MicroRNA expression and functional analysis in Hodgkin lymphoma
Yuan, Ye
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2019
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Yuan, Y. (2019). MicroRNA expression and functional analysis in Hodgkin lymphoma. University of Groningen.
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
Res Commun, 2017. 485(1): p. 181-188.
41. Patil, S., et al., Smad7 is induced by CD40 and protects WEHI 231 B-lymphocytes from transforming growth factor-beta -induced growth inhibition and apoptosis. J Biol Chem, 2000. 275(49): p. 38363-70.
42. Cottonham, C.L., S. Kaneko, and L. Xu, miR-21 and miR-31 converge on TIAM1 to regulate migration and invasion of colon carcinoma cells. J Biol Chem, 2010. 285(46): p. 35293-302. 43. Frankel, L.B., et al., Programmed cell death 4 (PDCD4) is an important functional target of
the microRNA miR-21 in breast cancer cells. J Biol Chem, 2008. 283(2): p. 1026-33. 44. Yan, L.X., et al., MicroRNA miR-21 overexpression in human breast cancer is associated
with advanced clinical stage, lymph node metastasis and patient poor prognosis. RNA, 2008. 14(11): p. 2348-60.
45. Zhang, J.G., et al., MicroRNA-21 (miR-21) represses tumor suppressor PTEN and promotes growth and invasion in non-small cell lung cancer (NSCLC). Clin Chim Acta, 2010. 411(11-12): p. 846-52.
46. Ou, H., Y. Li, and M. Kang, Activation of miR-21 by STAT3 induces proliferation and suppresses apoptosis in nasopharyngeal carcinoma by targeting PTEN gene. PLoS One, 2014. 9(11): p. e109929.
47. Go, H., et al., MicroRNA-21 plays an oncogenic role by targeting FOXO1 and activating the PI3K/AKT pathway in diffuse large B-cell lymphoma. Oncotarget, 2015. 6(17): p. 15035-49. 48. Jones, K., et al., Plasma microRNA are disease response biomarkers in classical Hodgkin
lymphoma. Clin Cancer Res, 2014. 20(1): p. 253-64.
49. van Eijndhoven, M.A., et al., Plasma vesicle miRNAs for therapy response monitoring in Hodgkin lymphoma patients. JCI Insight, 2016. 1(19): p. e89631.
CHAPTER 6
Summary, Discussion
and Future perspectives
SUMMARY AND DISCUSSION
Aberrant miRNA expression profiles have been reported in many human diseases including most cancer subtypes [1-3]. Several studies determined miRNA expression profiles in B-cell lymphoma and more specifically in Hodgkin lymphoma (HL) [4-9]. For a subset of these miRNAs important roles have been shown in the pathogenesis of B-cell lymphomas [10-13]. The knowledge on the functional role of differentially expressed miRNAs in HL is still limited. In this thesis, we aimed to (1) identify miRNAs differentially expressed in HL cell lines compared to germinal center (GC)-B cells using small RNA sequencing, (2) determine which miRNAs affect HL cell growth and (3) identify target genes of these miRNAs. To achieve this, we implemented and applied several high throughput screenings approaches.
The miRNA expression profile of HL cells
Several studies have generated miRNA expression profiles of HL cell lines, total tissue samples and purified HRS cells [8, 9]. In chapter 2, we were the first to use small RNA sequencing to identify miRNAs with deregulated expression in HL cell lines compared to GC-B cells. The expression levels of the top-10 most abundant miRNAs accounted for >60% of all reads in both HL and GC-B cells with 6 of the top-10 miRNAs overlapping between both cell types. Comparison of the miRNA profiles of HL cell lines and GC-B cells revealed 84 significantly differentially expressed miRNAs. MiR-23a-3p, miR-24-3p and miR-27a-3p, all three derived from one primary-miRNA transcript, were among the 84 miRNAs with increased levels in HL cell lines compared to GC-B cells. Validation of the small RNA-seq data by qRT-PCR confirmed the expected pattern for 11 of 15 miRNAs with increased and for 4 of 7 miRNAs with decreased expression levels in HL cell lines compared to GC-B cells.
High throughput screens
To explore the relevance and molecular mechanisms of differentially expressed miRNAs we applied two different high throughput screenings approaches. The first series of experiments aimed to identify miRNAs that are involved in proliferation of HL cells. In the second series, we performed Argonaute 2-RNA immunoprecipitation (Ago2-IP) chip experiments to identify HL relevant miRNA target genes and aimed to link individual miRNA targets to the observed phenotypes of specific miRNAs on
growth of HL cells.
Identification of miRNAs affecting cell growth Pilot studies
Before we studied the effects of miRNA modulation on HL cell growth in a high-throughput screen we performed a cellular barcoding experiment to test the feasibility of such an approach in HL. A potential problem might be caused by presence of cancer stem cells (CSCs) in various HL cell lines, as indicated in previous studies. We infected two HL cell lines with a barcoded empty vector (EV-BC) library in chapter 3A. The aim was to set up the technology, exclude potential bias caused by CSC and define cutoff values for follow-up studies with miRNA inhibition or overexpression constructs. We observed a broad variation in the read counts and marked changes in abundance of EV-BC constructs over time. This variation was caused by suboptimal experimental conditions, such as presence of inappropriate EV-BC inserts, multiple independently prepared EV-BC lentiviral pools, and differences in amplification efficiency due to use of different forward and reverse primers. In addition, we relied on prediction of BC sequences present in the pool based on next generation sequencing of the complete plasmid pool. To optimize the experimental conditions, we verified the insert sequences of all individual clones by Sanger sequencing and included only those constructs with appropriate inserts and fully matching primer binding sites in the second screen. In addition, we also made one large viral pool that was used for all infections. These changes led to a marked improvement in the consistency of read counts per construct over time. Nevertheless, we did notice marked differences in the range of the read count ratios over time between the two cell lines, which suggested that the background of such screens may vary per cell line. No obvious changes were observed in the abundance of EV-BC constructs that could be attributed to targeting of CSCs in our experiments. Based on the variation observed in the second screen, we set the cutoff values for true miRNA-induced effects at the mean ratio of read counts of day 13 / day 5 and day 21 / day 5 ± 2× the SD of both cell lines for the experiments described in chapters 3B and 4.
In chapter 3B, we performed a pilot high throughput screening experiment with miRNA inhibition and overexpression constructs in the KM-H2 HL cell line. We aimed to optimize the experimental conditions and design a pipeline to analyze the NGS data for follow-up studies using a larger set of constructs. To validate overexpression of the pCDH miRNA overexpression constructs, HEK-293T cells were infected with a virus
6
SUMMARY AND DISCUSSION
Aberrant miRNA expression profiles have been reported in many human diseases including most cancer subtypes [1-3]. Several studies determined miRNA expression profiles in B-cell lymphoma and more specifically in Hodgkin lymphoma (HL) [4-9]. For a subset of these miRNAs important roles have been shown in the pathogenesis of B-cell lymphomas [10-13]. The knowledge on the functional role of differentially expressed miRNAs in HL is still limited. In this thesis, we aimed to (1) identify miRNAs differentially expressed in HL cell lines compared to germinal center (GC)-B cells using small RNA sequencing, (2) determine which miRNAs affect HL cell growth and (3) identify target genes of these miRNAs. To achieve this, we implemented and applied several high throughput screenings approaches.
The miRNA expression profile of HL cells
Several studies have generated miRNA expression profiles of HL cell lines, total tissue samples and purified HRS cells [8, 9]. In chapter 2, we were the first to use small RNA
sequencing to identify miRNAs with deregulated expression in HL cell lines compared to GC-B cells. The expression levels of the top-10 most abundant miRNAs accounted for >60% of all reads in both HL and GC-B cells with 6 of the top-10 miRNAs overlapping between both cell types. Comparison of the miRNA profiles of HL cell lines and GC-B cells revealed 84 significantly differentially expressed miRNAs. MiR-23a-3p, miR-24-3p and miR-27a-3p, all three derived from one primary-miRNA transcript, were among the 84 miRNAs with increased levels in HL cell lines compared to GC-B cells. Validation of the small RNA-seq data by qRT-PCR confirmed the expected pattern for 11 of 15 miRNAs with increased and for 4 of 7 miRNAs with decreased expression levels in HL cell lines compared to GC-B cells.
High throughput screens
To explore the relevance and molecular mechanisms of differentially expressed miRNAs we applied two different high throughput screenings approaches. The first series of experiments aimed to identify miRNAs that are involved in proliferation of HL cells. In the second series, we performed Argonaute 2-RNA immunoprecipitation (Ago2-IP) chip experiments to identify HL relevant miRNA target genes and aimed to link individual miRNA targets to the observed phenotypes of specific miRNAs on
growth of HL cells.
Identification of miRNAs affecting cell growth
Pilot studies
Before we studied the effects of miRNA modulation on HL cell growth in a high-throughput screen we performed a cellular barcoding experiment to test the feasibility of such an approach in HL. A potential problem might be caused by presence of cancer stem cells (CSCs) in various HL cell lines, as indicated in previous studies. We infected two HL cell lines with a barcoded empty vector (EV-BC) library in chapter 3A. The aim
was to set up the technology, exclude potential bias caused by CSC and define cutoff values for follow-up studies with miRNA inhibition or overexpression constructs. We observed a broad variation in the read counts and marked changes in abundance of EV-BC constructs over time. This variation was caused by suboptimal experimental conditions, such as presence of inappropriate EV-BC inserts, multiple independently prepared EV-BC lentiviral pools, and differences in amplification efficiency due to use of different forward and reverse primers. In addition, we relied on prediction of BC sequences present in the pool based on next generation sequencing of the complete plasmid pool. To optimize the experimental conditions, we verified the insert sequences of all individual clones by Sanger sequencing and included only those constructs with appropriate inserts and fully matching primer binding sites in the second screen. In addition, we also made one large viral pool that was used for all infections. These changes led to a marked improvement in the consistency of read counts per construct over time. Nevertheless, we did notice marked differences in the range of the read count ratios over time between the two cell lines, which suggested that the background of such screens may vary per cell line. No obvious changes were observed in the abundance of EV-BC constructs that could be attributed to targeting of CSCs in our experiments. Based on the variation observed in the second screen, we set the cutoff values for true miRNA-induced effects at the mean ratio of read counts of day 13 / day 5 and day 21 / day 5 ± 2× the SD of both cell lines for the experiments described in
chapters 3B and 4.
In chapter 3B, we performed a pilot high throughput screening experiment with miRNA
inhibition and overexpression constructs in the KM-H2 HL cell line. We aimed to optimize the experimental conditions and design a pipeline to analyze the NGS data for follow-up studies using a larger set of constructs. To validate overexpression of the pCDH miRNA overexpression constructs, HEK-293T cells were infected with a virus
pool including 22 miRNA overexpression constructs and analyzed changes in miRNA expression levels by miRNA microarray. For 16 of the constructs we observed a >2 fold increase in expression values for at least one of the two miRNA strands (i.e. the 3p- or 5p-strand). The effects on the expression levels was limited because we used a virus pool for infection, which leads to presence of each construct in only a limited proportion of the cells. Based on these data we concluded that the majority of the miRNA overexpression constructs indeed led to an increase in the levels of the corresponding mature miRNA.
For the pilot screen, we prepared two virus pools, i.e. a pCDH pool with 30 miRNA overexpression constructs and a miRZIP pool with 34 miRNAs inhibition and 4 short hairpin RNA (shRNA) constructs targeting Dicer or Drosha (2 shRNAs for each gene). For the pCDH constructs, read counts were inversely correlated to the length of the inserts. This caused a broad range in total read counts per construct and was considered to be suboptimal. For one of the pCDH constructs, i.e. pCDH-miR-142, we observed a decrease in read counts at day 27 compared to day 5, whereas no effect was seen for the other constructs. For the miRZIP constructs, the initial read counts were quite similar, consistent with the minor variation in insert sizes (64-73nt). However, read counts for the 4 shRNA constructs targeting Dicer or Drosha were much lower despite similar insert sizes. Most likely, this is caused by the perfect matching stem regions in the design of the shRNAs, which are notoriously difficult to sequence. The miRZIP constructs are standardly designed with imperfectly matching stem sequences (2-3 mismatches), and these appear to be sequenced more effectively. We identified 7 miRZIP constructs with increased read counts and 7 with decreased read counts at day 27. In addition, the read counts of 3 shRNA constructs increased over time. Based on these pilot experiments we further improved the approach for the actual high throughput screens in several ways. We redesigned the pCDH constructs in such a way that all inserts sizes were approximately 500bp. To allow more effective sequencing of shRNA constructs we adapted the procedure for the library preparation for the next generation sequencing. After amplification of the insert sequence, we digested the PCR products with AgsI. This restriction enzyme binds to a unique recognition sequence present in the loop of all hairpin structures. After digestion, the 3’-part of the stem-loop region can be removed by size selection and the resulting 5’-parts are used as input for the next generation sequencing library preparation. In this way, formation of secondary structures that may inhibit sequencing efficiency will be prevented.
In addition, we noticed that the range of the read count ratios observed over time differed per cell line. This could not be explained by experimental differences. Potentially this could lead to false positive or false negative results in the high throughput screen. To avoid this, we explored an alternative data analysis approach using the data of the actual screen to determine threshold values for defining miRNA constructs that are outliers and thus truly affect cell growth of HL cell lines.
MiRNA overexpression screen
In chapter 4, we used the optimized conditions to perform a screen with a pool of 40 pCDH miRNA overexpression constructs in HL cell lines L540 and KM-H2. Compared to the pilot experiment the variation in read counts was clearly decreased although we still observed much higher read counts for pCDH-miR-19b-1. Based on the first data analysis approach, which was based on the cutoff values of the barcoded empty vector experiment as described in chapter 3, pCDH-miR-141 was found to be depleted at day 21 compared to day 5 in at least 3 of 4 infections (2 cell lines each infected in duplo). The alternative method to analyze the data (based on the distribution of the slopes of the read count ratios over time using the Tukey IQR test), revealed an increase of pCDH-miR-19b-1 in at least 3 of 4 infections. Based on the small RNA-seq data described in chapter 2, miR-141 is downregulated in HL compared to normal GC-B cells, albeit not significant. MiR-19b has been indicated as an oncogene in other GC- B-cell lymphoma subtypes [14], whereas the role of miR-141 has not been studied previously. Thus, based on our screening both miR-141 and miR-19b-1 may have potential roles on growth of HL, but this needs to be validated.
A main drawback of the overexpression vector used in our study is the relatively low GFP intensity, which might have hampered effective sorting of all infected cells. Small changes in the sorting gates from day to day might have led to false positive or negative results in our high throughput screen.
MiRNA inhibition screen
In chapter 5 we performed a high-throughput screen with a virus pool containing 63 miRNA inhibition constructs in L540, L428 and KM-H2 HL cell lines. In addition, the miRZIP constructs pool also included 180 short hairpin RNA (shRNA) constructs targeting multiple protein coding and noncoding genes, largely irrelevant for this thesis. Using the Tukey IQR test on slopes of the read count ratios over time, four miRNA inhibition constructs targeting miR-449a-5p, miR-625-5p, let-7f-2-3p and miR-21-5p
6
pool including 22 miRNA overexpression constructs and analyzed changes in miRNA expression levels by miRNA microarray. For 16 of the constructs we observed a >2 fold increase in expression values for at least one of the two miRNA strands (i.e. the 3p- or 5p-strand). The effects on the expression levels was limited because we used a virus pool for infection, which leads to presence of each construct in only a limited proportion of the cells. Based on these data we concluded that the majority of the miRNA overexpression constructs indeed led to an increase in the levels of the corresponding mature miRNA.
For the pilot screen, we prepared two virus pools, i.e. a pCDH pool with 30 miRNA overexpression constructs and a miRZIP pool with 34 miRNAs inhibition and 4 short hairpin RNA (shRNA) constructs targeting Dicer or Drosha (2 shRNAs for each gene). For the pCDH constructs, read counts were inversely correlated to the length of the inserts. This caused a broad range in total read counts per construct and was considered to be suboptimal. For one of the pCDH constructs, i.e. pCDH-miR-142, we observed a decrease in read counts at day 27 compared to day 5, whereas no effect was seen for the other constructs. For the miRZIP constructs, the initial read counts were quite similar, consistent with the minor variation in insert sizes (64-73nt). However, read counts for the 4 shRNA constructs targeting Dicer or Drosha were much lower despite similar insert sizes. Most likely, this is caused by the perfect matching stem regions in the design of the shRNAs, which are notoriously difficult to sequence. The miRZIP constructs are standardly designed with imperfectly matching stem sequences (2-3 mismatches), and these appear to be sequenced more effectively. We identified 7 miRZIP constructs with increased read counts and 7 with decreased read counts at day 27. In addition, the read counts of 3 shRNA constructs increased over time. Based on these pilot experiments we further improved the approach for the actual high throughput screens in several ways. We redesigned the pCDH constructs in such a way that all inserts sizes were approximately 500bp. To allow more effective sequencing of shRNA constructs we adapted the procedure for the library preparation for the next generation sequencing. After amplification of the insert sequence, we digested the PCR products with AgsI. This restriction enzyme binds to a unique recognition sequence present in the loop of all hairpin structures. After digestion, the 3’-part of the stem-loop region can be removed by size selection and the resulting 5’-parts are used as input for the next generation sequencing library preparation. In this way, formation of secondary structures that may inhibit sequencing efficiency will be prevented.
In addition, we noticed that the range of the read count ratios observed over time differed per cell line. This could not be explained by experimental differences. Potentially this could lead to false positive or false negative results in the high throughput screen. To avoid this, we explored an alternative data analysis approach using the data of the actual screen to determine threshold values for defining miRNA constructs that are outliers and thus truly affect cell growth of HL cell lines.
MiRNA overexpression screen
In chapter 4, we used the optimized conditions to perform a screen with a pool of 40
pCDH miRNA overexpression constructs in HL cell lines L540 and KM-H2. Compared to the pilot experiment the variation in read counts was clearly decreased although we still observed much higher read counts for pCDH-miR-19b-1. Based on the first data analysis approach, which was based on the cutoff values of the barcoded empty vector experiment as described in chapter 3, pCDH-miR-141 was found to be depleted at
day 21 compared to day 5 in at least 3 of 4 infections (2 cell lines each infected in duplo). The alternative method to analyze the data (based on the distribution of the slopes of the read count ratios over time using the Tukey IQR test), revealed an increase of pCDH-miR-19b-1 in at least 3 of 4 infections. Based on the small RNA-seq data described in chapter 2, miR-141 is downregulated in HL compared to normal GC-B cells, albeit not significant. MiR-19b has been indicated as an oncogene in other GC- B-cell lymphoma subtypes [14], whereas the role of miR-141 has not been studied previously. Thus, based on our screening both miR-141 and miR-19b-1 may have potential roles on growth of HL, but this needs to be validated.
A main drawback of the overexpression vector used in our study is the relatively low GFP intensity, which might have hampered effective sorting of all infected cells. Small changes in the sorting gates from day to day might have led to false positive or negative results in our high throughput screen.
MiRNA inhibition screen
In chapter 5 we performed a high-throughput screen with a virus pool containing 63
miRNA inhibition constructs in L540, L428 and KM-H2 HL cell lines. In addition, the miRZIP constructs pool also included 180 short hairpin RNA (shRNA) constructs targeting multiple protein coding and noncoding genes, largely irrelevant for this thesis. Using the Tukey IQR test on slopes of the read count ratios over time, four miRNA inhibition constructs targeting miR-449a-5p, miR-625-5p, let-7f-2-3p and miR-21-5p
showed a significant decrease in abundance. None of the constructs showed a significant increase in abundance in at least 4 of 6 infections. For some of the in HL overexpressed or abundant miRNAs for which we included constructs in our miRNA inhibition screen, previous studies suggested a role in HL pathogenesis, i.e. miR-9; [5] miR-96, miR-182 and miR-183 cluster; [15] and the miR-17 seed family (miR-17 and miR-106b) [12]. In our cell lines, miR-9 inhibition did not result in a significant change in abundance in any of the 6 infections. The miR-96, miR-182 and miR-183 cluster was shown to target FOXO1, an important tumor suppressor gene, in HL [15]. However, no study was done to establish the effects of modulating expression of these three miRNAs on the growth of HL cell lines. For miR-182-5p inhibition, we observed no changes in any of the infections and for miR-96 and miR-183 we did not have inhibition constructs in our pool. So, for the latter two miRNAs it remains unknown whether they affect HL cell growth or not. Effects on cell cycle progression were shown in KM-H2 upon inhibition of the miR-17 seed family [12]. We observed a significant decrease in abundance for miR-20a-5p for 2 of 6 infections and for miR-106b-5p for 3 of 6 infections. For miR-17-5p no effects were observed and for miR-106a an increase in abundance was observed in 1 out of 6 infections. Thus, for some of the miR-17 family members we did observe a phenotype upon inhibition, albeit not consistent enough to pass our selection criteria.
ShRNA construct screen
The optimized library preparation procedure resulted in a much smaller difference in read counts between miRNA inhibition (median 328, range 15 – 927) and shRNA constructs (median 103, range 32 – 436) as compared to the pilot experiment described in chapter 3B. Nine shRNA constructs targeting 7 genes were depleted in at least 4 of 6 infections, i.e. BAP1-3, BAP1-4, EZH2-2, shRNA-EZH2-4, shRNA-FLJ-2, shRNA-HNRNPL-2, shRNA-MEF2C-1, shRNA-REL-1 and shRNA-SETD2-3. Again, none of the shRNA constructs showed enrichment over time. In comparison to the pilot experiment, shRNA constructs targeting Dicer or Drosha were not consistently increased in the duplicate infections. The apparent increase in abundance of shRNA constructs targeting Drosha and Dicer in the first experiment might have been caused by the low and thus less reliable read counts of these shRNA constructs. It can also be caused by the strong decrease of several miRZIP constructs in the pilot experiment within a limited total number of constructs as compared to the second screen. Due to the normalization procedures strong decreases for abundant constructs can result in relative increases in read counts especially for constructs with
low initial read counts.
Overall, the decrease in read counts of miRZIP constructs was more pronounced in the first pilot experiment compared to the second set of experiments. This might have been caused by two main reasons. First of all, the last time point of harvesting cells was at day 27 in the first and at day 21 in the second experiment. A later time point likely would have resulted in a stronger decrease in read counts assuming a continuing decrease over time. Secondly, suboptimal experimental conditions in the pilot experiments as described above may have resulted in larger changes in construct abundances in these experiments.
In summary
Although the overall effects on proliferation are still limited, we do consider a high-throughput screen to identify miRNAs involved in regulating HL cell growth as feasible. Further improvement of the technical procedures should include longer follow-up and switch to another miRNA overexpression vector with a stronger and/or separate promotor for the GFP protein. Moreover, the number of miRNA overexpression and inhibition constructs in the pools should be increased to cover all potential relevant miRNAs in the screen.
Identification of miRNA target genes
MiRNAs are incorporated into Ago-protein containing RNA induced silencing complexes (RISC) and guide them to their target RNA transcripts [16]. In human, the Ago protein family consists of four members, i.e. Ago1, Ago2, Ago3 and Ago4. In general, both Ago1 and Ago2 are highly abundant, whereas the other Ago members are expressed at lower levels. To identify functional target genes relevant for HL, we performed an unbiased genome wide Ago2-RNA Immunoprecipitation (IP) approach in four HL cell lines in chapter 2.
Using a threshold of >2-fold enrichment in the IP fraction, we identified 1,142 genes (represented by 1,434 probes) consistently targeted by miRNAs in at least three out of four HL cell lines. Gene set enrichment analysis revealed a significant enrichment of the miRNA target gene sets of the top-10 most abundantly expressed miRNAs in HL. This confirms the efficiency of the IP procedure.
Based on the known oncogenic effects of miR-9 and the high levels of miR-155 in HL (Figure 1A) [5, 17], we aimed to pinpoint the targets of these two miRNAs in HL. Next
6
showed a significant decrease in abundance. None of the constructs showed a significant increase in abundance in at least 4 of 6 infections. For some of the in HL overexpressed or abundant miRNAs for which we included constructs in our miRNA inhibition screen, previous studies suggested a role in HL pathogenesis, i.e. miR-9; [5] miR-96, miR-182 and miR-183 cluster; [15] and the miR-17 seed family (miR-17 and miR-106b) [12]. In our cell lines, miR-9 inhibition did not result in a significant change in abundance in any of the 6 infections. The miR-96, miR-182 and miR-183 cluster was shown to target FOXO1, an important tumor suppressor gene, in HL [15]. However, no study was done to establish the effects of modulating expression of these three miRNAs on the growth of HL cell lines. For miR-182-5p inhibition, we observed no changes in any of the infections and for miR-96 and miR-183 we did not have inhibition constructs in our pool. So, for the latter two miRNAs it remains unknown whether they affect HL cell growth or not. Effects on cell cycle progression were shown in KM-H2 upon inhibition of the miR-17 seed family [12]. We observed a significant decrease in abundance for miR-20a-5p for 2 of 6 infections and for miR-106b-5p for 3 of 6 infections. For miR-17-5p no effects were observed and for miR-106a an increase in abundance was observed in 1 out of 6 infections. Thus, for some of the miR-17 family members we did observe a phenotype upon inhibition, albeit not consistent enough to pass our selection criteria.
ShRNA construct screen
The optimized library preparation procedure resulted in a much smaller difference in read counts between miRNA inhibition (median 328, range 15 – 927) and shRNA constructs (median 103, range 32 – 436) as compared to the pilot experiment described in chapter 3B. Nine shRNA constructs targeting 7 genes were depleted in at least 4 of 6 infections, i.e. BAP1-3, BAP1-4, EZH2-2, shRNA-EZH2-4, shRNA-FLJ-2, shRNA-HNRNPL-2, shRNA-MEF2C-1, shRNA-REL-1 and shRNA-SETD2-3. Again, none of the shRNA constructs showed enrichment over time. In comparison to the pilot experiment, shRNA constructs targeting Dicer or Drosha were not consistently increased in the duplicate infections. The apparent increase in abundance of shRNA constructs targeting Drosha and Dicer in the first experiment might have been caused by the low and thus less reliable read counts of these shRNA constructs. It can also be caused by the strong decrease of several miRZIP constructs in the pilot experiment within a limited total number of constructs as compared to the second screen. Due to the normalization procedures strong decreases for abundant constructs can result in relative increases in read counts especially for constructs with
low initial read counts.
Overall, the decrease in read counts of miRZIP constructs was more pronounced in the first pilot experiment compared to the second set of experiments. This might have been caused by two main reasons. First of all, the last time point of harvesting cells was at day 27 in the first and at day 21 in the second experiment. A later time point likely would have resulted in a stronger decrease in read counts assuming a continuing decrease over time. Secondly, suboptimal experimental conditions in the pilot experiments as described above may have resulted in larger changes in construct abundances in these experiments.
In summary
Although the overall effects on proliferation are still limited, we do consider a high-throughput screen to identify miRNAs involved in regulating HL cell growth as feasible. Further improvement of the technical procedures should include longer follow-up and switch to another miRNA overexpression vector with a stronger and/or separate promotor for the GFP protein. Moreover, the number of miRNA overexpression and inhibition constructs in the pools should be increased to cover all potential relevant miRNAs in the screen.
Identification of miRNA target genes
MiRNAs are incorporated into Ago-protein containing RNA induced silencing complexes (RISC) and guide them to their target RNA transcripts [16]. In human, the Ago protein family consists of four members, i.e. Ago1, Ago2, Ago3 and Ago4. In general, both Ago1 and Ago2 are highly abundant, whereas the other Ago members are expressed at lower levels. To identify functional target genes relevant for HL, we performed an unbiased genome wide Ago2-RNA Immunoprecipitation (IP) approach in four HL cell lines in chapter 2.
Using a threshold of >2-fold enrichment in the IP fraction, we identified 1,142 genes (represented by 1,434 probes) consistently targeted by miRNAs in at least three out of four HL cell lines. Gene set enrichment analysis revealed a significant enrichment of the miRNA target gene sets of the top-10 most abundantly expressed miRNAs in HL. This confirms the efficiency of the IP procedure.
Based on the known oncogenic effects of miR-9 and the high levels of miR-155 in HL (Figure 1A) [5, 17], we aimed to pinpoint the targets of these two miRNAs in HL. Next
to Ago2-IP based methods, another commonly used method to identify miRNA targets is stable isotope labeling by amino acids in cell culture (SILAC). The SILAC approach has been used successfully in several studies to identify miRNA specific target genes. Qian et al. found that the expression levels of 178 out of 1,498 proteins investigated were increased upon miR-21 inhibition. Using luciferase reporter assays they demonstrated that PIAS3 and PCBP1 are direct targets of miR-21 [18]. Christopher et al. identified 42 upregulated and 46 downregulated proteins upon ectopic overexpression of miR-155 in HEK-293T cells and validated CKAP5, KIF11, UBE2C and KPNA2 as miR-155 target genes by western blot analysis [19]. In addition, it has been reported miR-21 may contribute to increased viability and decreased apoptosis through targeting Bcl-2 [20] and PTEN [21] in DLBCL cells. SILAC experiments in MiaPaCa2 pancreatic cancer cells identified 93 more than 2-fold downregulated proteins upon miR-143 overexpression. Validation of 34 of these candidate targets by luciferase assays showed that 10 genes were direct targets of miR-143 [22].
Encouraged by these studies, we performed both SILAC and Ago2-IP experiments in 3 HL cell lines upon miR-9 or miR-155 inhibition (data not included in this thesis). Unfortunately, these two independent approaches resulted in a very limited overlap of miRNA specific target genes. The vast majority of the proteins identified as potential miRNA targets using the proteomics approach were not identified as miRNA specific targets in the Ago2-IP approach in the same cell line. The overlap between the 3 cell lines with the same approach was also quite limited, in contrast to the substantial overlap in targets for the Ago2-IP experiments performed on wild type cell lines. To unravel the cause of these disappointing results, we used the same constructs to inhibit these two miRNAs to demonstrate their potential effects on HL cell growth in GFP competition assays (Figure 1B). This revealed no effects upon miR-155 inhibition and mild effects upon miR-9 inhibition. The lack of a phenotype in combination with the failure to identify target genes for these miRNAs in HL might have been caused by (1) ineffective knockdown of these miRNAs possibly due to their high expression levels (Figure 1A), or (2) variation in cell line specific target genes, or (3) incomplete profile of the proteins detected in the proteomics analyses. In a previous study, we successfully identified 54 miR-155-specific target genes using Ago2-IP and confirmed 5 selected genes by luciferase assay in Burkitt lymphoma cell lines [23]. The main difference of our study with the previous study was that in the latter we overexpressed miR-155 in a miR-155 negative BL cell line. The limited effects that we saw upon inhibition of miR-155 in a miR-155 high HL cell line were in line with the limited overlap in target genes found in our study. Thus, to identify genes regulated by 9 or
miR-155 in HL we need a more robust miRNA knockdown in HL.
MiRNAs relevant for the pathogenesis of HL
We further investigated the role of two miRNAs on HL cell growth. In chapter 2 we showed that inhibition of miR-24-3p profoundly impaired HL cell growth, indicating an oncogenic role for 24-3p. To explore the mechanism, we tested the effect of miR-24-3p inhibition on cell death and cell cycle progression. We found that the percentages of apoptotic cells were significantly increased upon miR-24-3p inhibition, while no obvious effects on cell cycle were observed. To identify the HL-relevant targets of miR-24-3p we analyzed the Ago2-IP data. Functional annotation of the 52 Target scan predicted miR-24-3p target genes enriched in the Ago2-IP fraction revealed gene ontologies related to cell growth and apoptosis for 15 genes. The top-5 genes were selected for further validation by western blot. Two of the top-5 genes, i.e. CDKN1B/P27kip1 and MYC showed an increase in protein levels upon miR-24-3p
inhibition in HL cell lines. IHC showed variable expression of CDKN1B/P27kip1 and
MYC in HRS cells of HL cases without an obvious correlation between the expression patterns of these two proteins. In summary, miR-24-3p plays an oncogenic effect in HL and its inhibition impaired cell growth possibly via targeting CDKN1B/P27kip1 and MYC.
6
to Ago2-IP based methods, another commonly used method to identify miRNA targets is stable isotope labeling by amino acids in cell culture (SILAC). The SILAC approach has been used successfully in several studies to identify miRNA specific target genes. Qian et al. found that the expression levels of 178 out of 1,498 proteins investigated were increased upon miR-21 inhibition. Using luciferase reporter assays they demonstrated that PIAS3 and PCBP1 are direct targets of miR-21 [18]. Christopher et al. identified 42 upregulated and 46 downregulated proteins upon ectopic overexpression of miR-155 in HEK-293T cells and validated CKAP5, KIF11, UBE2C and KPNA2 as miR-155 target genes by western blot analysis [19]. In addition, it has been reported miR-21 may contribute to increased viability and decreased apoptosis through targeting Bcl-2 [20] and PTEN [21] in DLBCL cells. SILAC experiments in MiaPaCa2 pancreatic cancer cells identified 93 more than 2-fold downregulated proteins upon miR-143 overexpression. Validation of 34 of these candidate targets by luciferase assays showed that 10 genes were direct targets of miR-143 [22].
Encouraged by these studies, we performed both SILAC and Ago2-IP experiments in 3 HL cell lines upon miR-9 or miR-155 inhibition (data not included in this thesis). Unfortunately, these two independent approaches resulted in a very limited overlap of miRNA specific target genes. The vast majority of the proteins identified as potential miRNA targets using the proteomics approach were not identified as miRNA specific targets in the Ago2-IP approach in the same cell line. The overlap between the 3 cell lines with the same approach was also quite limited, in contrast to the substantial overlap in targets for the Ago2-IP experiments performed on wild type cell lines. To unravel the cause of these disappointing results, we used the same constructs to inhibit these two miRNAs to demonstrate their potential effects on HL cell growth in GFP competition assays (Figure 1B). This revealed no effects upon miR-155 inhibition and mild effects upon miR-9 inhibition. The lack of a phenotype in combination with the failure to identify target genes for these miRNAs in HL might have been caused by (1) ineffective knockdown of these miRNAs possibly due to their high expression levels (Figure 1A), or (2) variation in cell line specific target genes, or (3) incomplete profile of the proteins detected in the proteomics analyses. In a previous study, we successfully identified 54 miR-155-specific target genes using Ago2-IP and confirmed 5 selected genes by luciferase assay in Burkitt lymphoma cell lines [23]. The main difference of our study with the previous study was that in the latter we overexpressed miR-155 in a miR-155 negative BL cell line. The limited effects that we saw upon inhibition of miR-155 in a miR-155 high HL cell line were in line with the limited overlap in target genes found in our study. Thus, to identify genes regulated by 9 or
miR-155 in HL we need a more robust miRNA knockdown in HL.
MiRNAs relevant for the pathogenesis of HL
We further investigated the role of two miRNAs on HL cell growth. In chapter 2 we
showed that inhibition of miR-24-3p profoundly impaired HL cell growth, indicating an oncogenic role for 24-3p. To explore the mechanism, we tested the effect of miR-24-3p inhibition on cell death and cell cycle progression. We found that the percentages of apoptotic cells were significantly increased upon miR-24-3p inhibition, while no obvious effects on cell cycle were observed. To identify the HL-relevant targets of miR-24-3p we analyzed the Ago2-IP data. Functional annotation of the 52 Target scan predicted miR-24-3p target genes enriched in the Ago2-IP fraction revealed gene ontologies related to cell growth and apoptosis for 15 genes. The top-5 genes were selected for further validation by western blot. Two of the top-5 genes, i.e. CDKN1B/P27kip1 and MYC showed an increase in protein levels upon miR-24-3p
inhibition in HL cell lines. IHC showed variable expression of CDKN1B/P27kip1 and
MYC in HRS cells of HL cases without an obvious correlation between the expression patterns of these two proteins. In summary, miR-24-3p plays an oncogenic effect in HL and its inhibition impaired cell growth possibly via targeting CDKN1B/P27kip1 and MYC.
Figure 1. Endogenous miR-9 and miR-155 levels and effects on cell growth in HL cell lines. (A) RT-qPCR analysis results for miR-9 and miR-155 in HL cell lines and GC-B cells. Significance was analyzed by a Mann-Whitney test. *P<0.05. (B) GFP competition assay of miR-9 and miR-155 inhibitors (miRZIP-9 and miRZIP-155) and control miRZIP-SCR infected HL cell lines. The GFP percentage was measured triweekly for 22 days and the percentage at the first day of measurement (day 4) was set to 1.
The second miRNA selected for further functional follow-up was miR-21-5p. This miRNA was selected based on the high-throughput screening data described in chapter 5. Of the miRNAs with a phenotype on HL growth upon miRNA inhibition, miR-21-5p was the most abundant miRNA with significantly increased expression levels in HL cell lines compared to GC-B cells. We confirmed the significant decrease in cell
growth upon miR-21-5p inhibition in 3 HL cell lines. Functional studies revealed a significant increase in cell death upon miR-21-5p inhibition. Using the Ago2-IP data of chapter 2 with a threshold of probes showing an >2 fold enrichment in the IP/T ratio in at least 2 of 3 cell lines showing a phenotype, we identified 1,294 protein coding genes (represented by 1,664 probes) as being miRNA target genes. All 1,142 genes identified in chapter 2 are included in this list, with 152 additional genes being included because we used 2 out of 3 instead of 3 out of 4 HL cell lines as a criterion for selection. Of these 1,294 genes enriched in the Ago2-IP fractions, 36 genes were identified as putative targets of miR-21-5p according to prediction by Targetscan or previously proven miR-21-5p targets [24-28]. One of the identified genes, SESN1, was targeted by both miR-21-5p and miR-24-3p. Gene ontology analysis revealed that 13 of the miR-21-5p targets had a function related to cell growth and apoptosis. For 2 out of 13 targets, i.e. BTG2 and PELI1, the transcript abundance significantly decreased in HL cell lines as compared to GC-B cells. We confirmed targeting by miR-21-5p for both genes using luciferase assays and for PELI1 also at the protein level by western blotting.
Conclusions
In conclusion, in this thesis we determined the miRNA expression profile of HL cell lines and their normal counterparts, i.e. GC-B cells. We identified 84 significantly differentially expressed miRNAs. We setup and improved high throughput screens to identify miRNAs that can influence HL cell growth and identified several candidate miRNAs. Two miRNAs that both protect the HL cell lines from cell death were studied in more detail. MiR-24 protects HL cells from cell death possibly via targeting CDKN1B/P27kip1 and MYC and miR-21 protects HL cells from cell death via targeting
BTG2 and PELI1. An updated overview of the current knowledge on the role of miRNAs and their proven target genes involved in HL pathogenesis is shown in Figure 2.
6
Figure 1. Endogenous miR-9 and miR-155 levels and effects on cell growth in HL cell lines. (A)
RT-qPCR analysis results for miR-9 and miR-155 in HL cell lines and GC-B cells. Significance was analyzed by a Mann-Whitney test. *P<0.05. (B) GFP competition assay of miR-9 and miR-155 inhibitors (miRZIP-9 and miRZIP-155) and control miRZIP-SCR infected HL cell lines. The GFP percentage was measured triweekly for 22 days and the percentage at the first day of measurement (day 4) was set to 1.
The second miRNA selected for further functional follow-up was miR-21-5p. This miRNA was selected based on the high-throughput screening data described in
chapter 5. Of the miRNAs with a phenotype on HL growth upon miRNA inhibition,
miR-21-5p was the most abundant miRNA with significantly increased expression levels in HL cell lines compared to GC-B cells. We confirmed the significant decrease in cell
growth upon miR-21-5p inhibition in 3 HL cell lines. Functional studies revealed a significant increase in cell death upon miR-21-5p inhibition. Using the Ago2-IP data of chapter 2 with a threshold of probes showing an >2 fold enrichment in the IP/T ratio in at least 2 of 3 cell lines showing a phenotype, we identified 1,294 protein coding genes (represented by 1,664 probes) as being miRNA target genes. All 1,142 genes identified in chapter 2 are included in this list, with 152 additional genes being included because we used 2 out of 3 instead of 3 out of 4 HL cell lines as a criterion for selection. Of these 1,294 genes enriched in the Ago2-IP fractions, 36 genes were identified as putative targets of miR-21-5p according to prediction by Targetscan or previously proven miR-21-5p targets [24-28]. One of the identified genes, SESN1, was targeted by both miR-21-5p and miR-24-3p. Gene ontology analysis revealed that 13 of the miR-21-5p targets had a function related to cell growth and apoptosis. For 2 out of 13 targets, i.e. BTG2 and PELI1, the transcript abundance significantly decreased in HL cell lines as compared to GC-B cells. We confirmed targeting by miR-21-5p for both genes using luciferase assays and for PELI1 also at the protein level by western blotting.
Conclusions
In conclusion, in this thesis we determined the miRNA expression profile of HL cell lines and their normal counterparts, i.e. GC-B cells. We identified 84 significantly differentially expressed miRNAs. We setup and improved high throughput screens to identify miRNAs that can influence HL cell growth and identified several candidate miRNAs. Two miRNAs that both protect the HL cell lines from cell death were studied in more detail. MiR-24 protects HL cells from cell death possibly via targeting CDKN1B/P27kip1 and MYC and miR-21 protects HL cells from cell death via targeting
BTG2 and PELI1. An updated overview of the current knowledge on the role of miRNAs and their proven target genes involved in HL pathogenesis is shown in Figure 2.
Figure 2. An updated overview of miRNAs and target genes involved in HL pathogenesis. MiRNAs with increased expression in HRS cells include miR-9, miR-96, miR-155, miR-182 and miR-183, the miR-17 seed family, miR-24 and miR-21. Their increased expression in HL contributes to cytokines production, cell proliferation, cell cycle progression and inhibition of apoptosis via targeting specific mRNAs. Moreover, miR-135a is downregulated in HL, resulting in derepression of JAK2 and as a consequence resulting in increased Bcl-xL levels [5, 12-14, 29, 30]. Adapted from [31].
FUTURE PERSPECTIVES
Although the understanding of miRNA function in HL has increased in the past few years, the pathogenic consequences of altered miRNA expression remain unknown for the vast majority of the miRNAs. Therefore, we should further expand functional studies on miRNA in HL. To enhance the outcome of such studies it is important to further improve some of the experimental approaches used in this thesis.
High-throughput screens
We used high-throughput screens to identify miRNAs that influence HL cell growth. Although we optimized the screens and identified several miRNAs, the experimental conditions should be optimized further to more completely characterize the overall effects of miRNAs on HL cell growth. First of all, we could increase the library sizes to
include all potentially HL relevant miRNA constructs in the virus pool. Secondly, we should prolong the culturing time after infections from 21 days to at least 40 days to aid the identification of candidates with a more limited impact on cell growth. As the HL cell lines typically divide rather slowly it may take more time before changes in the abundance of constructs can be identified reliably. For the overexpression pool, we could try to increase the efficiency of infection of those miRNA overexpression constructs that had read counts below the minimal threshold, by reducing the insert sizes. In addition, it will be worthwhile to include more cell lines in the screen, as we know that the effects of miRNA modulation may vary per cell line. Finally, it might be worthwhile to change the experimental procedure of the screening to a single sort at day 5 and follow the abundance of the constructs in the sorted culture. This will reduce potential bias due to differences introduced by small variations in the applied sorting gates at different time points.
In our high-throughput miRNA loss-of-function studies, we failed to detect some miRNAs previously reported to influence the growth of HL i.e. miR-9 and the miR-17 family. This might have been caused by poor efficiency to inhibit the miRNAs by specific miRZIP constructs. A powerful new technology - clustered, regularly interspaced, short palindromic repeat (CRISPR) – might be applied to more effectively eliminate miRNA expression or function by deleting or mutating them at the genomic level [32]. Hong et al. demonstrated long-term miRNA knockdown phenotype by CRISPR/cas9 in in vitro and in vivo models [33]. Weixia et al. successfully mutated the endogenous miR-155 gene using CRISPR/Cas9 technology and obtained multiple knockout clones [34]. Using CRISPR/Cas9 technology, they performed a global loss-of-function screen to simultaneously test the functions of individual miRNAs and protein coding genes during the growth of a myeloid leukemia cell line. Using this screen, they identified miRNAs that suppressed or promoted cell growth [35]. Although a very promising new tool, this system can only be used to knockout miRNAs and not to induce miRNA expression. To enable inhibition and activation of transcription the catalytic site of Cas9 has been inactivated (dCas9) and the protein was coupled to either inhibiting (KRAB) or activating (VP64 and P300) domains [36]. More recently, dCas9 was also coupled to epigenome-modifying factors leading to new insights into the function of epigenetic marks in gene expression [37-39].
To improve experimental miRNA target gene identification in HL other promising new technologies should be considered, e.g. high-throughput sequencing of RNAs isolated by crosslinking immunoprecipitation (HITS-CLIP) [40], crosslinking ligation and sequencing of miRNA-RNA hybrids (CLASH) [41] and “chimiRic” [42]. Another
6
Figure 2. An updated overview of miRNAs and target genes involved in HL pathogenesis. MiRNAswith increased expression in HRS cells include miR-9, miR-96, miR-155, miR-182 and miR-183, the miR-17 seed family, miR-24 and miR-21. Their increased expression in HL contributes to cytokines production, cell proliferation, cell cycle progression and inhibition of apoptosis via targeting specific mRNAs. Moreover, miR-135a is downregulated in HL, resulting in derepression of JAK2 and as a consequence resulting in increased Bcl-xL levels [5, 12-14, 29, 30]. Adapted from [31].
FUTURE PERSPECTIVES
Although the understanding of miRNA function in HL has increased in the past few years, the pathogenic consequences of altered miRNA expression remain unknown for the vast majority of the miRNAs. Therefore, we should further expand functional studies on miRNA in HL. To enhance the outcome of such studies it is important to further improve some of the experimental approaches used in this thesis.
High-throughput screens
We used high-throughput screens to identify miRNAs that influence HL cell growth. Although we optimized the screens and identified several miRNAs, the experimental conditions should be optimized further to more completely characterize the overall effects of miRNAs on HL cell growth. First of all, we could increase the library sizes to
include all potentially HL relevant miRNA constructs in the virus pool. Secondly, we should prolong the culturing time after infections from 21 days to at least 40 days to aid the identification of candidates with a more limited impact on cell growth. As the HL cell lines typically divide rather slowly it may take more time before changes in the abundance of constructs can be identified reliably. For the overexpression pool, we could try to increase the efficiency of infection of those miRNA overexpression constructs that had read counts below the minimal threshold, by reducing the insert sizes. In addition, it will be worthwhile to include more cell lines in the screen, as we know that the effects of miRNA modulation may vary per cell line. Finally, it might be worthwhile to change the experimental procedure of the screening to a single sort at day 5 and follow the abundance of the constructs in the sorted culture. This will reduce potential bias due to differences introduced by small variations in the applied sorting gates at different time points.
In our high-throughput miRNA loss-of-function studies, we failed to detect some miRNAs previously reported to influence the growth of HL i.e. miR-9 and the miR-17 family. This might have been caused by poor efficiency to inhibit the miRNAs by specific miRZIP constructs. A powerful new technology - clustered, regularly interspaced, short palindromic repeat (CRISPR) – might be applied to more effectively eliminate miRNA expression or function by deleting or mutating them at the genomic level [32]. Hong et al. demonstrated long-term miRNA knockdown phenotype by CRISPR/cas9 in in vitro and in vivo models [33]. Weixia et al. successfully mutated the endogenous miR-155 gene using CRISPR/Cas9 technology and obtained multiple knockout clones [34]. Using CRISPR/Cas9 technology, they performed a global loss-of-function screen to simultaneously test the functions of individual miRNAs and protein coding genes during the growth of a myeloid leukemia cell line. Using this screen, they identified miRNAs that suppressed or promoted cell growth [35]. Although a very promising new tool, this system can only be used to knockout miRNAs and not to induce miRNA expression. To enable inhibition and activation of transcription the catalytic site of Cas9 has been inactivated (dCas9) and the protein was coupled to either inhibiting (KRAB) or activating (VP64 and P300) domains [36]. More recently, dCas9 was also coupled to epigenome-modifying factors leading to new insights into the function of epigenetic marks in gene expression [37-39].
To improve experimental miRNA target gene identification in HL other promising new technologies should be considered, e.g. high-throughput sequencing of RNAs isolated by crosslinking immunoprecipitation (HITS-CLIP) [40], crosslinking ligation and sequencing of miRNA-RNA hybrids (CLASH) [41] and “chimiRic” [42]. Another
powerful tool to pull down Ago protein complexes is the approach of “Ago protein Affinity Purification by Peptides” (APP). This method uses a high affinity Ago-binding peptide to pull down all Ago-containing complexes [43]. These technological improvements might improve the identification of miRNA target genes.
Functional miRNA studies
We investigated the role of miR-24 and miR-21 in HL in more detail. We identified several target genes of these miRNAs whose function fits with the observed role of miR-24 and miR-21 in HL, i.e. a pro-proliferative and/or anti-apoptotic function. For both miRNAs, we did not yet proof whether one or more of the target genes indeed has a causative role in the observed phenotype. This can be shown by performing a phenotype rescue experiment in which the target gene will be inhibited simultaneously with the miRNA.
Besides these 2 miRNAs, there are several other potentially interesting miRNAs that may be worthwhile to study in HL, including miR-449a-5p, miR-625-5p, let-7f-2-3p. Inhibition of these miRNAs resulted in a decrease in read counts of the corresponding constructs in at least 4 of 6 infections in the high throughput screening (chapter 5). However, the low expression levels for all these 3 miRNAs in HL cell lines (chapter 2), makes a pathogenic role for them in HL less likely.
So far, we only focused on miRNA target genes that are protein coding. However, long non-coding RNAs (lncRNAs) have also been recognized as miRNA interacting molecules [44]. For most of the miRNA-lncRNA interactions, it is currently unknown whether this will result in inhibition of the lncRNA, the miRNA or both. LncRNA transcripts that compete with mRNAs for miRNA binding have received much attention in recent years. These RNA transcripts are commonly known as competing endogenous RNAs (ceRNAs) or natural miRNA sponges [45]. This category includes lncRNAs, circular RNAs and pseudogenes [46]. The competition for miRNA binding can result in alleviation of miRNA-mediated repression of specific mRNA targets. MiR-9 targets the Metastasis Associated Lung Adenocarcinoma Transcript 1 (MALAT-1) in HL, which is among the most abundantly expressed and conserved lncRNAs [29]. In our Ago2-IP data, we focused on protein coding transcripts, but our array also contained probes for a large number of lncRNAs. Next to the 1,142 mRNAs described in this thesis, we also identified 133 lncRNAs consistently enriched in at least 3 out of the 4 studied HL cell lines. Future studies aiming to disrupt the interaction between the
lncRNA and miRNA could provide insight into the relevance of these interactions for the pathogenesis of HL.
The results of our study are based on in vitro experiments in HL cell lines. It is definitely important to also perform studies using patient tissues. Although it is challenging to obtain large number of purified HRS tumor cells, it is very likely that (small) RNA sequencing experiments in primary HRS cells will yield additional insights as compared to our analysis in cell lines. At this moment, miRNA in situ hybridization and immunohistochemistry for their target genes can be used to validate miRNA and target gene expression patterns in the tumor cells of primary HL biopsies and to establish mutually exclusive expression patterns. Unfortunately there is no HL animal model and HL tumor biopsies do not engraft in immunodeficient mouse models [47]. Alternatively, we could study the role of specific miRNAs in NHL animal models or use HL cell line xenograft models.
miRNAs in the clinic
MiRNAs play key regulatory roles in nearly every aspect of biology and, as we and others have shown, also in the pathogenesis of HL. So it is worthwhile to explore the use of miRNAs as biomarkers or potential therapeutic agents for HL. Many studies have reported the value of miRNAs as biomarkers with potential diagnostic and prognostic significance in various cancer types [48]. MiR-15a and miR-16-1 have been described as prognostic biomarkers in chronic lymphocytic leukemia (CLL) [49] and let-7a can act as a biomarker for lung cancer [50]. High levels of miR-21 have been associated with poor prognosis in many tumor types [51, 52]. Circulating cell-free miRNAs have received increasing attention for their potential diagnostic and prognostic value in diseases. Several miRNAs, such as miR-155, miR-210 and miR-21 were reported to have value as circulating biomarkers for diffuse large B-cell lymphoma [53]. In HL, clinical value of miRNAs is largely unknown. In a group of 89 cHL patients, low miR-135a expression levels were associated with a higher probability of relapse (P=0.04) and a shorter disease free survival (P=0.02), indicating prognostic value of this miRNA. [13] Van Eijndhoven et al. determined levels of miRNAs in plasma derived extracellular vesicles (EV) in cHL patients compared to healthy donors. This revealed increased levels of miR-21-5p, miR-127-3p, miR-24-3p, let-7a-5p, and miR-155-5p in cHL. The authors proposed that miRNA levels in vesicles might be a good biomarker for metabolic active tumor tissue reflecting the presence of vital tumor cells [54]. So, these miRNAs might potentially be suitable for monitoring therapy response and
6
powerful tool to pull down Ago protein complexes is the approach of “Ago protein Affinity Purification by Peptides” (APP). This method uses a high affinity Ago-binding peptide to pull down all Ago-containing complexes [43]. These technological improvements might improve the identification of miRNA target genes.
Functional miRNA studies
We investigated the role of miR-24 and miR-21 in HL in more detail. We identified several target genes of these miRNAs whose function fits with the observed role of miR-24 and miR-21 in HL, i.e. a pro-proliferative and/or anti-apoptotic function. For both miRNAs, we did not yet proof whether one or more of the target genes indeed has a causative role in the observed phenotype. This can be shown by performing a phenotype rescue experiment in which the target gene will be inhibited simultaneously with the miRNA.
Besides these 2 miRNAs, there are several other potentially interesting miRNAs that may be worthwhile to study in HL, including miR-449a-5p, miR-625-5p, let-7f-2-3p. Inhibition of these miRNAs resulted in a decrease in read counts of the corresponding constructs in at least 4 of 6 infections in the high throughput screening (chapter 5). However, the low expression levels for all these 3 miRNAs in HL cell lines (chapter 2), makes a pathogenic role for them in HL less likely.
So far, we only focused on miRNA target genes that are protein coding. However, long non-coding RNAs (lncRNAs) have also been recognized as miRNA interacting molecules [44]. For most of the miRNA-lncRNA interactions, it is currently unknown whether this will result in inhibition of the lncRNA, the miRNA or both. LncRNA transcripts that compete with mRNAs for miRNA binding have received much attention in recent years. These RNA transcripts are commonly known as competing endogenous RNAs (ceRNAs) or natural miRNA sponges [45]. This category includes lncRNAs, circular RNAs and pseudogenes [46]. The competition for miRNA binding can result in alleviation of miRNA-mediated repression of specific mRNA targets. MiR-9 targets the Metastasis Associated Lung Adenocarcinoma Transcript 1 (MALAT-1) in HL, which is among the most abundantly expressed and conserved lncRNAs [29]. In our Ago2-IP data, we focused on protein coding transcripts, but our array also contained probes for a large number of lncRNAs. Next to the 1,142 mRNAs described in this thesis, we also identified 133 lncRNAs consistently enriched in at least 3 out of the 4 studied HL cell lines. Future studies aiming to disrupt the interaction between the
lncRNA and miRNA could provide insight into the relevance of these interactions for the pathogenesis of HL.
The results of our study are based on in vitro experiments in HL cell lines. It is definitely important to also perform studies using patient tissues. Although it is challenging to obtain large number of purified HRS tumor cells, it is very likely that (small) RNA sequencing experiments in primary HRS cells will yield additional insights as compared to our analysis in cell lines. At this moment, miRNA in situ hybridization and immunohistochemistry for their target genes can be used to validate miRNA and target gene expression patterns in the tumor cells of primary HL biopsies and to establish mutually exclusive expression patterns. Unfortunately there is no HL animal model and HL tumor biopsies do not engraft in immunodeficient mouse models [47]. Alternatively, we could study the role of specific miRNAs in NHL animal models or use HL cell line xenograft models.
miRNAs in the clinic
MiRNAs play key regulatory roles in nearly every aspect of biology and, as we and others have shown, also in the pathogenesis of HL. So it is worthwhile to explore the use of miRNAs as biomarkers or potential therapeutic agents for HL. Many studies have reported the value of miRNAs as biomarkers with potential diagnostic and prognostic significance in various cancer types [48]. MiR-15a and miR-16-1 have been described as prognostic biomarkers in chronic lymphocytic leukemia (CLL) [49] and let-7a can act as a biomarker for lung cancer [50]. High levels of miR-21 have been associated with poor prognosis in many tumor types [51, 52]. Circulating cell-free miRNAs have received increasing attention for their potential diagnostic and prognostic value in diseases. Several miRNAs, such as miR-155, miR-210 and miR-21 were reported to have value as circulating biomarkers for diffuse large B-cell lymphoma [53]. In HL, clinical value of miRNAs is largely unknown. In a group of 89 cHL patients, low miR-135a expression levels were associated with a higher probability of relapse (P=0.04) and a shorter disease free survival (P=0.02), indicating prognostic value of this miRNA. [13] Van Eijndhoven et al. determined levels of miRNAs in plasma derived extracellular vesicles (EV) in cHL patients compared to healthy donors. This revealed increased levels of miR-21-5p, miR-127-3p, miR-24-3p, let-7a-5p, and miR-155-5p in cHL. The authors proposed that miRNA levels in vesicles might be a good biomarker for metabolic active tumor tissue reflecting the presence of vital tumor cells [54]. So, these miRNAs might potentially be suitable for monitoring therapy response and
identify relapse in individual HL patients.
Targeting miRNAs has entered the preclinical and clinical stage for a few of them and might become available for therapy of HL patients in the future [55]. For example, miR-122 is the most abundant miRNA in adult liver and a central player in liver biology and disease. Therefore, it is considered to be a therapeutic target in liver disease [56, 57]. MiR-155 is one of the most commonly overexpressed miRNAs in Acute Myeloid Leukemia (AML). Inhibition of miR-155 induced significant anti-leukemic effects in AML cell lines and primary AML samples in vitro and in vivo, which underscores the potential of miR-155 as a therapeutic target in AML [58]. MRG-106, an inhibitor of miR-155, is currently being tested in clinical trials in cutaneous T-cell lymphoma patients. It would be interesting to test the efficiency of this inhibitor to completely inhibit the high miR-155 levels in HL cells and explore its potential relevance to treat HL patients. RG-012, a potent inhibitor of miR-21, has received an orphan drug status from the U.S. Food and Drug Administration and European Commission as a potential therapeutic option for Alport syndrome in preclinical studies. The marked effects we observed on HL cell growth upon miR-21 inhibition, is promising for the possible implementation of RG-012 as a novel treatment option for HL patients. Based on these reports and our own data it is evident that miRNAs are attractive biomarkers and therapeutic targets for various diseases, including HL.
REFERENCES
1. He, L., et al., A microRNA polycistron as a potential human oncogene. Nature, 2005. 435(7043): p. 828-33.
2. Lawrie, C.H., MicroRNAs and lymphomagenesis: a functional review. Br J Haematol, 2013. 160(5): p. 571-81.
3. Mraz, M., et al., MicroRNAs in chronic lymphocytic leukemia pathogenesis and disease subtypes. Leuk Lymphoma, 2009. 50(3): p. 506-9.
4. Huang, X., et al., CD99 triggers upregulation of miR-9-modulated PRDM1/BLIMP1 in Hodgkin/Reed-Sternberg cells and induces redifferentiation. Int J Cancer, 2012. 131(4): p. E382-94.
5. Leucci, E., et al., Inhibition of miR-9 de-represses HuR and DICER1 and impairs Hodgkin lymphoma tumour outgrowth in vivo. Oncogene, 2012. 31(49): p. 5081-9.
6. Gibcus, J.H., et al., Hodgkin lymphoma cell lines are characterized by a specific miRNA expression profile. Neoplasia, 2009. 11(2): p. 167-76.
7. Landgraf, P., et al., A mammalian microRNA expression atlas based on small RNA library sequencing. Cell, 2007. 129(7): p. 1401-14.
8. Navarro, A., et al., MicroRNA expression profiling in classic Hodgkin lymphoma. Blood, 2008. 111(5): p. 2825-32.
9. Van Vlierberghe, P., et al., Comparison of miRNA profiles of microdissected Hodgkin/Reed-Sternberg cells and Hodgkin cell lines versus CD77+ B-cells reveals a distinct subset of differentially expressed miRNAs. Br J Haematol, 2009. 147(5): p. 686-90.
10. Dorsett, Y., et al., MicroRNA-155 suppresses activation-induced cytidine deaminase-mediated Myc-Igh translocation. Immunity, 2008. 28(5): p. 630-8.
11. Kluiver, J., et al., Lack of BIC and microRNA miR-155 expression in primary cases of Burkitt lymphoma. Genes Chromosomes Cancer, 2006. 45(2): p. 147-53.
12. Gibcus, J.H., et al., MiR-17/106b seed family regulates p21 in Hodgkin's lymphoma. J Pathol, 2011. 225(4): p. 609-17.
13. Navarro, A., et al., Regulation of JAK2 by miR-135a: prognostic impact in classic Hodgkin lymphoma. Blood, 2009. 114(14): p. 2945-51.
14. Yuan, Y., et al., miR-24-3p Is Overexpressed in Hodgkin Lymphoma and Protects Hodgkin and Reed-Sternberg Cells from Apoptosis. Am J Pathol, 2017. 187(6): p. 1343-1355. 15. Xie, L., et al., FOXO1 is a tumor suppressor in classical Hodgkin lymphoma. Blood, 2012.
119(15): p. 3503-11.
16. Ender, C. and G. Meister, Argonaute proteins at a glance. J Cell Sci, 2010. 123(Pt 11): p. 1819-23.
17. Kluiver, J., et al., BIC and miR-155 are highly expressed in Hodgkin, primary mediastinal and diffuse large B cell lymphomas. J Pathol, 2005. 207(2): p. 243-9.
18. Xiong, Q., et al., Identification of novel miR-21 target proteins in multiple myeloma cells by quantitative proteomics. J Proteome Res, 2012. 11(4): p. 2078-90.
19. Lossner, C., et al., Quantitative proteomics identify novel miR-155 target proteins. PLoS One, 2011. 6(7): p. e22146.
20. Liu, K., J. Du, and L. Ruan, MicroRNA-21 regulates the viability and apoptosis of diffuse large B-cell lymphoma cells by upregulating B cell lymphoma-2. Exp Ther Med, 2017. 14(5):
