 points) were included in the analysis. Each domain of the MCID was assessed according to the score that was most strongly correlated with (if the correlation was statistically significant). Therefore, the determination of the MCID for general well-being, as measured on a 15 point scale, was determined using the change in SF-36 total score. For pain, the subjects’ self-assessment was determined using the Bodily Pain domain of the SF-36. For function, the MCID was determined using the change in ODI total score for the lumbar group and the PhyFcn SF-36 domain for the cervical group. For mood, the Mental Health component of the SF-36 was used to determine the MCID. For relationships, the Social Function Domain of the SF-36 was used to determine the MCID for relationships, although there was not even a moderate correlation. Only patients reporting improvement in a specific domain were included in the analysis of that domain. If the direction of change in outcomes scores conflicted with the direction of change of patient self-report, subjects were excluded from the analysis. We compared patients reporting a minimum positive change (≤4 points) with patients reporting a maximum positive change (≥5 points). The values for the MCID and the MaCID are reported in Table 4a and 4b.
DISCUSSION: Our findings suggest that the value of outcomes change scores that reflect a minimal clinically important difference from the perspective of the patient are, indeed, small. Our results demonstrate, importantly, that the MCID should be determined in regard to specific domains, as the majority of patients reported different scores in each domain (well-being, pain, function, mood, relationships) and the MCIDs differed according the domain examined. While the relative importance of each specific domain can be compared to the overall well-being change rating, it would be interesting to examine these relationships in further detail. Although our data collection is not yet complete and we have not yet reached sample size in each domain, we found it interesting to discuss some of our preliminary findings in this pilot report.
Our sub group analysis (lumbar/cervical) show that the use of the VAS as an anchor is feasible and may be best used when accounting for global change in health status not related to a specific disease state. The VAS scores were more strongly correlated across the board to changes in SF-36 outcomes scores than the 15-item assessment (which allows more freedom than the typically used 6 or 7 item questionnaires). The use of the VAS as anchor eliminates one important criticism of the anchor-based method; that the meaning of change differs according to the arbitrary divisions of the anchor’s scale. Although, conceptually, the MCID is the difference between a “no change” category and a “minimum change” category on a scale, actually evaluating the MCID in this manner is problematic and self-limiting. The use of a categorical anchor may not represent the patient’s actual change in health status, therefore use of a continuous anchor, such as the VAS, would be a better assessment tool. Because the VAS is continuous, unlike a Likert scale, and is not subject to arbitrary divisions, the meaning of the MCID determined using a VAS as an anchor will not differ according to the level or number of divisions of the anchor scale. One potential problem is that the level at which a minimum change vs. no change is delineated has not been established. However, to be able to validate the use of the VAS, we still need further enrollment to reach our estimated sample size.
While this preliminary data is able to serve as a starting point for determining the MCID for degenerative spinal disorders, much work remains to be done in order for a truly comprehensive and clinically useful MCID value to be determined. Different methods of determining the MCID (anchor-based vs. distributional) yield vastly different results and it remains open to discussion which method is more accurate.
Several limitations of the data relate to the number of patients on which data was able to be collected. We were able to collect data on 115 consecutive patients with degenerative lumbar conditions (See Table 1). While we were able analyze the pilot data by combining all diagnoses and timepoints and report significant results, we feel that more data must be collected until we have reached the sample size predicted by our power calculation (n=38) for each diagnosis and timepoint. Furthermore, to be able to form a complete picture of the MCID, we believe that the calculated sample size must be reached for each group (“no change”, “minimum change”, “maximum change”) in addition to diagnosis and timepoints.
We collected data on 43 consecutive patients with degenerative cervical conditions. Once patients were categorized as either “small improvement”, “no change” or “large improvement” the resulting groups sizes varied between 13-22 with disparate diagnoses and timepoints included in each group reporting minimal change. More data should be collected to validate the results of the cervical group.
Additionally, for both the cervical and lumbar branches of the study, too few patients reported deteriorations to be able to calculate MCID values for deterioration. Although it would be beneficial to calculate an MCID for deterioration, it does not seem practical at this time. As predicted, the number of patients reporting deteriorations is very small. It would take an inordinate amount of time to amass a significant number of patients rating themselves as worse than baseline.
The MCID is influenced by a wide variety of factors and it is important to be able to tease these factors apart. We were unable, due statistical power issues, to stratify patients based on diagnosis and surgery. Additionally, we did not separately analyze the changes scores while controlling for the floor and ceiling effect exhibited in baseline outcomes scores. We believe that the MCID is temporally dependent and will change over time. Because of this, patients should be grouped based, not only on diagnosis, but also on time out from surgery. In further work, it is important to determine the specific impact of baseline status, post-operative time point, and diagnosis on the MCID.
A further consideration is the impact of the patient’s expectations at the outset of treatment and the impact pre-operative expectations on post-operative assessments of change. While this is beyond the current scope of our project, it would be interesting to examine this in the future.
CONCLUSIONS: An ideal means of determining the MCID for a given intervention is yet to be determined. While we were able to establish MCIDs, these numbers may be subject to change in the future. Our pilot MCID values are subject to change as more data is amassed and sample size is reached. Additionally, because most of the follow-up reported on is short term (lumbar average= 4.49 months; cervical average= 5.38 months), subject reports of pain and function may still be influenced by the proximity of surgery and follow-up timepoint. As subjects are further out from surgery (1 year, 2 year), we believe the outcomes data will not reflect pain and loss of function due to the surgery itself and will, therefore, be more accurate. On the contrary, it may take longer to reach a significant sample size further out from surgery, as many patients may report significant improvement at later timepoints compared to earlier one.
There are several limitations in the accurate determination of an MCID have been identified: 1) the multiplicity of MCID determinations (i.e. different methods yield different results) 2) patients report significant changes early on in the post-op period and we predict will continue to report significant changes further out from surgery, making it take longer to enroll a big enough sample of patients reporting minimal change at each timepoint 3) the dependence of the MCID on pre-treatment baseline status and patient expectations. We believe it is possible to develop a useful method provided that the assumptions and methodology are initially declared. Our efforts toward the establishment of a MCID rely upon the establishment of specific external criteria based upon the symptoms of the patient and treatment intervention being evaluated.
Based on our preliminary results, an ideal means of determining the MCID for a given intervention is yet to be determined. Even though the MCID should be applicable to any follow-up time point, subject reports of pain and function may be influenced by the proximity of surgery with short follow-up time point (≤ 6 months). As subjects are further out from surgery we believe the outcomes data will be more evenly distributed. Several limitations are detected in the accurate determination of the MCID: the multiplicity of MCID determinations, conflicting patient self-report, individual perception, wide variation among patient’s scores, and the relationship between pre-treatment baseline and post-treatment change scores.
*Abstracts/permission forms not received at the time of publication
**Current and ongoing research