A parthenogenetic quasi-program causes teratoma-like tumors during aging in wild-type C. elegans

A long-standing belief is that aging (senescence) is the result of stochastic damage accumulation. Alternatively, senescent pathology may also result from late-life, wild-type gene action (i.e., antagonistic pleiotropy, as argued by Williams) leading to non-adaptive run-on of developmental programs (or quasi-programs) (as suggested more recently by Blagosklonny). In this study, we use existing and new data to show how uterine tumors, a prominent form of senescent pathology in the nematode Caenorhabditis elegans, likely result from quasi-programs. Such tumors develop from unfertilized oocytes which enter the uterus and become hypertrophic and replete with endoreduplicated chromatin masses. Tumor formation begins with ovulation of unfertilized oocytes immediately after exhaustion of sperm stocks. We show that the timing of this transition between program and quasi-program (i.e., the onset of senescence), and the onset of tumor formation, depends upon the timing of sperm depletion. We identify homology between uterine tumors and mammalian ovarian teratomas, which both develop from oocytes that fail to mature after meiosis I. In teratomas, futile activation of developmental programs leads to the formation of differentiated structures within the tumor. We report that older uterine tumors express markers of later embryogenesis, consistent with teratoma-like activation of developmental programs. We also present evidence of coupling of distal gonad atrophy to oocyte hypertrophy. This study shows how the Williams Blagosklonny model can provide a mechanistic explanation of this component of C. elegans aging. It also suggests etiological similarity between teratoma and some forms of senescent pathology, insofar as both are caused by quasi-programs.

Quantitation of chromatin mass development in senescent uterine tumors. Figure S2 Expression of markers of development in older uterine tumors. Figure S3 Timing of sperm depletion affects timing of tumorigenesis. Figure S4 Effect of RNAi on tumor size. Figure S5 VIT-2::GFP accumulation in uterine tumors. Figure S6 Relationship between other pathologies and uterine tumors. Figure S7 Is the uterus a latent tumor niche? Figure S8 Older vs newer ideas about the proximate action of antagonistic pleiotropy. Figure S9 Red autofluorescence in senescent uterine tumors. Video S1 3D view of nuclei within uterine tumors (light sheet microscopy [SPIM]). Video S2 3D view of individual nuclei within tumors (light sheet microscopy [SPIM]). Video S3 3D view of VIT-2::GFP accumulation in uterine tumor (confocal microscopy).

lin-53 is not a suppressor of neuronal gene expression in uterine tumors
The status of germline chromatin prevents expression of somatic genes, and loss of the LIN-53 histone chaperone can promote neuronal differentiation within germline 1 . To probe the role of lin-53 as a teratoma suppressor in uterine tumors, we tested effects of lin-53 RNAi on tumor expression of an embryonic neuronal marker (punc-119::GFP). However, no effect of lin-53 RNAi was detected ( Figure S2f).

No correlation between pharyngeal and uterine infection
Under standard culture conditions some 40% of wild-type N2 hermaphrodites die as the result of pharyngeal infection with E. coli 2 . One possibility is that such worms have weaker organism-wide immunity which also leads to uterine tumor infection. To test this we examined worms on days 10 and 14, when major pharyngeal infections are relatively frequent, to see whether those with infected pharynxes had a higher frequency of infected tumors, but they did not ( Figure S6d). Thus, independent mechanisms lead to bacterial infection of the pharynx and uterus.

Evidence of possible latent tumor niche properties in the uterus
That extra-uterine oocytes do not develop into tumors raises the possibility that the uterus might possess tumor niche properties that become manifest only after sperm depletion, when unfertilized, immature oocytes start to enter the uterus 3 . Proliferation of mitotic germ cells is normally promoted by Notch signaling from the distal tip cell 4 . However, mutations such as pro-1(na48) cause abnormal juxtaposition of mitotic germ cells with the gonad sheath, which stimulates mitosis and consequently the development of proximal tumors (the Pro phenotype) 5 . This has been attributed to production of Notch ligands by the gonad sheath, particularly APX-1, ARG-1 and DSL-5, since RNAi or mutation of these genes suppresses pro-1(na48) Pro, and apx-1 and arg-1 are expressed in the gonad sheath 3 .
One possibility is that dsl-5(ok588) delays tumor development by reducing oocyte production rate and/or delaying sperm depletion. To test this we examined dsl-5 mutant reproductive parameters, but detected no delay in the reproductive schedule or, after sperm depletion, reduction of unfertilized oocytes laid ( Figure S7d), or level of stacking of oocytes in the proximal gonad 6 (data not shown). Thus, the delay in tumor growth in dsl-5 mutants appears to be a consequence of slower tumor growth rather than delayed onset of tumor development.
A further possibility is that the effect of dsl-5 on tumor growth is due to an overall delay in the aging process. To probe this we examined the effect of dsl-5 on other senescent pathologies: pharyngeal deterioration, distal gonad atrophy, intestinal atrophy and yolk accumulation, as previously described 7,8 . The only effect detected was a slight increase in yolk accumulation (data not shown), possibly attributable to reduced yolk uptake by tumors.
APX-1 and ARG-1 are predicted to be membrane tethered, but DSL-5 is predicted to be secreted 9 . This suggests the possibility that DSL-5 is secreted into the uterine lumen. However, the site(s) of dsl-5 expression remain unidentified: in a prior analysis of 10 lines with pdsl-5::GFP reporters including ~3.5 kb upstream of dsl-5 no expression was detected 3 . This could mean that distant regulatory elements control dsl-5 expression. Notably, dsl-5 is the most downstream of six closely spaced genes transcribed in the same orientation, the first two of which are part of a predicted operon. This suggests the possibility that dsl-5 is downstream in a large operon, in which case thus the dsl-5 promoter would lie far upstream of its open reading frame (ORF). To try to identify the site of dsl-5 expression we used a counter-selection recombineering protocol 10 to replace the dsl-5 ORF with GFP in the fosmid WRM0639dC09, which contains the entire six gene cluster. However, although multiple C. elegans transformants were generated containing the recombinant fosmid (4 germline, 5 transient, confirmed by PCR), no GFP was detected (data not shown). A remaining possibility is that dsl-5 expression requires sequence elements that are downstream of the gene's translational start site, e.g. in intronic sequences.
Next we tested the effect on tumor growth of adult-specific abrogation of expression of the germline-expressed Notch receptor GLP-1, which also plays a major role in embryogenesis 11 . For this we used the temperature sensitive glp-1(e2141) mutant, which is virtually wild-type at 15˚C 12 . glp-1 hermaphrodites were shifted to the non-permissive temperature at several time points, including after self-sperm depletion (L4, day 2, 4, 6, 8), and tested for reduction in final tumor size (day 10), in comparison to similarly temperature-shifted N2 controls. Reduced tumor size was seen only upon upshift prior to sperm depletion (L4, day 2) ( Figure S7e). This suggests that synthesis of GLP-1 in intra-uterine oocytes after sperm depletion does not promote tumor growth; however, it does not rule out a role for GLP-1 synthesized prior to the temperature shift. The possibility that a uterine tumor niche promotes oocyte hypertrophy warrants further investigation.

Age increase in red autofluorescence is attributable to uterine tumor growth
During epifluorescence analysis we noted the presence of red autofluorescence within the tumors of older wild-type hermaphrodites that was not present in earlier stage tumors ( Figure S9a,b). Such red autofluorescence occurred in small patches at variable positions within the tumor. Notably, red autofluorescence is a biomarker of senescence in C. elegans 13 . This biomarker appears to be largely attributable to tumor patho-development since overall red fluorescence increased with age more strongly in rrf-3(b26) worms (tumors present) than glp-4(bn2) worms (tumors absent) ( Figure S9c), and mainly occurred in the mid-body ( Figure S9d).

Construction of dsl-5 transgenic lines
To construct dsl-5 reporter lines including including potential distant gene-regulatory elements, we used the fosmid WRM0639dC09, which includes a cluster of six predicted genes, of which dsl-5 is the most downstream. The open reading frame of the dsl-5 gene was substituted with that of GFP using a counter selection recombineering protocol, as previously described 10 . Transgenic lines were generated by either microinjection or biolistic transformation, using rol-6(su1006) as a cotransformational marker. The presence of the fosmid in Rol transformants was confirmed by PCR.
Mating protocol N2 or fog-2(q71) males were used for mating. A brief mating protocol was used to reduce the lifeshortening effects of mating 22 . Sexes were separated at L4 stage and left overnight to develop into adults before being combined for mating. Animals were incubated at a ratio of 3:1 males to hermaphrodites/females for 5 hours (84 males + 28 hermaphrodites/females per plate), after which males were removed.
Progeny and unfertilized oocytes production assays L4 larvae were cultured at a population density of one animal per plate, and transferred daily to fresh plates during the first five days of adulthood. After removal of the parent worm, each plate was then maintained at 20˚C for 1 day, and then larvae and unfertilized oocytes on the plate were counted using a dissecting microscope.

Microscopy
Nematodes were placed on 2% agarose pads and anesthetized with 0.2% levamisole. For most trials, Nomarski and epifluorescence microscopy was performed on a Zeiss Axioskop 2 Plus microscope connected to a Hamamatsu C10600 -Orca ER digital camera. Images were acquired and quantified using Volocity 6.3 software. 100x or 400x images are shown in this report. For viewing red autofluorescence, a rhodamine filter was used (λ ex 545/25 nm, λ em 605/70 nm).
Selective plane illumination microscopy (SPIM) was used to make 3D reconstructions of C. elegans uterine tumors. Anaesthetized worms were transferred into a plastic capillary tube with 1.5% low melting point agarose, 0.03% levamisole and 0.5 µm sized FluoSphere microspheres (beads) (1:1000, F8813 from Life Technologies). Agarose embedding and anesthetic kept the animals immobile during imaging, and FluoSphere beads allowed registration and 3D construction of 2D images taken from 5 angles. We used a 1 ml BD Plastikpak (REF 300013) syringe to mount the plastic capillaries into the OpenSPIM chamber full of M9 buffer. OpenSPIM was performed to take images as previously described 23 . Acquired data was processed using Fiji software. The beads registration algorithm and the multi-view deconvolution plugin were performed for reconstruction of 3D structure 24,25 .

DAPI staining and quantitation of genomic copies
Nuclei of uterine tumors were stained with the DNA-binding dye 4', 6'-diamidino-2-phenylindole (DAPI). Worms of different ages were fixed with methanol and incubated on ice, then washed with M9 buffer and stained with 500 ng/µl DAPI staining in darkness for 30 min. Finally, worms were washed again with M9 buffer before imaging. Genomic copy number was estimated using real-time PCR as described 26 .

Bodipy staining
Lipid staining was performed in fixed worms as described 27 , except that animals were manipulated in 15μL droplets within parafilm micro-wells. Worms were washed 2x by transferring them successively into two droplets of M9, and then transferred to a drop of 2% paraformaldehyde solution for 15-20min and frozen/thawed 3x at -80˚C/room temperature (RT). They were then carefully washed 3x by transferring them into fresh M9 droplets. Worms were then transferred to 1μg/mL BODIPY 493/503 (Invitrogen) in M9 for 1-2hr at RT in darkness. Lastly, they were washed 3x in M9 droplets and mounted for imaging (λ ex 488 nm, λ em 505-575 nm) using a Zeiss LSM710 confocal microscope.

Pathology scoring system
Severity of uterine tumors was scored using a five stage classification as previously described 28 . Score 1 denotes a uterus containing fertilized eggs or oocytes of normal size and morphology. Score 2 denotes a uterus containing unfertilized oocytes only, with a slightly abnormal appearance. Score 3 and score 4 denote a small tumor and a large tumor, respectively. Score 5 denotes a very large tumor which fills the body cavity in mid-body region.
To measure different levels of nuclear hypertrophy we created another five stage classification. Score 1 denotes small, spherical early stage oocyte nuclei. Score 2 denotes larger but still spherical nuclei. Score 3 denotes larger nuclei with irregular morphology. Score 4 denotes highly hypertrophic nuclei with grossly abnormal phenotype (e.g. with major protrusions), such that that some individual nuclei are barely distinguished. Score 5 denotes large chromatin masses where most individual nuclei can no longer be distinguished.
To describe expression of embryonic reporters at different levels we used a three stage classification system. Score 1 denotes no reporter fluorescence in the tumor. Score 2 denotes weak reporter fluorescence in the tumor. Score 3 denotes strong reporter fluorescence in the tumor.

RNA-mediated interference
The E. coli feeding strains used for RNAi were obtained from the Ahringer RNAi strain library. DNA sequencing was used to verify the identity of all plasmid inserts. RNAi treatment was performed as described previously 29 . Worms were raised on E. coli HT115 and transferred to plates seeded with RNAi feeding strains at the L4 stage, or on day 1, 3 or 6 of adulthood. E. coli HT115 containing the empty plasmid vector L4440 was used as a control.

Statistical analysis
Correlation analysis was performed using linear regression analysis. The non-parametric Wilcoxon-Mann Whitney test was performed to compare uterine status and nuclear morphology. Fluorescence intensity was measured using Volocity and Fiji software. A multiple comparisons t test was used to compare fluorescence intensity and tumor size.        Blagosklonny suggests that genes with major, global effects on aging control entire developmental and reproductive programs, and promote their futile run-on into non-adaptive quasi-programs, with pathogenic consequences. For example, as described in this study, complex developmental programs of oocyte maturation and embryogenesis, requiring action of numerous wild-type genes, run on to become quasi-programs causing uterine tumor formation. Mutation of the growth-promoting DAF-2 insulin/IGF-1 receptor inhibits uterine tumor development. Different colors of tissues, organs indicate distinct states of differentiation; e.g. in C. elegans, grey could be the germline, blue the intestine, and so on. TOR, target of rapamycin kinase; IGF-1R, insulin-like growth factor 1 receptor. Pale green and red indicate promotion of fitness and senescent pathology, respectively.   Candidate tumor promoter genes, including genes specifying germline development and the cell cycle, were selected from those listed in several sources [33][34][35] . Summed data for three trials, all trials, n≥10. Linear regression analysis. Red font, p < 0.05.