TTP family member transcripts in cultured WT BMDM before and after LPS stimulation

TTP (or ZFP36), ZFP36L1, and ZFP36L2 are paralogues that all emerged during early vertebrate evolution, and are products of different genes on different chromosomes (Taylor et al, 1991; Blackshear & Perera, 2014). In addition, as described above, knocking out the individual genes in mice led to very different phenotypes (Taylor et al, 1996; Stumpo et al, 2004, 2009). Nonetheless, they all have highly conserved tandem zinc finger and CNOT1 binding domains, and exhibit similar biochemical activities in RNA binding, deadenylation, and decay assays (Lai et al, 2000, 2003). This has raised questions as to whether the three proteins could be interchangeable or have redundant functions in certain circumstances. TTP is best known for its effects on pro-inflammatory cytokines and chemokines in myeloid cells (Carballo et al, 1998; Kratochvill et al, 2011; Qiu et al, 2012; Lee et al, 2017), but the potential functions and targets of ZFP36L1 and ZFP36L2 in myeloid cells are largely unknown.

In the experiments described here, basal levels of Zfp36, Zfp36l1, and Zfp36l2 mRNAs were roughly equivalent in control mouse macrophages before LPS stimulation. However, Zfp36 mRNA rapidly increased after 1 h of LPS stimulation and remained elevated even 24 h after the addition of LPS, whereas Zfp36l1 and Zfp36l2 mRNA levels slightly increased after LPS treatment and then rapidly decreased to below basal levels. These data, combined with area under the curve analyses, suggest that Zfp36 mRNA is expressed to by far the greatest extent of the TTP family mRNAs after a pro-inflammatory stimulus in macrophages (Qiu et al, 2012). Previous experiments have shown that these changes in Zfp36 mRNA levels are initially reflected in rapid and large changes in TTP protein accumulation, and it remains at high levels after LPS for many hours (Qiu et al, 2012, 2015; Patial et al, 2016a; Lai et al, 2018). Less is known about the behavior of the other two proteins under these conditions.

Genome-wide effects of M-TTP KO and M-triple KO genotypes on transcript levels in unstimulated BMDM

In unstimulated BMDM from the M-TTP KO mice, TTP deficiency alone led to significant accumulation of only two transcripts by more than 1.3fold, of which only one (Asb1) contained one or more potential TTP binding sites (see Fig 8A and Table S1). A handful of transcripts were down-regulated under these conditions, presumed to be because of secondary or otherwise downstream events.

In contrast, measurement of steady-state transcript levels in the unstimulated M-triple KO macrophages revealed that 1,415 transcripts were significantly up-regulated by 1.3fold or more, including Asb1, and 1,223 transcripts were significantly down-regulated by 1.3fold or more, when compared with control macrophages (see Fig 8C and Table S3). There was significant enrichment of potential TTP binding site sequences in the up-regulated transcripts compared with the down-regulated transcripts, and many of the up-regulated transcripts were previously known or suspected targets of TTP (Brooks & Blackshear, 2013; Patial et al, 2016b). We presume that there are many other direct targets of TTP and its paralogues in this collection, and that the rest of the up-regulated, and probably all of the down-regulated, transcripts are the results of secondary or otherwise downstream events.

Although much previous work has demonstrated that the presence of at least one TTP binding site is necessary for TTP to bind to and destabilize its target transcripts (Carballo et al, 1998; Lai et al, 1999; Lai et al, 2000; Lai and Blackshear, 2001; Lai et al, 2003; Brewer et al, 2004; Lai et al, 2005; Qiu et al, 2015; Lai et al, 2018), many apparently unaffected or down-regulated transcripts in the present experiments contain potential TTP family member binding sites within their 3′-UTRs, but did not appear to be stabilized. This phenomenon has been observed in experiments of this type in many different cell types, and there is no simple explanation (Patial et al, 2016b). One possibility is that the potential binding sites could be occupied by other AU-rich element-binding proteins or even other RNAs (Sobolewski et al, 2022). Another possibility is that there could be an RNA secondary structure involving the potential binding site, which has been shown to prevent binding by TTP family proteins (Hudson et al, 2004). TTP and its family members also may occupy specific locations within the cytosol that may not expose them to the relevant mRNAs (Phillips et al, 2002). There may be multiple mechanisms at work, but the general consensus from many experiments is that the number of transcripts with potential binding sites is much greater than those demonstrated to decay in response to TTP-like proteins.

What is the mechanism of this remarkable difference in transcript level changes between the unstimulated cells from the M-TTP KO and M-triple KO mice? The most likely mechanism is that the three TTP paralogues, present at similar levels, act on the same biochemical pathways in the WT cells. In this proposed mechanism, the proteins would bind to the same AU-rich motifs in the same transcripts, and promote their deadenylation and decay, in a synergistic fashion. One corollary of this idea is that the three proteins are behaving essentially identically as “TTP equivalents.” This will be difficult to prove, but it is supported by the similar effects exhibited by the proteins in cell transfection mRNA decay assays, and cell-free assays of RNA binding and promotion of deadenylation (Lai et al, 2000, 2003). On the other hand, there have been several reports of, for example, TTP-specific activities based on primary amino acid sequences that are unique to TTP (Lykke-Andersen & Wagner, 2005; Bulbrook et al, 2018). Ongoing studies are underway to test the “equivalence” concept, including expressing comparable levels of the three proteins in cells, and determining whether the induced changes in gene expression are identical or different. We are also extending the present studies to investigate the effects of knocking out two of the three family member genes on gene expression patterns under the same conditions.

It should be noted that the macrophages used in our experiments were incubated in low-serum medium for 16–18 h before harvesting, so some changes in actual target transcripts, and at least some of the presumed downstream effects, may be in part because of autocrine/paracrine effects of secreted cytokines such as TNF, whose transcript is a well-known target of TTP in these cells.

Phenotypes of M-triple KO mice compared with control mice, and single myeloid-specific KO mice, under normal vivarium conditions

As described previously for M-TTP KO (Kratochvill et al, 2011; Qiu et al, 2012) and M-ZFP36L1 KO (Hyatt et al, 2014) mice, the single myeloid-specific KO mice for each of the three genes tested in this study did not exhibit any weight loss, arthritis, premature death or any other external phenotypes in the first several months of age. In marked contrast, the M-triple KO mice had a severe, spontaneous inflammatory phenotype, with marked peripheral arthritis, failure of weight gain, and early death, with median survivals of only about 8 wk for both sexes. This was characterized by increased levels of many pro-inflammatory cytokines in the serum, and histological evidence of widespread, severe soft tissue infiltration by inflammatory cells, bone destruction and osteopenia, splenomegaly, and many other pathological changes. Overall, the arthritis in both the whole-body Zfp36-KO mice and the M-triple KO mice is similar to the pathology observed in human rheumatoid arthritis (Komatsu & Takayanagi, 2022) and in mouse models of rheumatoid arthritis, such as collagen antibody-induced arthritis (Caplazi et al, 2015).

As detailed above in the Results section, several cytokines and chemokines were elevated in the the serum of the M-triple KO mice, and many of these were also found in the medium of their cultured BMDM. Many of these are undoubtedly involved in the pathogenesis of the severe M-triple KO syndrome, and their involvement in inflammatory diseases and disease models has been documented in many previous studies (Deshmane et al, 2009; Lee et al, 2017; Karin & Razon, 2018; Steinkamp et al, 2018; Tokunaga et al, 2018; House et al, 2020; Yao et al, 2021). On the other hand, some well-known pro-inflammatory cytokines, such as G-CSF and IL-6, were increased in the serum of M-triple KO mice, but not in serum from M-TTP KO mice. These have also been implicated in previous models of inflammatory disease, and in myeloid lineage specification (Lawlor et al, 2004; Roberts, 2005; Kaplan et al, 2011; Theyab et al, 2021).

It seems probable that the pathogenetic mechanism for this phenotype is a consequence of the gene expression changes discussed above in unstimulated macrophages. Specifically, even under normal vivarium conditions, there is likely to be increased stabilization of many transcripts encoding pro-inflammatory cytokines in myeloid cells; this is reflected in the increases we saw in serum cytokine and chemokine levels, and increased levels of these proteins in culture media from the KO cells. TNF is a good example of how the gene expression changes seen in the unstimulated, triple KO macrophages are reflected in mouse phenotype. Tnf mRNA levels were not elevated in the unstimulated macrophages from the M-TTP KO mice, but were significantly increased by 3.5fold in the M-triple KO macrophages. This was reflected in increases in TNF protein levels in serum from the M-triple KO mice, even without innate immune stimulation. The severe phenotype observed in the M-triple KO mice could therefore be in part because of a combination of direct pathogenic effects of the elevated TNF, and autocrine/paracrine effects of TNF to promote the secretion of many other pro-inflammatory cytokines, including itself (Grivennikov et al, 2005; Kruglov et al, 2008; Moudgil & Choubey, 2011; Steinkamp et al, 2018; Yao et al, 2021). Similar severe phenotypes have been observed in mice with direct transgenic overexpression of TNF (Keffer et al, 1991; Probert et al, 1993), or mice in which an AU-rich instability region of the Tnf mRNA have been removed, resulting in systemic TNF overexpression (Kontoyiannis et al, 1999). It will be of great interest to see whether the M-triple KO phenotype can be modified by interfering with TNF activity, as has been done in the case of the conventional TTP KO mice (Taylor et al, 1996; Carballo & Blackshear, 2001), and other pro-inflammatory pathways that have been shown to have an ameliorating effect on the whole-body TTP-deficiency syndrome (Molle et al, 2013).

Examples of how this might work in the present study come from several groups of genes whose expression was altered in the M-triple KO macrophages in the absence of stimulation. For example, levels of 15 chemokine transcripts were significantly increased by more than 1.3fold in the unstimulated M-triple KO macrophages, with one of them, Ccl12 mRNA, increased by 238fold (adjusted P-value 6.02 × 10⁻¹¹). This transcript contained a single potential TTP family binding site, but two other highly up-regulated chemokine transcripts, Ccl7 (46fold increase) and Ccl8 (37fold increase) mRNAs, did not contain obvious binding sites, and may be examples of autocrine/paracrine secondary stimulation. Similarly, five transcripts from the Cxcl family of chemokines were increased, but only two of them, Cxcl1 and Cxcl2 mRNAs, are well-known TTP targets (Jalonen et al, 2005; Datta et al, 2008; Qiu et al, 2015), whereas the other three do not contain obvious TTP family binding sites. A final example is interferon β1 (Ifnb1) mRNA, whose levels were increased by about 10.5fold. This transcript has a single potential TTP binding site and was the only interferon transcript up-regulated in this collection. Perhaps in response, there were 22 interferon-activated or -induced transcripts among the up-regulated mRNAs, of which only Ifit1 mRNA (increased 15fold) contained an obvious TTP binding site.

Taken together, the data suggest that the severe phenotype of the M-triple KO mice was because of a combination of primary increased synthesis and secretion of certain pro-inflammatory proteins whose transcripts are direct targets of TTP family proteins, such as Tnf mRNA, and secondary effects of those primary pro-inflammatory proteins on the myeloid cells themselves and other cells and tissues that they influence.

Transcript stability changes in LPS-stimulated macrophages from M-TTP KO mice compared with cells from M-triple KO mice

We attempted to measure mRNA decay rates in WT, M-TTP KO, and M-triple KO macrophages to identify likely primary targets for TTP and its family members in this system, and to look for additivity or synergy between the family members. As described above, both Zfp36l1 and Zfp36l2 mRNA levels decreased rapidly to below baseline levels after LPS stimulation of WT macrophages, and it seemed possible that the high remaining levels of TTP in this situation would result in transcripts from the M-triple KO cells decaying at the same rate as those from the M-TTP KO cells.

However, this was not the case. Under the stringent criteria we used to measure changes in mRNA decay, transcripts could be divided into several groups. The largest group included transcripts in which no stability differences could be measured; most of these transcripts decayed too slowly in the control cells to be included in the analysis, and a smaller subset decayed fast enough to be analyzed but exhibited no differences among the three genotypes. However, for those transcripts in which significant differences in decay rates could be measured, three major groups could be delineated. In the first group, stability of transcripts from M-TTP KO cells did not differ from transcript stability in control cells, but stability was significantly increased when the M-triple KO cells were compared with control cells. The most striking examples of transcripts in this group are illustrated in Fig 10. For example, Rab3a mRNA decayed at the same rate as control in the M-TTP KO cells, but was completely stabilized in the M-triple KO cells. This transcript contains four potential TTP family member binding sites, some of them overlapping, and its encoded protein is involved in GTP-dependent protein binding, especially in neuronal processes (Geppert et al, 1997; Sheehan et al, 2016). Its roles in myeloid cells are unclear, but previous studies have suggested a role in plasma membrane repair and lysozyme exocytosis in response to bacterial toxins (Abu-Amer et al, 1999; Vieira, 2018). In the present situation, it is the clearest example of a transcript whose normal instability seems to be unaffected by TTP deficiency, but is completely stabilized in the absence of all three family members. As such, it may be a good marker transcript to use in addressing future questions about specificity among the family members, for example, is it a specific target for one of the three proteins, or is its normal processsing dependent on the presence of normal levels of all three proteins, acting in concert and perhaps synergistically?

In a relatively small group of transcripts, the mRNA decay rates were slowed to approximately the same extent in both the M-TTP KO and M-Triple KO cells. Good examples of members of this group are Cxcl2 and Ccl2 mRNAs, as shown in Fig 9. Both mRNAs have been described previously as TTP targets (Jalonen et al, 2005; Sauer et al, 2006; Qiu et al, 2015), and both encode important chemokines involved in leukocyte chemotaxis and pathogenesis of inflammatory diseases (Sauer et al, 2006; Deshmane et al, 2009; De Filippo et al, 2013).

A final group consists of transcripts in which significant stabilization occurred in the setting of TTP deficiency alone, but was increased further in the M-triple KO cells. This group contains some of the best known targets of TTP in these cells (Taylor et al, 1996; Carballo et al, 1998, 2000; Lai et al, 2006; Datta et al, 2008), including Tnf, Ier3, Cxcl1, and Csf2 mRNAs (see Fig 9). In these cases, the effects of knocking out all three family members appeared closer to additive.

Given the behavior of Zfp36l1 and Zfp36l2 transcripts in the control cells after LPS stimulation, that is, decreases below starting levels within a short period after LPS, it is difficult to explain their ability to contribute to the decay rates of the transcripts described above in the control cells. One possible explanation has to do with the fact that their levels are approximately the same as Zfp36 mRNA levels in the pre-stimulated, steady-state condition, possibly resulting in similar levels of all three encoded proteins that could persist for the relatively short duration of the ActD experiment. These protein levels could actually increase in response to the rapid, transient peaks in mRNA levels seen in the short term after LPS, and then persist for the duration of the decay experiment, despite the apparent suppression of their transcripts. Other explanations are possible, including changes in phosphorylation status, interactions with cytoplasmic binding proteins, and others. It seems clear, however, that the three encoded proteins, acting together, are often more potent in promoting mRNA decay in these experimental circumstances than TTP alone.

Physiological importance of the three expressed TTP family members in myeloid cells

An advantage of the severe, universal, early onset phenotype seen in the M-triple KO mice is that it provides an assay for the contributions of the different family members to the phenotype. Although the numbers of mice were small in each group, we found that two normal alleles of any of the three genes, in the setting of myeloid cell deficiency of the other two, could completely prevent the early death, growth inhibition, and peripheral arthritis seen in the M-triple KO mice. Even one normal allele of Zfp36 or Zfp36l2, in the absence of all others, could prevent the development of the syndrome, whereas one normal allele of Zfp36l1 in the absence of the others led to a less severe version of the M-triple KO phenotype. This is additional evidence that the protein activities of the three paralogues are to some extent overlapping and redundant.

In contrast to the situation in mice living under normal vivarium conditions, the M-TTP KO mice were extremely susceptible to small doses of LPS, and died of a sepsis-like syndrome that their WT counterparts survived without difficulty (Kratochvill et al, 2011; Qiu et al, 2012). It will be of great interest to see how the apparently phenotypically normal mice that express only one of the three genes will respond to this kind of inflammatory stimulus. It will also be interesting to correlate those results with the effects of the double KOs on transcript expression and stability in the presence and absence of LPS. These types of experiments are currently under way.

LysM-Cre has been shown to provide conditional KOs in several different types of myeloid cells (Clausen et al, 1999), and it will be of great interest to try to dissect the effects of deficiencies of the different family members in specific cell types, using cell type-specific Cres. Another approach to the question of cell specificity might be to sort the various cell types present in the bone marrow and other tissues from the M-TTP KO and M-triple KO mice, and examine their gene expression patterns and transcript stability with or without an inflammatory stimulus.

Although the phenotype of the M-triple KO mice was very severe and affected many organ systems, it was almost as striking to note abnormalities that might be expected that did not occur. For example, most models of TNF excess, either by transgenic overexpression or by deletion of the AU-rich instability element, are associated with a severe form of inflammatory bowel disease (Kontoyiannis et al, 1999). We have never seen this in the conventional TTP KO mice (Taylor et al, 1996), nor was it present in the M-triple KO mice described here, despite evidence of TNF overexpression from macrophages and presumably other myeloid cells. Similarly, a severe form of ocular inflammation leading to blindness was seen in the T-cell conditional triple KO mice described recently by Cook et al (2022), whereas we did not observe eye abnormalities in the M-triple KO mice. In their study, Cook et al (2022) found that mice with simultaneous deletion of Zfp36, Zfp36l1, and Zfp36l2 in T cells under the control of the Cd4-Cre developed a lethal inflammatory syndrome with multi-organ involvement and overproduction of inflammatory cytokines (Cook et al, 2022).

Previous work has shown that mice with double KO of Zfp36l1 and Zfp36l2 in T cells during thymopoiesis develop T cell acute lymphoblastic leukemia because of changes in Notch-1 signaling (Hodson et al, 2010). However, mice with single KO of Zfp36l1 or Zfp36l2 individually in T cells during thymic development do not develop T cell acute lymphoblastic leukemia, suggesting that Zfp36l1 and Zfp36l2 may have redundant functions in T cells (Hodson et al, 2010). Mice with deletion of either Zfp36l1 or Zfp36l2 individually in muscle cells using Pax-7 Cre did not display any growth defects or phenotype, yet mice with deficiency of Zfp36l1 and Zfp36l2 simultaneously in Pax7-expressing satellite cells had reduced body weight, reduced skeletal muscle mass, and reduced capacity to regenerate muscle after muscle injury (Bye et al, 2018).

In summary, the complete absence of all three expressed TTP family members from myeloid cells in mice led to a spontaneous, severe inflammation syndrome that affected many organ systems. This could be prevented by normal expression of any of the three family member genes, and, in two instances, by only a single allele of those genes. This shows a remarkable degree of redundancy among the three expressed family members, and may help explain certain experiments of nature, like the ability of birds to survive despite their apparent lack of TTP expression (Lai et al, 2013). Many questions remain unanswered, but among the most important is the question of specificity and cross-reactivity among the different family members, both in cell-free experiments with purified components of the mRNA deanylation apparatus, and in intact cells.

Mice

Mice with loxP sites flanking the second exon of each respective gene (Zfp36^flox/flox, Zfp36l1^flox/flox, or Zfp36l2^flox/flox) were generated by gene targeting in C57Bl/6 embryonic stem cells by Xenogen Biosciences, as previously described (Qiu et al, 2012; Hyatt et al, 2014; Dumdie et al, 2018). Expression of Cre recombinase under the control of the murine M lysozyme promoter (LysM-Cre) was used to effect deletion specifically in cells of the myeloid lineage (monocytes, macrophages, and granulocytes). Homozygous LysM-Cre (B6.129P2-Lyz2^tm1(cre)Ifo) mice on a C57Bl/6 background were purchased from the Jackson Laboratory (JAX stock #018956) (Clausen et al, 1999; Takeda et al, 1999). Individual KO mice for each TTP family member were generated by crossing either Zfp36^flox/flox, Zfp36l1^flox/flox, or Zfp36l2^flox/flox mice with LysM-Cre mice. For experiments using Zfp36^flox/flox; LysM-Cre^+/− mice, Zfp36^flox/flox; LysM-Cre^−/− littermates were used as controls.

A mouse line in which all three TTP family members were floxed, referred to as “triple floxed” mice, was generated by inter-crossing and breeding Zfp36^flox/flox, Zfp36l1^flox/flox, and Zfp36l2^flox/flox mice to homozygosity for the floxed alleles of Zfp36, Zfp36l1, and Zfp36l2. Next, myeloid-specific triple (M-triple) KO mice were achieved by crossing the triple-floxed (Zfp36^flox/flox; Zfp36l1^flox/flox; Zfp36l2^flox/flox) mice with LysM-Cre mice. In this breeding scheme, various myeloid-specific (M-double) KO combinations were also generated. The deletion of Zfp36, Zfp36l1, Zfp36l2 in macrophages was confirmed (data not shown) and the efficiency of the LysM-Cre in myeloid cells has been previously reported (Clausen et al, 1999; Takeda et al, 1999; Qiu et al, 2012). For experiments using Zfp36^flox/flox; Zfp36l1^flox/flox; Zfp36l2^flox/flox; LysM-Cre^+/− mice, Zfp36^flox/flox; Zfp36l1^flox/flox; Zfp36l2^flox/flox; LysM-Cre^−/− littermates were used as controls.

Tissue processing

Mice were euthanized at 8–10-wk-old, gross examinations of all organs were performed, and the following tissues were harvested: adrenal gland, brain with olfactory nerve, cervix, esophagus, eyes with optic nerve, femur, gallbladder, Harderian glands, heart/aorta, small and large intestines (duodenum, jejunum, ileum, cecum, colon, rectum), kidney, liver, lungs, lymph nodes, mammary gland, skeletal muscle, sciatic nerves, nose, nasal cavity, ovary, pancreas, pituitary gland, prostate gland, salivary gland, seminal vesicle, haired skin, spinal cord, spleen, stomach (glandular and nonglandular), testes, thymus, thyroid gland, tongue, trachea, urinary bladder, uterus, and vagina. Internal organs were fixed in 10% neutral-buffered formalin and used for paraffin embedding, sectioning, and hematoxylin and eosin staining. Tissue samples containing bones, such as paws, were decalcified with Immunocal (StatLab) for greater than 8–36 h. Microscope images were captured using a Hamamatsu C13220 Microscope with a 20X objective with a 40X lens and 0.75 aperture. The magnification of each image is specified in the respective figure legend.

Peripheral blood analysis

For hematological analysis, peripheral blood was collected by cardiac puncture, collected into K3-EDTA-containing Microvette blood collection tubes (Sarstedt), and analyzed using a Procyte Hematology Analyzer (IDEXX).

Measurement of cytokine levels in serum

Peripheral blood was collected by cardiac puncture into a Micro Z-Gel serum separation tube (Sarstedt), followed by centrifugation for 5 min at 10,000g at RT, after which the supernatant was collected. Cytokine and chemokine analyses were performed using a Mouse Cytokine/Chemokine 31-Plex Array (Eve Technologies).

MicroCT

Front and hind paws were fixed in 10% neutral-buffered formalin for 72 h, followed by storage in 70% ethanol. The paws were removed from alcohol storage and mounted within a sealed plastic tube sample holder (Corning) with internal physical anchoring by foam to prevent movement during scanning. The samples were scanned using a SKYSCAN 1272 (Bruker microCT) at a nominal resolution of 6.5 microns, using a 0.25-mm-thick aluminum filter, and an applied x-ray voltage of 65 kV and 153 μA. Camera pixel binning of 2 × 2 was applied. The scan orbit was 360° with a rotation step of 0.32°. Reconstruction was carried out with a modified Feldkamp algorithm using the SkyScan NRecon software accelerated by a graphics processing unit. Appropriate ring artefact reduction and 30% beam hardening correction were applied. A SkyScan CT-Analyzer software suite was used for 3D model reconstructions and bone morphometric analyses. Reconstructed data were volume rendered in CTVoxx (Bruker microCT) and a custom transfer function was created to completely reduce the tissue opacity from the window/level of view. In addition, an RGB color scale was applied to the data to better translate the surface of the bone. All samples were consistently handled such they all received the same transfer function, RGB, and lighting settings.

Flow cytometry analysis

8–10-wk-old mice were euthanized by CO₂ inhalation, and femurs were dissected. Bone marrow cells were isolated by flushing dissected femurs with 5 ml PBS containing 2% FBS, using a 25-gauge needle. The cell suspension was filtered over a 100-μM nylon mesh strainer (Thermo Fisher Scientific), followed by centrifugation for 5 min at 500g at 4°C. Cells were resuspended as a single cell suspension, and 1 × 10⁵ cells were stained with 25 μl of Brilliant Stain Buffer (BD Biosciences) and the following antibodies for 30 min at 4°C in the dark: BV421–anti-CD45, AF488–anti-CD3, PE–anti-CD4, AF647–anti-CD8, PE-Cy7–anti-CD19, AF700–anti-CD11b, PerCP-Cy5.5–anti-CD11c, BV480–anti-MHC-II, APC-Cy7–anti-Ly-6G, PE-CF594–anti-Ly-6C (BD Biosciences). Red blood cells were lysed with FACSLyse (BD Biosciences), followed by centrifugation for 5 min at 500g, and resuspension in staining buffer. Data were acquired on an LSRII Flow Cytometer (BD Biosciences) and analyzed with FlowJo software (FlowJo BD).

For the analysis of hematopoietic stem and progenitor cells, 1 × 10⁷ bone marrow cells were treated with 1 ml ACK lysis buffer (0.1 M NH₄CH₃CO₂, 10 mM KHCO₃, 0.1 mM EDTA, produced by NIEHS Media and Glassware Units) at RT for 30 s to deplete red blood cells, then quenched with staining medium (PBS supplemented with 2% FBS and 2 mM EDTA). Cells were then centrifuged at 300g for 5 min, resuspended in the staining medium, and stained on ice for 30 min with the following antibodies: CD117 (c-Kit) APC (105808; BioLegend), Sca-1 APC-Cy7 (108126; BioLegend), CD3 Biotin (100244; BioLegend), Ly-6G/Ly-6C (Gr-1) Biotin (108404; BioLegend), B220 Biotin (103204; BioLegend), TER-119 Biotin (116204; BioLegend), CD11b Biotin (101204; BioLegend), CD150 (SLAM) PE-Cy7 (115914; BioLegend), CD41 FITC (133904; BioLegend), CD105 (Endoglin) PerCP/Cyanine5.5 (120416; BioLegend), CD16/32 (FcγRII/III) eFluor 450 (48-0161-82; eBioscience). After the primary staining, cells were then washed once, incubated with Streptavidin APC (SA1005; Invitrogen) on ice for 15 min to reveal biotin-conjugated lineage antibodies, and then resuspended at 5 × 10⁶ cells/ml for flow cytometry. Data were collected using an LSRFortessa instrument (BD Biosciences) and analyzed using BD FACSDiva and FlowJo (BD Life Sciences).

The following surface marker combinations were used to identify adult HSPC subtypes (Nakorn et al, 2003; Pronk et al, 2007; Weksberg et al, 2008; Cabezas-Wallscheid et al, 2014): HSC (Lin−, Sca1+, c-Kit+, CD150+); multipotent progenitor (Lin−, Sca1+, c-Kit+, CD150−); Mye Pro(Lin−, Sca1−, c-Kit+); MKP (Lin−, Sca1−, c-Kit+, CD150+, CD41⁺); PreMegE (Lin−, Sca1−, c-Kit+, CD41−, CD16/32−, CD150+, Endoglin−); Pre CFU-E (Lin−, Sca1−, c-Kit+, CD41−, CD16/32−, CD150+, Endoglin+); CFU-E and Pro Ery (Lin−, Sca1−, c-Kit+, CD41−, CD16/32−, CD150−, Endoglin+); PreGM (Lin−, Sca1−, c-Kit+, CD41−, CD16/32−, CD150−, Endoglin−); granulocyte–macrophage progenitor (Lin−, Sca1−, c-Kit+, CD41−, CD16/32+).

Culture and treatment of BMDM

8–9-wk-old mice were euthanized by CO₂ inhalation, and both femurs were isolated aseptically. In a tissue culture hood, both femurs from each animal were flushed with RPMI 1640 Medium (Gibco; Thermo Fisher Scientific), using a 25-gauge needle over a 100-μM nylon mesh strainer (Thermo Fisher Scientific), followed by centrifugation for 5 min at 500g at RT. The pelleted cells from each animal were resuspended and plated into five, 10 cm TC-treated culture dishes (Corning) and cultured in RPMI 1640 Medium (Gibco; Thermo Fisher Scientific) supplemented with 30% L929-conditioned medium, 10% ES Cell-screened FBS (HyClone; Cytiva), 15 mM HEPES (Sigma-Aldrich), 2 mM glutamine, and 100 U; 100 μg/ml penicillin/streptomycin (Gibco; Thermo Fisher Scientific). On day 4 after plating the cells, the floating cells remaining in suspension were collected and plated onto new plates for experiments. Supplemental medium was added every other day until the cells reached ∼80% confluency around days 7–10 (Qiu et al, 2012; Lai et al, 2019b, 2019c).

Once reaching confluency, BMDM were then incubated for 16–18 h in a serum starvation medium (RPMI supplemented with 1% FBS, 2 mM glutamine, 100 U penicillin/100 μg/ml streptomycin), followed by treatment with 1 μg/ml LPS (L6529; Sigma-Aldrich) and/or ActD (A4262; Sigma-Aldrich) for the designated time points (Lai et al, 2019a). For mRNA decay experiments, four mice of each genotype were used, and three plates of cells from each mouse were pooled for each time point. For mRNA accumulation experiments, four plates of cells were pooled and used as controls (untreated), and three plates of cells were pooled and used for each time point of LPS treatment (1 μg/ml).

RNA extraction

Total cellular RNA was isolated from cultured bone marrow macrophages using the illustra RNAspin RNA isolation kit, according to the manufacturer’s instructions (GE Healthcare). The RNA content and purity were determined by measuring absorbance at 260/280 nm on NanoDrop One (Thermo Fisher Scientific) and the quality of RNA was determined using TapeStation System (Agilent).

NanoString RNA analysis

Total cellular RNA from BMDM used in the RNA accumulation experiments was analyzed using the NanoString nCounter method (Fortina & Surrey, 2008) for Zfp36, Zfp36l1, and Zfp36l2 mRNAs. The observed counts were normalized by a series of internal spike-in controls, and a set of housekeeping controls that had been previously validated as stable during LPS treatment.

Library preparation and RNA sequencing

Library preparation was performed using the TruSeq RNA Library Prep Kit (Illumina Inc.) at the NIEHS Epigenomics Core Laboratory. RNA deep-sequencing was performed using 75-bp paired-end reads on the NovaSeq6000 platform (Illumina Inc.) at the NIEHS Epigenomics Core Laboratory.

RNA-seq data analysis

RNA-seq data analysis was performed by beginning with fastq sequence files of 75-base pair-ended reads and subjecting them to read quality and length trimming using bcl2fastq2. Read pairs were filtered to remove those in which either mates’ mean quality score fell below 20. 50 million reads per sample were aligned to the mm10 transcriptome using STAR 2.7.0f. Counts per gene were determined using featureCounts 1.5.1. Differentially expressed transcripts were identified by DESeq2 using filtered thresholds of FDR < 0.05 and log₂ fold change ≥ 0.3785 (or a raw fold change of 1.3), with the additional requirement of a mean control FPKM > 0.1 and an adjusted P-value ≤ 0.05.

Analysis of mRNA decay from RNA-seq data

mRNA decay curve analysis was performed by first removing non-mRNAs, the knocked out genes (Zfp36, Zfp36l1, and Zfp36l2), the artefactually up-regulated gene Plekhg2, mRNAs whose mean levels in the control cells were < 0.1 FPKM, and mRNAs that did not decrease below 85% mRNA remaining in control BMDM after 120 min of ActD. The data were then converted to the percentage of the original mRNA remaining after the 1 h LPS time point. To narrow down the list of transcripts to analyze the decay curves, an initial screen was performed using two-tailed t tests with Bonferroni correction between the control and M-TTP KO cells or the control and M-triple KO cells at three consecutive time points, with a cutoff of P < 0.008. After this initial list of candidates was identified, two-way ANOVA with Geisser–Greenhouse correction and Šídák’s multiple comparison test was performed with a cutoff of an ANOVA P-value < 0.05, used as a more stringent statistical test to identify genotype differences between the full decay curves. The apparently stabilized transcripts were then ranked by the biggest difference in the average percent mRNA remaining at the 60- and 120-min time points between the control and M-TTP KO samples, or the control and M-triple KO samples. One thing to note is that addition of ActD causes a large number of transcripts to decay rapidly, resulting in a decline in the total RNA isolated at later time points. Standard normalization methods used in RNA-seq assume that libraries are prepared from comparable pools of RNA. Therefore, levels of non-decaying or slowly decaying transcripts will artefactually appear to increase (Lai et al, 2019a). However, in the experiments described above, each pair of WT and KO cells was treated in the same experiment, and any drift in the baselines should be comparable between the genotypes throughout the experiment.

Identification of TTP binding sites

A custom TTP binding site search application (available upon request) was used to scan the 3′-UTR of potential target transcripts for the presence of possible high-affinity TTP binding sites: UAUUUAU (7-mer), UAUUUUAU (8-mer), UUAUUUAUU (9-mer), and UUAUUUUAUU (10-mer), as previously described (Lai et al, 2013; Patial et al, 2016b).

IPA

The differentially expressed transcripts were subjected to IPA to investigate the biological networks and pathways that were enriched under these conditions (QIAGEN).

Statistical analysis

Results are expressed as means ± SD unless otherwise specified. The statistical significance was assessed as indicated using GraphPad Prism 9.0. The following tests were used: unpaired, two-tailed t tests with Welch’s correction, one-way ANOVA, and two-way ANOVA test with Šidák’s multiple comparison test. A P-value less than 0.05 was considered significant.

Study approval

All animal breeding and other procedures were approved by the Institutional Animal Care and Use Committee of the National Institute of Environmental Health Sciences. Animals were maintained in a specific pathogen-free facility with ad libitum access to food and water.