Fine tuning of TCR signaling via CD8αβ and PD-1 and the fate of autoreactive thymocytes during negative selection

Abstract

During thymocyte development, positive selection produces cells whose T cell receptors (TCRs) bind to self MHC. Then, negative selection culls most thymocytes whose TCRs have too high an affinity for self MHC+peptide. Signal transduction events control these processes. CD8-αβ (via CD8-β) recruits p56^lck to the immunological synapse and promotes signaling through the TCR. Conversely, PD-1 attenuates TCR signal transduction. We examined the roles of CD8-β and PD-1 in the survival of thymocytes in H-2^k haplotype mice expressing a transgenic BM3 TCR, which has high affinity for the allogeneic H-2K^b MHC I molecule. In transgenic mice expressing both H-2K^b in the thymic medulla and the BM3 TCR, apoptosis eliminates most (but not all) post-selection thymocytes. To analyze the roles of CD8-β and PD-1 in the survival of post-selection thymocytes, we devised a novel probabilistic gating strategy employing Gaussian mixture models and statistical methods using sliding windows and changepoint detection. We found that at high levels of CD8-β and therefore high levels of CD8-αβ), thymocytes are prone to apoptosis, regardless of the PD-1 level. At intermediate levels of CD8-β, thymocyte survival increases concordantly with increasing PD-1 levels. At low levels of CD8-β, thymocyte survival is high regardless of the PD-1 level. Surviving DP^lo post-selection thymocytes give rise to PD-1⁺CCR7⁺DN and PD-1⁺CCR7^-DN post-selection thymocytes, which appear to become DN T cells and IELs, respectively. Thus, PD-1 appears to promote the survival of both IEL precursors and thymocytes destined for other fates. More strikingly, downregulation of CD8-β is a hallmark of autoreactive MHC I-restricted thymocytes that survive negative selection.

Introduction

To respond to an ever-changing antigenic universe, immunoglobulin and T cell receptor (TCR) gene rearrangements occur in a largely random manner. To eliminate useless and autoreactive receptors, thymocytes bearing newly rearranged αβ-TCRs then undergo two successive selection steps. Positive selection allows the survival of CD4⁺CD8⁺ (DP) thymocytes whose αβ-TCRs bind to self-MHC + peptide present on thymic cortical epithelial cells (cTECs). Next, thymocytes undergo negative selection both at the corticomedullary junction¹ and in the medulla. Post-positive selection DP^lo thymocytes and CD4 and CD8 single-positive (SP) thymocytes encounter self-MHC+self-peptide presented on medullary epithelial cells (mTECs) and on dendritic cells. Those thymocytes whose αβ-TCRs bind too avidly die by apoptosis.

Negative selection does not eliminate all autoreactive thymocytes. In both non-transgenic and TCR transgenic mice, some autoreactive thymocytes survive and differentiate into intraepithelial lymphocyte precursors (IELps).^2–5 What determines life versus death of autoreactive thymocytes? Robey and colleagues showed that intermediate MHC I+self-peptide reactivity results in negative selection while excessively strong MHC I+self-peptide reactivity promotes differentiation into IELps.⁶ However, TCR affinity for its ligand is not the sole determinant of thymocyte fate. Signal transduction events involving the TCR and other cell-surface molecules also play decisive roles. DP thymocytes, CD8⁺ SP thymocytes, and mature CD8⁺ T cells express the CD8 coreceptor as a heterodimer of CD8-α and CD8-β. In contrast, IELs express CD8 as a CD8-αα homodimer.^{2,  7} CD8-αα and CD8-αβ have a similar affinity for MHC I.⁸ However, they are not functionally equivalent because of properties unique to the CD8-β chain. CD8-β enhances TCR-MHC I + peptide binding.⁹ Furthermore, CD8-β, but not CD8-α, is palmitoylated. Palmitoylation of CD8-β greatly increases its association with p56^lck.10,11 This enables CD8-αβ to localize into lipid rafts,¹² where it interacts with LAT.¹¹ Additionally, CD8-β promotes TCR clustering within lipid rafts.^11,12 Collectively, the amassed p56^lck-CD8-αβ, LAT, and associated signal transduction components promote TCR signaling. This is presumably the reason that the CD8-β chain is required for positive selection of MHC I-specific thymocytes and for their development into CD8⁺ SP thymocytes.^13–15 But how about CD8-β’s role in negative selection? A reduction in the density of cell-surface CD8-αβ heterodimers should compromise signaling through the TCR, which may allow escape from negative selection.^16,17 Indeed, it has been suggested that down-regulation of CD8-β in autoreactive thymocytes diminishes the TCR signal strength to a level below that required for apoptosis.^16,17

Another important regulator of TCR signaling is PD-1, which attenuates signaling through the TCR.^18–21 PD-1 expression is high in thymocytes that have survived strong agonist signaling. These thymocytes, which are DP^lo but have increased levels of CD3 and CD5, include IELps.^{4,5,  22}

To study the fates of autoreactive thymocytes that survive negative selection, we used transgenic and double transgenic mice derived from the BM3.3 mouse. This mouse strain is of the CBA (H-2^k) background and expresses a TCR specific for the allogeneic H-2K^b molecule.²³ We wished to answer 2 important questions: How do autoreactive thymocytes escape negative selection? What happens to them next—do they all become IELps? We examined the development of thymocytes in 6 types of transgenic mice that express either the BM3.3 TCR alone or express the BM3.3 TCR and H-2K^b driven by one of two promoters previously shown to be expressed by thymic epithelial cells (Table 1). Importantly, negative selection does not occur prematurely in the double transgenic strains. This allowed us to study the fates of autoreactive thymocytes in the face of agonist H-2K^b expression.

Table 1.

Mouse strains used in these studies.

Mouse strain	Important characteristics	Ref. or source
BM3.3	CBA (H-2k) strain, transgenic for H-2Kb-specific TCR	23
BM3 rag−/−	BM3.3 TCR transgenes bred onto a CBA rag1 gene knockout background	Gift from Nicholas Jones
KAL	CBA strain carrying an H-2Kb transgene driven by the guinea pig α-lactalbumin promoter	24
KQ	CBA strain carrying an H-2Kb transgene driven by the human keratin 14 promoter	This paper
KALxBM3	Fully homozygous hybrid of KAL bred to BM3.3	This paper
KBBR	Fully homozygous hybrid of KAL bred to BM3 rag−/−	This paper
KQxBM3	Fully homozygous hybrid of KQ bred to BM3.3	This paper
KQBR	Fully homozygous hybrid of KQ bred to BM3 rag−/−	This paper

We developed a novel probabilistic analytical approach to determine the roles of CD8-β and PD-1 in the survival of post-selection DP^lo thymocytes. Post-selection DP^lo thymocytes having high levels of CD8-αβ are prone to apoptosis regardless of the level of PD-1. Down-regulation of CD8-β strongly correlates with the survival of post-selection DP^lo thymocytes. Furthermore, the survival of autoreactive thymocytes having low to intermediate levels of CD8-αβ increases with rising PD-1 expression. The surviving thymocytes appear to go on to develop into DN T cells that populate lymph nodes and into DN and CD8-αα IELs in the gut. Thus, CD8-β downregulation and PD-1 upregulation allow the survival of autoreactive thymocytes that may go on to populate diverse peripheral lymphoid tissues.

Materials and methods

Mice used in these studies

The Ker14-H-2K^b transgenic mouse line is called KQ was generated at the Yale University School of Medicine by microinjection of DNA into fertilized eggs derived from (C3HxC57Bl/10) × (C3HxC57Bl/10) matings. In this transgenic mouse line, we used a vector bearing the human keratin 14 (Ker14) promoter, as wells as human growth hormone gene sequences to promote gene expression.²⁵ We cloned an H-2K^b cDNA into the unique BamHI site of the Ker14/hGX expression vector.²⁵ Mice expressing the Ker14-H-2K^b transgene were identified by polymerase chain reaction (PCR) using primers specific for the human Ker14 promoter.²⁵ Because the transgenic mice could be b/b, b/k, or k/k haplotype, the founder transgenic mice were haplotyped by PCR using primers specific for the I-E-β gene.²⁶ Founder mice were bred onto the C3H background for over 15 generations before use in experiments. Subsequently, the Ker14-H-2K^b transgene was bred onto the CBA background.

The previously characterized BM3.3 transgenic mouse line carries the functionally rearranged TCR-α and -β chain genes derived from an allogeneic CTL line specific for H-2K^b.27 The KAL transgenic mouse line carries the H-2K^b gene driven by the guinea pig α-lactalbumin promoter.²⁴ Both transgenic mouse lines are on the CBA strain background.

The KALxBM3 and KQxBM3 mice were generated by crossing KAL and KQ with BM3.3, respectively. The KALxBM3xBM3 rag^−/− (KBBR) and KQxBM3xBM3 rag^−/− (KQBR) were generated by crossing the KALxBM3 and KQxBM3 with BM3 rag^−/−, followed by subsequent multiple rounds of breeding to yield mice homozygous for rag deficiency and for expression of H-2K^b. Mouse strains and their sources are summarized in Table 1.

All mice used in these studies were maintained and bred in-house at the University of California, Santa Cruz Vivarium. All procedures were in accordance with the University of California, Santa Cruz Chancellor’s Animal Research Committee (protocol ID = Zunim2212dn_r2) and with the recommendations from the Guide for the Care and Use of Laboratory Animals by the National Research Council (eighth edition). These studies were done over a twenty-year time span (but chiefly within the last decade). Some of the experiments were done numerous times over several years. Hence, the numbers of mice in some of the data figures are large. We also routinely isolated lymph nodes from mice at the time of thymus harvest.

Preparation of tissues for flow cytometry and RNA isolation

Thymic lobes and lymph nodes were isolated from mice at 6–8 wk of age. Single cell suspensions were obtained by mashing tissues between the frosted edges of glass slides.

IELs were isolated from 11 to 13-wk-old mice according to a previously published protocol,²⁸ with the following modifications. Harvested intestines were cut into 4 pieces before everting. Hanks was used instead of RPMI in the IEL isolation medium. Samples were agitated in IEL isolation medium in capped 250 ml Nalgene Erlenmeyer flasks in a 37 °C shaking water bath for 37 min.

For thymic stromal cell preparations, thymic lobe tissues were triturated to deplete thymocyte fragments and then agitated for 10 min at 37 °C. Stromal cells were released from the tissue fragments by digestion at 37 °C with Blendzyme digestion mix (Roche) supplemented with 150 U/ml DNase in Ca²⁺ free Mg²⁺ free Hanks and 6 mM CaCl_2.²⁹ The released stromal cell preparation was then filtered through a 100-micron cell strainer. Stromal cell preparations were depleted of CD45⁺ cells with biotinylated anti-CD45 bound to MagVigen-Streptavidin magnetic nanoparticles on ice for 35 min. Reagent suppliers are listed in Table 2.

Table 2.

Catalog and Purchase Information on antibodies and reagents used in this study.

Item	Catalog number	Company
Goat anti-mouse IgG (H&L), F(ab′)2 Fragment—Affinity Pure, FITC Conjugate	35-8054	Tonbo Biosciences
Normal mouse serum	sc-45051	Santa Cruz Biotechnology
Normal rat serum	sc-45135	Santa Cruz Biotechnology
CD3epsilon (145-2C11) Armenian hamster mAb FITC	100305	BioLegend
Mouse S1P1/EDG-1 Alexa Fluor 647 Antibody (Clone 713412)	FAB7089R-100UG	Bio-Techne
Mouse S1P1/EDG-1 Allophycocyanin MAb (Clone 713412), 100 UG	FAB7089A-100UG	Bio-Techne
MagVigen-Streptavidin for cell separation	21005C	Nvigen Inc
BASE PAIR 10 NMOL SCALE	DNA001	Fisher Scientific Company
PE/Cy7 anti-mouse CD8β.2 Antibody	140415/25 μg	BioLegend
	140416/100 μg
PE/Cy5 anti-mouse/human CD44 Antibody	103009/25 μg	BioLegend
	103010/100 μg
PE/Cyanine5 anti-mouse CD122 (IL-2Rβ) Antibody	123219/25 μg	BioLegend
	123220/100 μg
Pacific BlueTM anti-mouse CD8α Antibody	100728/25 μg	BioLegend
	100725/100 μg
PE anti-mouse CD279 (PD-1)	135205	BioLegend
PE/Cy7 anti-mouse CD197 (CCR7)	120123/25UG	BioLegend
	120124/100UG
PE/DazzleTM 594 anti-mouse CD103	121429	BioLegend
PE/DazzleTM 594 anti-mouse CD5	100643/25UG	BioLegend
	100644/100UG
AF488 anti-mouse IgG	A11017	Invitrogen
7-AAD	420403	BioLegend
APC Annexin V	640920	BioLegend
Biotin anti-mouse CD45.2	109803	BioLegend
Biotin anti-mouse CD5	100603	BioLegend
PE/DazzleTM 594 anti-mouse CD69	104535	BioLegend
APC anti-mouse CD62L	104412	BioLegend
APC/Cyanine7 anti-mouse CD127 (IL-7Rα)	135039	BioLegend
APC/Cyanine7 anti-mouse CD8β.2	140421	BioLegend
PE/DazzleTM 594 anti-mouse CD62L	104447	BioLegend
Brilliant Violet 605 anti-mouse CD25	102036	BioLegend
Biotin anti-mouse CD11c	117303	BioLegend
Alexa Fluor® 700 anti-mouse CD5	100635	BioLegend
T4 Gene 32 Protein	M0300S	New England BioLabs
RNase H	M0297S	New England BioLabs

Flow cytometric analysis

Single-cell suspensions were incubated with the monoclonal antibodies and reagents shown in Tables 2 and 3. Cells treated with biotinylated anti-CD5 were washed and subsequently stained with streptavidin bound to PE-Dazzle. For detection of H-2K^b expression, cells were washed and subsequently stained with anti-mouse IgG conjugated to AF488 or FITC.

Table 3.

Antibodies and reagents used for flow cytometry studies.

Clone/reagent	Marker/specificity	Fluor/ligand
53-6.7	CD8α	Pac Blue, PE, APC, FITC
53-7.3	CD5	Biotin, PE/Dazzle, AF700
Streptavidin	Biotin	AF700, PE/Dazzle
17A2	CD3	APC, FITC, AF700
GK1.5	CD4	AF488, BV605, AF647, FITC, PE
Mel-14	CD62L	APC, PE/Dazzle, PE/Cy7
4B12	CCR7	PE/Cy7
145-2C11	CD3	PE/Cy7, PE, AF488
53-5.8	CD8β	PE/Cy7
29F.1A12	PD-1	PE
30-F11	CD45	PE/Cy5
IM7	CD44	PE/Cy7, PE/Cy5
B20.6	Vβ2	FITC, PE
2E7	CD103	PE/Dazzle
M5/114	MHC Class II	VF450
UEA-1		FITC
6C3	Ly51	PE
G8.8	EpCAM	APC
Purified antibody or antibody fragments	Mouse IgG	FITC, AF4888
Annexin V	Phosphatidyl serine	APC
Propidium iodide (PI)
34-1-2S	H-2Kb biotin	Biotin
93	CD16/CD32	None

After blocking with mAb 93 (Fc block) or with a mixture of 5% normal mouse and 5% rat serum, cells were incubated with antibodies diluted into Hanks supplemented with 2% fetal bovine serum for 30 min on ice. When staining for CCR7, incubations were at 37 °C for 30 min. The samples were resuspended in PBS supplemented with 2% fetal bovine serum and 20 mM EDTA (FACS stain buffer) then filtered through Nylon Mesh prior to analysis. When analyzing cells for apoptosis, stained cells were resuspended in 10 mM Hepes/NaOH, pH 7.4 supplemented with 140 mM NaCl and 2.5 mM CaCl₂ (Annexin V Binding buffer) and incubated at room temperature for 15 minutes. Cells were then incubated with propidium iodide and filtered through Nylon mesh prior to analysis. For all flow cytometry experiments, samples were analyzed on the LSRII flow cytometer (Becton Dickinson) or on the Cytoflex (Beckman Coulter).

Immunohistochemistry

Thymus and skin tissues were cryosectioned at a thickness of 5–10 μm and fixed in acetone. Immunohistochemistry was done using a published protocol.³⁰ Briefly, to detect MHC I expression, tissue sections were incubated with digoxigenin-conjugated H-2K^b-specific mAb 20.8.4 or mAb Y-3. (The 20.8.4 and Y-3 monoclonal antibodies were purified from hybridoma culture supernatants and conjugated to digoxigenin in our lab.) After washing, the sections were incubated with anti-Dig-POD. Visualization of antibody binding was achieved with diaminobenzidine. Skin grafting experiments were performed as described previously.³¹

RNA isolation and cDNA synthesis

For RNA isolation, thymus tissue was first disrupted in the Bioruptor sonicator (Diagenode) using three 30-s cycles on high, followed by manual disruption methods with forceps. RNA was extracted with Trizol (Qiagen) according to manufacturer’s instructions. The RNA was treated with DNase at 37 °C for 30 min. DNase was inactivated by incubation with DNase STOP at 65 °C for 10 min. cDNA was synthesized with the GoScript Reverse Transcription System (Promega), with the following modification: T4 gene 32 protein was included in the reactions to increase the yield of cDNA.^32,33 Samples were treated with E. coli RNase H after cDNA synthesis.

Primer sequences for genotyping of mice via PCR and for RT-PCR

Primers for genotyping are:

Primers for genotyping KAL, KALxBM3, and KBBR mice are—KAL forward primer: GGGGCTGAGTGTGTTGGAGT and KAL reverse primer: GAGTAGGCACCT ATGCAGCC. Primers for genotyping KQ, KQxBM3, and KQBR mice are—hGH forward primer: TAGGAAGAAGCCTATATCCCAAAGG and hGH reverse primer: ACAGTCTCTCAAAGTCAGTGGGG. Primers for genotyping BM3, BM3rag^−/−, KALxBM3, KBBR, KQxBM3, and KQBR mice are—BM3 Vβ2 forward primer: CCCAAGGTGGCGTCT GGTAC and BM3 Vβ2 reverse primer: CAATGAGCCGGCTTCCTTCT. Primers for distinguishing rag-sufficient from rag-deficient mice are—Rag-1 forward primer: CTCTTGCCGATATCCGTGCT and Rag-1 reverse primer: CCTCTAATTCATCAGCTTGCCTG. Primers for genotyping mice for H-2^k versus H-2^b haplotype background are: H-2E_β forward primer: CGACTGTAGAACCTTAGCCTG and H-2E_β reverse primer: GTGGACACAATTCCTGTTTT.

Primers for RT-PCR experiments include the following: H-2K^b forward primer: TACATGGAAGTCGGCTACG; H-2K^b reverse primer: CAGAGATCACCTGAATAGTG; Keratin 14 forward primer: CTAGCCGCATGTCCTCCATC; Keratin 14 reverse primer: GCAGGAGGACATTGGCATTG; mouse β-actin forward primer: ATATCGCTGCGCTGGTCGTC; mouse β-actin reverse primer: AGGATGGCGTGAGGGAGAGC; CD11c forward primer: ACAACCCCGTCCCTCTTATCG; CD11d reverse primer: TGCAGAATGCTTCTTTACCATTGG. PCR programs are available on request.

Data analysis

Flow cytometry data for routine analysis of thymocyte, thymic stromal cell populations, and lymph node cells were analyzed with FlowJo software (FlowJo LLC). Statistical analysis of FlowJo data was performed with GraphPad Prism 9.0 (GraphPad Software Inc., San Diego, California, USA). Column and group comparisons were performed by employing unpaired, 2-tailed Welch’s t test. P values > 0.05 were considered nonsignificant. Data were all represented as mean ± SEM. Statistical results are provided in the figure legends and in Table S1.

For detailed analysis of post-selection thymocytes, we devised a novel probabilistic gating strategy to isolate the CD4^loCD8^lo (DP^lo) population (Fig. S1). We first conducted soft clustering based on Gaussian mixture models (GMM,³⁴) to effectively identify relevant cell populations. Specifically, we applied a 15-cluster GMM (K = 15 found using Bayesian information criterion) to each mouse’s data (in 4 dimensions—CD4, CD8-α, CD3, and CD5) to identify fifteen overlapping subpopulations of cells. We picked the estimated clusters of cells that were either DN or DP and probabilistically gated and removed these cells, retaining the rest. We did the probabilistic removal by randomly drawing a membership from the posterior membership probability according to the Gaussian mixture model. We applied soft clustering to identify cells that are CD3⁺CD5⁺ DP^lo (ie thymocytes that had encountered antigen or post-selection thymocytes) based on probabilistic membership in the corresponding Gaussian clusters. Cells having simultaneous membership in both the CD3⁺CD5⁺ and the DP^lo clusters were classified as CD3⁺CD5⁺DP^lo post-selection thymocytes.

To assess the extent of apoptosis we applied a sliding elliptical window (with a total of 400 overlapping sliding windows) across the CD4 versus CD8-α plot (where DNs are furthest to the left and DPs are furthest to the right) and calculated the proportion of live (Annexin V^-PI^-) cells in each window (Fig. S2).

We next conducted for each cell a thorough analysis of the relationship between cell viability (with Annexin V^-PI^- cells being scored as live) and each of 2 variables¹ the CD8-β/α ratio (which is the ratio of the mean fluorescence intensity (MFI) of CD8-β to the MFI of CD8-α and² the PD-1 level. We used sliding windows or bins in the two variables (CD8-β/α ratio and PD-1 level) and calculated the viability of the cells in each bin or window. These approaches are described in the legends for the figures in which this strategy was employed. Representative CD4 versus CD8-α plots of thymocytes having low, intermediate, and high CD8-β/α ratios are shown in Fig. S3. Next, we translated the CD8-β/α ratios or PD-1 fluorescence values to a bin number as described and illustrated in Figs. S4 and S5, respectively. To define a single point in CD8-β/α ratio at which a sudden drop of more than 20% occurs (which we refer to as a viability “inflection point”), we employed a statistical technique called the 1d fused lasso.³⁵ Further details are provided in Figs. S4 and S5. The negative values for cells with little or no fluorescence in those 2 figures are due to the combined effects of baseline subtraction during fluorescence detection off the PMT and to aggregation of measurement errors from all the compensation channels.³⁶

Results

The fate of BM3 TCR-expressing thymocytes in the presence of thymic H-2K^b

As reported previously,²⁴ in BM3.3 mice, which lack H-2K^b, thymocytes develop normally. We investigated the impact of the expression of the BM3 TCR ligand, H-2K^b, on the developmental fates of BM3 TCR-bearing thymocytes. We used double transgenic mouse strains—KALxBM3 and KQxBM3 and their rag-deficient counterparts (KBBR and KQBR, respectively)—expressing H-2K^b driven by 2 different promoters.

In KAL, KALxBM3, and KBBR mice, H-2K^b expression is driven by the guinea pig α-lactalbumin promoter (Table 1). In the thymus, this promoter is deemed to drive H-2K^b expression chiefly in medullary epithelial cells based on the following. First, previous studies showed that radioresistant, but not bone marrow-derived cells, express the H-2K^b transgene.²⁴ Second, flow cytometric (Fig. S6A, B) and immunohistochemical (Fig. S6C) studies demonstrate thymic medullary epithelial expression of H-2K^b in KALxBM3 mice. There is little or no expression of H-2K^b expression in cortical epithelial cells (Fig. S6A, B).

In KQ, KQxBM3, and KQBR (rag-deficient KQxBM3) mice, H-2K^b expression is driven by the human keratin-14 (Ker14) promoter (Table 1), which has been shown to be active in thymic epithelial cells.³⁷ PCR analysis of thymic cDNAs (Fig. S6D) demonstrates H-2K^b expression in KQxBM3 thymus. Flow cytometric analysis of thymic preparations enriched for thymic epithelial cells (Fig. S7) indicates that a very small proportion of KQxBM3 medullary epithelial cells express H-2K^b. However, the expression of H-2K^b by these cells is far less than the level expressed by KALxBM3 medullary epithelial cells (Fig. S7).

BM3.3 T cells express both the BM3 TCR and a TCR composed of the BM3 TCR-β chain paired with an endogenously rearranged TCR-α chain.³⁸ To determine if the BM3 TCR alone is responsible for the observed effects of H-2K^b on thymocyte development, we used both rag-sufficient and rag-deficient strains (Table 1) in our studies. Given the high affinity of the BM3 TCR for H-2K^b, one might expect pronounced thymocyte death and thus reduced cellularity in both KQxBM3 and KALxBM3 thymi. Indeed, the cellularity of KALxBM3 thymi is ∼50% that of BM3.3 thymi (Fig. 1A). KQxBM3 thymi have slightly lower cellularity than BM3.3 thymi, but the differences are not statistically significant (Fig. 1A). Regardless of differences in cellularity, the proportions of DP thymocytes in BM3.3, KALxBM3, KQxBM3, BM3rag^−/−, KBBR, and KQBR mice are largely similar (Fig. 1B, C).

Figure 1.

Thymocytes in KQxBM3 and KALxBM3 mice have different fates. (A) Total cell number in the indicated mouse strains as determined by counting on a hemacytometer. Data are presented as mean ± SEM. The total number of mice and sexes for each strain and the number of experiments were: BM3.3 (29 mice; 8 males and 21 females in 17 experiments), KALxBM3 (64 mice; 22 males and 42 females in 27 experiments), KQxBM3 (28 mice; 10 males, 16 females and two mixtures each consisting of 1 male and 2 females in 19 experiments), BM3rag^−/− (36 mice; 20 males and 16 females in 20 experiments), KBBR (35 mice; 17 males and 18 females in 14 experiments), KQBR (6 mice; 4 males and 2 females in 2 experiments), KAL (12 mice total; 2 males and 10 females in 7 experiments), and KQ (10 mice; 5 males, 3 females, and two of unidentified sex; 8 experiments). (B, C) Flow cytometric analysis of thymocytes was performed using the gating strategy shown above the FACS plots. Panel (B) shows data for thymocytes from the rag-sufficient parental strains, KAL, KQ, and BM3.3 and double-transgenic strains, KQxBM3 and KALxBM3. Panel C shows data for the rag-deficient parental BM3rag^−/− and double-transgenic KQBR and KBBR strains. In some experiments, the antibody-cytofluor conjugates used for these experiments and their specificities were: 53-6.7-Pacific Blue (CD8-α), GK1.5-BV605 (CD4), 17A2-AF488 (CD3), 53-7.3-PE/Dazzle (CD5), and Mel-14-APC (CD62L). In other experiments, the antibody-cytofluor conjugates used for these experiments and their specificities were: 53-6.7-Pacific Blue (CD8-α), GK1.5-BV605 (CD4), 53-5.8-PE/Cy7 (CD8-β), 17A2-FITC (CD3), 53-7.3-PE/Dazzle (CD5), 30-F11-PE/Cy5 (CD45), and 29F.1A12-PE (PD-1). The percentages of the thymocyte subpopulations are shown in the bar graphs on the left. Percentages of thymocyte subpopulations are presented as mean ± SEM. Quantification of data are provided on the left of each panel. Representative CD4 vs. CD8-α plots are provided on the right side of each panel. Gates for CD4^-CD8^- (DN), CD4⁺CD8⁺ (DP), CD4⁺ SP, and CD8⁺ SP thymocytes are indicated. The total number of mice and sexes for each strain and the number of experiments were: BM3 (15 mice; 3 males and 12 females in 8 experiments), BM3rag^−/− (15 mice; 9 males and 6 females in 8 experiments), KALxBM3 (30 mice; 11 males and 19 females in 11 experiments), KBBR (13 mice; 5 males and 8 females in 7 experiments), KQxBM3 (15 mice; 3 males and 12 females in 6 experiments), KQBR (5 mice; 3 males and 2 females in 2 experiments), KAL (10 mice; 4 males and 6 females in 7 experiments), and KQ (4 mice; 2 males and 2 mice sex not recorded in 3 experiments). Asterisk indicates statistical significance using the unpaired 2-tailed t test with Welch’s correction. *P<0.05, **P<0.01, ****P<0.0001. Note that there are no bars for cell population designations that had no cells (eg, CD4 SPs in KALxBM3 and KBBR).

KQxBM3 and KQBR mice and their parental BM3, BM3rag^−/−, and KQ strains have similar proportions of DN thymocytes (Fig. 1B, C). In striking contrast, KALxBM3 and KBBR mice have an elevated proportion of DN thymocytes as compared to the parental BM3, BM3rag^−/−, and KAL parental strains (Fig. 1B, C). Apoptosis of KQxBM3 DN thymocytes is exceedingly low and is far lower than that of KALxBM3 DN thymocytes (Fig. S8). Thus, the smaller proportion of DN thymocytes in KQxBM3 mice as compared to KALxBM3 mice is not due to premature thymocyte deletion in the KQxBM3 thymus. Taken together, the data indicate that H-2K^b does not induce premature thymocyte death and that negative selection occurs after the DP stage in KALxBM3 mice.

KALxBM3 thymi have no CD8⁺ SP thymocytes, while KQxBM3 mice have ∼60% as many CD8⁺ SP thymocytes as do BM3.3 mice (Fig. 1B). KALxBM3 mice also have fewer CD4⁺ SP thymocytes than BM3.3 mice, as observed previously by others,²⁴ while KQxBM3 mice have the same proportion (Fig. 1B). Flow cytometry experiments with the clonotypic Ti98 antibody, which is specific for the BM3 TCR,^37,38 yielded similar results (Fig. S9). Since thymocyte populations of KQxBM3 mice are largely comparable to those of BM3.3 mice, we henceforth regard the KQxBM3 strain as another control for KALxBM3.

As expected, BM3rag^−/−, KQBR, and KBBR thymi lack CD4⁺ SP thymocytes (Fig. 1C). BM3rag^−/− and KQBR thymi have CD8⁺ SP thymocytes, while KBBR thymi do not (Fig. 1C). Thus, the BM3 TCR alone (and not a TCR composed of the BM3 TCR-β chain paired with an endogenous TCR-α chain) causes CD8⁺ SP deletion in KALxBM3 mice.

The DP^lo thymocyte stage—the point of great reckoning

The proportion of CD3^hiCD5^hiDP^lo thymocytes is significantly greater in KALxBM3 and KBBR mice than it is in BM3.3, KQxBM3, BM3rag^−/−, and KQBR mice (Fig. 2A, B). For example, KALxBM3 mice have 12–15 times more CD3^hiCD5^hi DP^lo thymocytes than BM3.3 mice have (Fig. 2C). We are particularly interested in this population because other researchers have shown that these CD3^hiCD5^hiDP^lo thymocytes (referred to as post-selection thymocytes) are subject to negative selection and give rise to IEL precursors.^4,5

Figure 2.

Elevated numbers of CD3^hiCD5^hi DP^lo and DN thymocyte populations in KALxBM3 and KBBR mice. P (A) Thymocytes from the indicated mouse strains were stained for the cell-surface markers specified in the gating strategy. The antibody-cytofluor conjugates used for these experiments and their specificities are: 53-6.7-Pacific Blue (CD8-α), GK1.5-BV605 (CD4), 53-5.8-PE/Cy7 (CD8-β), 17A2-FITC (CD3), 53-7.3-PE/Dazzle (CD5), 30-F11-PE/Cy5 (CD45), and 29F.1A12-PE (PD-1). P(B) Percentages of CD8, CD4, DP, DP^lo, and DN thymocytes that are CD3^hiCD5^hi in the indicated single and double transgenic strains. The total number of mice and sexes for each strain and the number of experiments were: BM3.3 (6 mice; 3 males and 3 females in 2 experiments), KALxBM3 (17 mice; 6 males and 11 females in 6 experiments), KQxBM3 (11 mice; 3 males and 8 females in 4 experiments), BM3rag^−/− (13 mice; 7 males and 6 females in 7 experiments), KBBR (11 mice; 5 males and 6 females in 4 experiments), and KQBR (4 mice; 3 males and 1 female in 2 experiments). Asterisk indicates statistical significance using the unpaired 2-tailed t test with Welch’s correction. *P<0.04, **P=0.0011, ***P<0.001, ****P<0.0001. (C) KALxBM3 mice have a greater proportion of CD3^hiCD5^hi DP^lo and DN thymocytes than do BM3.3 and KQxBM3 mice. Shown is the ratio of the percentage of CD3^hiCD5^hiDP^lo thymocytes in the indicated KALxBM3 or KQxBM3 population (CD4⁺, CD8⁺, DP, DP^lo, or DN) relative to the percentage of CD3^hiCD5^hiDP^lo thymocytes in that same population in BM3.3 mice. Statistical analysis for panels (B) and (C) is provided in Table S1.

Because BM3.3 mice do not express H-2K^b, we expected negative selection to occur at low levels in them. Indeed, the level of apoptosis is very low in BM3.3 Ti98⁺ thymocytes (Fig. S10). Conversely, as indicated above, we expected the post-selection thymocyte population in KALxBM3 and KBBR mice to be subject to strong negative selection. To identify CD3⁺CD5⁺ thymocytes that are undergoing apoptosis, we employed the flow cytometry data analysis pipeline described in Materials and Methods in Fig. S1. This analysis revealed apoptotic cells both in CD3⁺CD5⁺DP and in CD3⁺CD5⁺DP^lo thymocyte populations in KALxBM3 and KBBR mice (Fig. S2). Representative Annexin V versus PI plots for KALxBM3 and KBBR DN, DP, and DP^lo thymocyte populations are shown in Fig. S11. These data show that apoptosis is greatest in DP^lo thymocytes and least in DN thymocytes (Fig. S11).

We next examined the levels of PD-1 and CD8-β in the CD3⁺CD5⁺DP^lo (ie post-selection) thymocytes. In the mouse, CD8-β occurs at the cell surface only in dimeric association with CD8-α.³⁹ Therefore, we can use the CD8-β/CD8-α ratio (henceforth referred to as the CD8-β/α ratio) as a surrogate for the relative proportions of CD8-αβ and CD8-αα on the cell surface. A comparison of the CD4 versus CD8-α plots of post-selection DP^lo thymocytes having different CD8-β/α ratios reveals that a high CD8-β/α ratio is not restricted to DP^lo thymocytes (Fig. S3). We created histograms of the CD8-β/α ratio and of the PD-1 expression level for all post-selection DP^lo thymocytes (Fig. 3A′, B′). We determined the proportion of live post-selection DP^lo thymocytes across the 2 histograms using 200 overlapping increments as described in Fig. 3. This sliding window approach dramatically improved our ability to observe the relationship between cell survival and (1) the CD8-β/α ratio and (2) the PD-1 level. These effects are obscured if one coarsely gates the cells into a smaller number of distinct regions across the entire range of CD8-β/α ratios or PD-1 levels, as is commonly done with FlowJo.

Figure 3.

A sliding window strategy reveals correlations of the CD8-β/CD8-α ratio and PD-1 with the survival of KALxBM3 and KBBR post-selection thymocytes. (A, B) Histogram analysis of CD8-β/CD8-α and PD-1 levels, respectively. The histograms in A′ and B′ are overlays of CD8-β/CD8-α ratio (A1′) and of PD-1 level (B′) for live CD3⁺CD5⁺DP^lo thymocytes (blue histogram) and for all CD3⁺CD5⁺DP^lo thymocytes (gray histogram). The gating strategy leading up to the histograms is provided above the plots. A sliding window with a total of 200 overlapping increments was moved across each of the histograms. The antibody-cytofluor conjugates used for these experiments and their specificities are: 53-6.7-Pacific Blue (CD8α), GK1.5-BV605 (CD4), 53-5.8-PE/Cy7 (CD8β), 17A2-FITC (CD3), 53-7.3-PE/Dazzle (CD5), 30-F11-PE/Cy5 (CD45), and 29F.1A12-PE (PD-1). Data are from flow cytometry experiments using the reagents and gating strategies described in the Materials and Methods and in Figure S2. (C, D) Proportion of live cells across the plots of the CD8β/CD8α ratio (panel C) and of the PD-1 level (panel D) for all CD3⁺CD5⁺DP^lo thymocytes from 6 KALxBM3 and 6 KBBR mice. The line width is not uniform across the plot and reflects the number of cells (enumerated to the right of the graphs and below the strain names). The 200 increments were overlapping, resulting in a smooth curve.

This analysis revealed an apparent correlation between the CD8-β/α ratio and post-selection DP^lo thymocyte survival in KALxBM3 and even more so in KBBR mice. Specifically, viability is lowest for DP^lo thymocytes having a very high CD8-β/α ratio (Fig. 3C). For PD-1, the situation is more complicated. Thymocyte survival is observed in post-selection DP^lo thymocytes having either a very low or a very high PD-1 level. However, as indicated by the line thicknesses in Fig. 3C–D, very few KALxBM3 thymocytes have extremely high CD8-β/α ratios or very high levels of PD-1.

While the results in Fig. 3 are interesting, they do not inform us of the relative contributions of the CD8-β/α ratio and of PD-1 to the survival of post-selection DP^lo thymocytes. To learn more about their relative importance, we performed a more detailed analysis (Fig. S12A, B). We interrogated 336 regions (grid squares in left panel of Fig. S12B) of the CD-8β/α versus PD-1 plots of post-selection DP^lo thymocytes for live cells (right panel of Fig. S12). We then grouped these 336 populations into fourteen cohorts based on the level of PD-1 expression. For each cohort of PD-1 level, we plotted the percentage of live cells as a function of the CD8-β/α ratio (Fig. 4A). For ease of visualization, we are showing data for seven of the 14 PD-1 expression groups (Fig. 4A). These studies led us to 3 major conclusions. First, the association of the CD8-β/α ratio with the survival of post-selection thymocytes is far more striking than is the PD-1 level. Second, in agreement with the data in Fig. 3C, thymocytes having the highest CD8-β/α ratio also have the highest proportion of dead cells. We present a possible explanation for these 2 observations later in this paper. Finally, in both KALxBM3 and KBBR mice, thymocyte viability remains high across a range of CD8-β/α ratios until a specific inflection point, at which it drops markedly (Fig. 4A).

Figure 4.

The ratio of CD8-β to CD8-α appears to be more important than PD-1 for the survival of KALxBM3 and KBBR post-selection thymocytes. (A) Viability (ie, % live cells) as a function of the CD8-β/α ratio for CD3⁺CD5⁺DP^lo thymocytes from 6 KALxBM3 and 6 KBBR mice faceted by the PD-1 level. For each strain (KALxBM3 on the left and KBBR on the right), the percentage of live cells in each of the fourteen PD-1 groups from Figure S12B was interrogated as a function of the CD8-β/α ratio. The data were faceted by PD-1 level, and the cell groups or bins (groups 1–14) were determined as shown in Fig. S12A. For ease of visualization, only a subset of the fourteen groups is shown (groups 2, 3, 5, 7 9, 11, and 13). The vertical blue dashed lines in each plot indicate the viability inflection points (defined as the CD8-β/α ratio at which the viability rapidly dropped by 20% or more for each mouse). The solid red lines indicate the average viability to the left and right of the inflection point that is marked by the vertical dashed blue line. (Fainter pink lines were used instead of red to mark the averages in mice whose sudden drop was below 20%.) The mice are the same ones used in Figure 3. (B) Viability inflection point plotted versus the PD-1 level for 6 KALxBM3 mice (top plot) and for 6 KBBR mice (lower plot). Each gray line is a spline regression curve of the inflection points regressed on PD-1 levels for each mouse. The thick blue line is the average of the viability inflection points for all the individual mice in that cohort at each PD-1 level. The mice are the same ones used in Figure 3.

We were struck by the viability inflection point and thus examined it more closely (Fig. 4B). In both KALxBM3 and KBBR thymocytes, the viability inflection point is higher when the PD-1 level is higher. This is evident in the 2 graphs shown in Fig. 4B in which the viability inflection point (as a function of the CD8-β/α ratio) is plotted versus the PD-1 level. Interestingly, the effect of PD-1 is not the same in the 2 strains. Apoptosis is observed in KBBR post-selection DP^lo thymocytes having low PD-1 levels but not KALxBM3 post-selection DP^lo thymocytes having low PD-1 levels (Fig. 4B). Thus, the expression of 2 TCRs in KALxBM3³⁹ may mitigate apoptosis of post-selection DP^lo thymocytes, despite their low PD-1 levels (Fig. 4A).

KALxBM3 and KBBR mice have post-selection DN thymocytes

As noted previously, KALxBM3 and KBBR mice have an unusually high proportion of DN thymocytes. Importantly, most of these have are CD3^hiCD5^hi. KALxBM3 thymi have ∼7 times more CD3^hiCD5^hi DNs than do BM3 and KQxBM3 thymi (Fig. 2B). KBBR thymi have ∼3 times more CD3^hiCD5^hi DNs than do BM3rag^−/− thymi and ∼10 times more than do KQBR thymi (Fig. 2C). Furthermore, the CD3⁺CD5⁺ DNs share with CD8 SP thymocytes (such as those in BM3rag^−/− and KQxBM3 mice) 2 features that distinguish them from antigen inexperienced DN thymocytes. First, most new DN thymic immigrants are CD62L^hi (as illustrated by BM3rag^−/− DNs in Fig. 5A). Conversely, virtually all BM3rag^−/− CD8 SP thymocytes and up to 30% of KALxBM3 and KBBR DN thymocytes are CD62L^lo and express higher levels of CD5 (Fig. 5A, B). A far smaller proportion of KQxBM3 and BM3rag^−/− DNs have this phenotype (Fig. 5A, B). Second, the CD3 MFI’s of BM3 rag^−/− CD8⁺ SP thymocytes (which are obviously post-selection) and of KALxBM3 CD5⁺CD62L^lo DNs are equivalent (Fig. 5C). In contrast, the CD3 MFI of KALxBM3 CD5⁺CD62L^hi DN thymocytes is only slightly over half of the CD3 MFI for BM3 rag^−/− CD8 SP thymocytes (Fig. 5C). These results are consistent with the idea that post-selection CD3⁺CD5⁺DP^lo thymocytes in KALxBM3 and KBBR mice give rise to CD3⁺CD5⁺CD62L^lo DNs. We thus shall henceforth refer to these CD3⁺CD5⁺CD62L^lo cells as post-selection DN thymocytes.

Figure 5.

Identification of post-selection DN thymocytes in KALxBM3 and KBBR mice. The gating strategy for distinguishing phenotypically immature DN and post-selection DN thymocytes is shown above the FACS plots depicted in panel A. (A) CD62L versus CD5 plots for the following cell populations: CD3⁺ DN thymocytes from KQxBM3, KALxBM3, and KBBR mice (plots in the top row), and as controls, CD3⁺ DN and CD3⁺CD8⁺ SP thymocytes from BM3rag^−/− mice (plots in bottom row). (B) The percentages of post-selection (ie, CD5⁺CD62L^lo) DN thymocytes quantified for several BM3rag^−/−, KQxBM3, KALxBM3, and KBBR mice. Data are presented as mean ± SEM. The total number of mice and sexes for each strain and the number of experiments were: BM3rag^−/− (13 mice; 7 males and 6 females in 7 experiments), KQxBM3 (12 mice; 4 males and 8 females in 4 experiments), KALxBM3 (21 mice; 9 males and 12 females in 8 experiments), and KBBR (10 mice; 5 males and 5 females in 4 experiments). Asterisk indicates statistical significance using the 2-tailed t test with Welch’s c. ***P=0.0004. (C) KALxBM3 post-selection DN thymocytes express higher levels of CD3 than do CD62L^hi DN thymocytes. Data are presented as the ratio of the CD3 MFI for the indicated population of post-selection DNs relative to the CD3 MFI for BM3rag^−/− CD8⁺ SP thymocytes. The results are the averages for 11 KALxBM3 mice (3 males and 8 females) from 4 experiments. Asterisks indicate statistical significance using the two-tailed t test with Welch’s c. **P=0.0042.

We next focused on the fates of the post-selection DN thymocytes. Thymocytes that migrate toward the thymic medulla express CCR7,⁴⁰ while DN Type A IEL precursors do not.⁵ To determine if all KALxBM3 and KBBR post-selection DNs are IEL precursors, we interrogated them for expression of CCR7 (Fig. 6A, B). In brief, ∼70% of post-selection KALxBM3 DNs and post-selection KBBR DNs express CCR7 (Fig. 6B). These proportions are comparable to the proportions of BM3.3 CD8⁺ SP thymocytes and KALxBM3 CD4⁺ SP thymocytes (Fig. 6A, B). Absolute numbers of CCR7⁺ post-selection DN thymocytes in KALxBM3 and KBBR mice also are comparable to the number of CD5⁺CD62L^loCCR7⁺ CD8⁺ T cells in BM3.3 mice (Fig. 6C). While ∼90% of the post-selection KALxBM3 and KBBR DN thymocytes express PD-1 (Fig. 6D), only slightly more than 40% of these PD-1⁺ cells are CCR7⁺ (Fig. 6D). These results indicate that PD-1 expression is not unique to type A IELps but rather is a feature of thymocytes that escape negative selection. This is consistent with its role in attenuating TCR signal transduction. Its expression might also predispose surviving post-selection thymocytes for a role in enforcing self-tolerance.⁴¹

Figure 6.

A significant proportion of KALxBM3 and KBBR PD-1⁺ post-selection DN thymocytes express CCR7 at levels comparable to those of BM3.3 CD8⁺ thymocytes. (A) Relative fluorescence intensity of CCR7 for each cell in post-selection CD3⁺ DN thymocytes. For comparison the CCR7 profiles of BM3.3 CD3⁺ CD8⁺ SP thymocytes and KALxBM3 CD3⁺ CD4⁺ SP thymocytes are shown (plots 1 and 2, respectively). The boxed area indicates 99% of CCR7 expression in the BM3.3 CD8⁺CD5^hiCD62L^lo thymocytes. The gating strategy is shown above the plots. (B) Percentage of CCR7⁺ cells in the indicated thymocyte populations: BM3.3 CD3⁺ CD8⁺SP, KALxBM3 post-selection CD3⁺ DN, KALxBM3 CD3⁺ CD62L^hi DN, KBBR post-selection DN, and KBBR CD3⁺ CD62L^hi DN. Post-selection DN data are in green bars; CD62L^hi DN are in red bars. Data are presented as mean +/- SEM. The total number of mice and sexes for each strain and the number of experiments were: BM3.3 (3 mice, all female in 2 experiments), KALxBM3 (11 mice; 4 males and 7 females in 5 experiments), and KBBR (7 mice; 4 males and 3 females in 4 experiments). Asterisk indicates significant difference of the indicated samples using the unpaired 2-tailed t test with Welch’s correction. *P<0.05. Complete statistical analysis is provided in Table S1. (C) The number of CCR7⁺ post-selection DN thymocytes in the indicated cell populations and mouse strains. The total number of mice and sexes for each strain and number of experiments are as in panel B. Data are presented as mean ± SEM. (D) The percentage of post-selection CD3⁺ CCR7⁺ DN thymocytes that are PD-1⁺ (two bars on left half of graph) and the percentage of post-selection CD3⁺ PD-1⁺ DN thymocytes that are CCR7⁺ (2 bars on right half of graph) in KALxBM3 and KBBR mice. The total number of mice and sexes for each strain are as in panel B. Data are presented as mean ± SEM.

We have referred to type A IELps, but why do we suspect that the PD-1⁺CCR7^- DNs are type A IELps? We have 3 reasons. First, at least some thymocytes undergo negative selection at the corticomedullary junction (CMJ) by interacting with CMJ mTECs expressing essential costimulatory ligands.⁴² Hence, at the CMJ, the interaction of H-2K^b-expressing mTECs with cortical post-selection KALxBM3 and KBBR thymocytes could promote their development into type A IELps.⁵ Second, like type A IELps (and unlike type B IELps), the PD-1⁺CCR7^-DN thymocytes in KALxBM3 and KBBR mice are specific for a classical MHC I molecule.⁵ Finally, and importantly, the thymi used in these studies were from young (6–8 wk of age) mice. At this age, there are almost 10 times more type A IELps than type B IELps.⁵ Collectively, these considerations make it highly likely that the PD-1⁺CCR7^-DN post-selection thymocytes are type A IELps.

A significant proportion of KALxBM3 and KBBR lymph node DN T cells have a memory phenotype and express PD-1

DN T cells have been found in other TCR transgenic mice that express the transgenic TCR’s cognate antigen.^43,44 Furthermore, these DN T cells have an activated phenotype.^43,44 We thus examined the phenotypes of lymph node cells from BM3, BM3rag^−/−, KQxBM3, KQBR, KALxBM3, and KBBR mice. While lymph node cells in KALxBM3 and KBBR mice are not numerous (Fig. 7A), they do include CD3⁺ and Ti98⁺ cells (Fig. 7B). DNs make up ∼25% of CD3⁺ cells and ∼35% of Ti98⁺ cells in KALxBM3 lymph nodes, with the remainder being CD4⁺ (Fig. 7C–F). In KBBR lymph nodes, virtually all CD3⁺ cells (Fig. 7C and E) and Ti98⁺ (Fig. 7D and F) are DN. In contrast, BM3, BM3rag^−/−, KQxBM3, and KQBR lymph nodes have virtually no CD3⁺ DN cells (Fig. 7C and E).

Figure 7.

KALxBM3 and KBBR lymph node DN T cells express the BM3 TCR. (A) Total number of lymph node cells. Data are presented as mean ± SEM. The total number of mice and sexes for each strain and the number of experiments were: BM3.3 (7 mice; 2 males and 5 females; 3 experiments), BM3rag^−/− (21 samples; 13 males, 6 females, 1 pool of two males, and 1 pool of 1 male and 1 female; 9 experiments), KQxBM3 (13 mice; 5 males and 8 females; 4 experiments), KQBR (3 mice; 1 male and 2 females; 1 experiment), KALxBM3 (28 samples; 12 males, 15 females, and 1 pool of 2 males and 2 females; 11 experiments), and KBBR (11 samples; 4 males, 5 females, 1 pool of 2 males, and 1 pool of 2 females; 7 experiments). (B) Total number of CD3⁺ or Ti98⁺ lymph node cells in the various strains. Data are presented as mean ±SEM. The total number of mice and sexes for each strain and the number of experiments were: BM3.3 (18 mice total; 9 males and 9 females in 9 experiments), BM3rag^−/− (10 mice total; 5 males and 5 females in 10 experiments), KQxBM3 (23 mice total, 7 males and 16 females in 8 experiments), KQBR (7 mice total; 5 males and 2 females in 2 experiments), KALxBM3 (35 mice total; 15 males and 20 females in 14 experiments), and KBBR (12 mice total; 7 males and 5 females in 7 experiments). (C) CD4 vs CD8-α plots of BM3.3, KQxBM3, and KALxBM3 CD3⁺ lymph node cells. The gating strategy is shown above the plots. (D) CD4 vs CD8-α plots of BM3.3, Continued.KQxBM3, and KALxBM3 Ti98⁺lymph node cells. The gating strategy is shown above the plots. (E) Stack graphs for several mice of the relative proportions of each CD3⁺ cell type shown in panel C. Data are presented as mean ± SEM. The total number of mice and sexes for each strain and the number of experiments were: BM3.3 (10 mice; 4 males and 6 females in 4 experiments), BM3rag^−/− (9 samples total, as follows: 6 males, 1 female, 1 pool of 2 males, and 1 pool of 1 male plus 1 female; 5 experiments), KQxBM3 (12 mice total; 4 males and 8 females in 4 experiments), KQBR (6 mice total; 4 males and 2 females in 2 experiments), KALxBM3 (24 samples total; 10 males and 13 females and 1 pool of 2 females plus 2 males; 9 experiments), and KBBR (8 samples total; 4 males, 3 females and 1 pool of 2 males; 4 experiments). (F) Stack graphs for several mice of the relative proportions of each Ti98⁺ cell type shown in panel D. Data are presented as mean ± SEM. The total number of mice and sexes for each strain and the number of experiments were: BM3.3 (7 mice total; 4 males and 3 females in 3 experiments), KALxBM3 (11 mice total; 3 males and 8 females in 3 experiments), KBBR (2 samples total; 1 male and 1 pool of 2 females in 1 experiment) and KQxBM3 (10 mice total; 3 males and 7 females in 3 experiments). Complete statistical analysis for panels A, B, E, and F is provided in Table S1.

Both the Ker14 promoter and the guinea pig α-lactalbumin promoter are active in epithelial cells; thus, peripheral T cells might encounter H-2K^b. Accordingly, we interrogated CD3⁺ lymph node cells for expression of CD44, a marker of memory T cells. All—or almost all—CD3⁺CD8⁺ cells in BM3rag^−/− and KQxBM3 mice are CD44^- and thus are regarded as being naive (Fig. 8A). The paucity of memory T cells in KQxBM3 lymph nodes is not surprising because H-2K^b (driven by the Ker14 promoter) is weakly expressed in resting KQxBM3 skin (Fig. S13). Conversely, of the CD3⁺ DN cells in KALxBM3 and KBBR lymph nodes, ∼30% are of the CD44⁺ phenotype (Fig. 8A). This finding is consistent with the high level of H-2K^b expression in the skin driven by the α-lactalbumin promoter (Fig. S13). Adoptive transfer experiments are necessary to definitively prove that these DN T cells are derived from the post-selection DN thymocytes. Unfortunately, the lack of suitable genetic markers in this model makes such experiments impossible. However, the occurrence of CD44⁺ DN T cells in KALxBM3 and KBBR mice is akin to findings in other TCR transgenic models in which the TCR’s cognate antigen is expressed.^43,44

Figure 8.

A significant fraction of CD3⁺ DN T cells in KALxBM3 and KBBR lymph nodes have a memory phenotype and are PD-1⁺. (A) CD44 vs CD62L plots of BM3rag^−/− CD3⁺CD8⁺ T cells, KQxBM3 CD3⁺CD8⁺ T cells, KALxBM3 CD3⁺ DN T cells, and KBBR CD3⁺ DN T cells. The gating strategy for interrogating lymph node cells for the memory phenotype is shown above representative FACS plots. CM = central memory; EM = effector memory. The relative proportions of naive, CM, and EM T cells for each of the indicated mouse strains and cell populations are plotted in the stack graph on the right half of panel A. The total number of mice and sexes for each strain and the number of experiments were: BM3rag^−/− (10 samples; 6 males, 2 females, and 2 pools of 2 males each; 5 experiments), KQxBM3 (13 mice; 5 males and 8 females in 4 experiments), KALxBM3 (24 samples; 10 males, 13 females, and 1 pool consisting of 2 males and 2 females; 9 experiments), and KBBR (7 samples; 4 males, 2 females, and 1 pool of 2 males; 4 experiments). (B) The frequency of PD-1⁺ cells in naive, CM, and EM populations for several mice for each strain is shown. The total number of mice and sexes for each strain and the number of experiments were: BM3rag^−/− (8 samples total; 5 males, 1 female, 1 pool of 1 male and 1 female, and 1 pool of 2 males; 4 experiments), KQxBM3 (8 mice; 3 males and 5 females; 2 experiments), KALxBM3 (17 samples; 7 males, 9 females, and 1 pool of 2 males and 2 females; 6 experiments), and KBBR (4 samples; 2 males, 1 female, and 1 pool of 2 males; 3 experiments). (C) The percentages of PD-1⁺ cells for BM3rag^−/− CD3⁺CD8⁺ T cells, KAxMB3 CD3⁺CD8⁺ T cells, KALxBM3 CD3⁺ DN cells, and KBBR CD3+ DN cells. The data are for the same mice as in Panel B. Asterisk indicates significant difference of the indicated samples using the unpaired 2-tailed t test with Welch’s correction. *P<0.0001. Complete statistical analysis for bar graphs in panels A and C is provided in Table S1.

We wished to determine if H-2K^b expression in skin correlates with PD-1 expression by lymph node T cells. As expected, very few BM3rag^−/− CD8⁺ T cells express PD-1 (Fig. 8B). In brief, ∼30% KQxBM3 CD8⁺ T cells express PD-1 (Fig. 8B), and virtually all of them have an effector memory phenotype (Fig. 8C). In contrast, PD-1 is expressed by a majority of KALxBM3 and KBBR lymph node CD3⁺ DN cells (Fig. 8B). Both CD44⁺ and CD44⁻ CD3⁺ DN cells in KALxBM3 and KBBR mice are PD-1⁺ (Fig. 8C). PD-1 may render the CD3⁺ DN cells anergic to H-2K^b and explain the lack of pathology in KALxBM3 and KBBR skin.

IELs in KQBR, BM3rag^−/−, KALxBM3, and KBBR mice are phenotypically similar

Thymic expression of agonist antigen in αβ-TCR transgenic mice promotes the development of thymic IEL precursors and an increase in the number of αβTCR⁺CD8-αα⁺ IELs in the gut.^{2,  5,  45–49} We therefore examined gut IELs in BM3rag^−/−, KALxBM3, KBBR, and KQBR mice. To exclude contaminating blood-borne T cells from our analysis, we gated on CD103⁺CD3⁺ intestinal cells. We gated on Vβ2⁺ cells to restrict our analysis to cells expressing the transgenic TCR (Fig. 9A). There are more than 100 times as many gut CD103⁺CD3⁺V-β2⁺ IELs in KALxBM3 mice than in BM3rag^−/− mice and at least twenty times more CD103⁺CD3⁺V-β2⁺ IELs than in KQBR mice (Fig. 9A). The difference is even greater for KBBR IELs (Fig. 9A). Notably, H-2K^b agonist doesn’t influence the relative proportions of DN and CD8-αα IELs, which are equivalent in all four strains examined (Fig. 9B). Instead, exposure to H-2K^b seems to promote the development of more IELs.

Figure 9.

KALxBM3 and KBBR mice have elevated numbers of TCR-bearing CD8-αα and DN IELs, a significant proportion of which express PD-1. (A) Representative FACS plots for BM3rag^−/−, KALxBM3, KBBR, and KQBR IELs. Gating strategy for identifying CD8αα and DN IELs expressing the transgenic BM3 TCR Vβ2 variable region is above the plots. The total number of CD3⁺CD103⁺ IELs per intestine is indicated in the graph to the right of the FACS plots. IEL samples were pooled to ensure that enough cells were run per sample. The total number of samples was 19, and the total number of mice was 67, with sample pools containing both sexes. There were 5 BM3rag^−/− pools, which collectively contained 6 males and 10 females, for a total of 16 mice. The 6 KALxBM3 pools collectively contained 7 males and 13 females, for a total of 20 mice. The 6 KBBR pools collectively contained 13 males and 7 females, for a total of 20 mice. The 2 KQBR pools collectively contained 7 males and 1 female, for a total of 8 mice. Data are from a total of 10 experiments. (B) Stack plot showing the percentages of DN and CD8αα IELs in BM3rag^−/−, KALxBM3, KBBR, and KQBR mice. Total numbers of mice and experiments are as in panel A. (C) The PD-1 expression profiles for BM3rag^−/− and KALxBM3 IELs (left) and for KQBR and KBBR IELs (right) are displayed as relative fluorescence intensity of PD-1 for each cell. Gating strategy is shown above. The plots on the left and on the right are from two separate experiments. (D) Percentages of PD-1⁺ IELs for BM3rag^−/−, KQBR, KALxBM3, and KBBR mice. Total numbers of mice and experiments are as in panel A. Complete statistical analysis for bar graphs in panels A and B is provided in Table S1.

IELs may play an important role in maintaining immune homeostasis in mucosal epithelial tissues, such as the gut.⁵⁰ For example, IELs may interact with intestinal epithelial cells (IECs), some of which express both MHC I and PD-L1.⁵¹ PD-1 binding to its molecular ligands attenuates signaling through the TCR.^20,21 Hence, expression of PD-1 by IELs might suppress activation of these cells and promote immune tolerance. We thus interrogated the gut IELs in BM3rag^−/−, KQBR, KALxBM3, and KBBR mice for expression of PD-1 (Fig. 9C, D). In all four strains, ∼30–40% of both CD8αα and DN IELs are PD-1⁺ (Fig. 9D), indicating that H-2K^b does not influence PD-1 expression. It will be fascinating to see if these PD-1⁺ IELs play a role in intestinal immune tolerance.

Discussion

The focus of this paper was to determine correlates of survival versus death of H-2K^b-specific TCR-bearing thymocytes in the face of thymic epithelial expression of H-2K^b. Specifically, we wished to understand how the regulation of TCR signal transduction by CD8-β and/or by PD-1 might influence the survival of autoreactive thymocytes. We found that in KALxBM3 and KBBR mice, which express the H-2K^b-specific BM3 TCR and thymic medullary H-2K^b, the viability of post-selection DP^lo thymocytes remains high across a range of CD8-β/α ratios. Then, viability drops precipitously at a specific CD8-β/α ratio, which we refer to as the inflection point (Fig. 4A). These observations are interesting in light of CD8β’s crucial role in initiating TCR-mediated signal transduction. As discussed earlier, CD8β recruits p56^lck and targets CD8-αβ to lipid rafts and thus, to the immunological synapse.^{44,  52} Hence, in cells having low CD8-β/α ratios, the paucity of p56^lck-associated CD8αβ molecules should result in diminished TCR signaling. This may allow thymocyte survival, despite TCR engagement by H-2K^b.

Our results further indicate that PD-1 is an additional determinant of thymocyte viability but not when the CD8-β/α ratio is high. CD48 recruits PD-1 into lipid rafts,⁵² where it is phosphorylated by p56^lck and attenuates TCR signaling.^18–21 Thus, robust levels of PD-1 should promote thymocyte survival. Why then might PD-1 be beneficial only at lower CD8-β/α ratios? Let us consider the possibility that PD-1 molecules are in competition with CD8-αβ heterodimers for space in lipid rafts. In support of this idea is the estimation that no more than 10–30 proteins can occur in a lipid raft.^53,54 Let us further consider the possibility that CD48 is less effective in recruiting PD-1 to lipid rafts than CD8-β is in targeting CD8-αβ to them. If this is the case, then when CD8-αβ is abundant, there may not be enough PD-1 in the immunological synapse to diminish signal transduction through the TCR (Fig. S14).

KBBR post-selection DP^lo thymocytes are more prone to apoptosis than are their KALxBM3 counterparts (Fig. 4A). This result isn’t surprising. Other researchers have reported that thymocytes expressing 2 TCRs, one of which is autoreactive, have reduced sensitivity to the autoantigen.^55–60 We suggest that the increased susceptibility of KBBR post-selection thymocytes to negative selection is due to 4 factors. First, the BM3 TCR α chain need not compete with another TCR-α chain for binding to the BM3 TCR-β chain. Second, during biosynthesis, the BM3 TCR in KBBR thymocytes doesn’t compete with another TCR for association with the CD3 complex. Hence, in KBBR thymocytes, newly synthesized BM3 TCR has unfettered ability to assemble into a functional TCR-CD3 complex and to migrate to the thymocyte surface. Third, once at the cell surface, the BM3 TCR need not compete with another TCR for p56^lck-associated CD8-αβ within the immunological synapse. Finally, the BM3 TCR has a high affinity for its H-2K^b ligand, enabling it to initiate the transduction of signals that result in apoptosis.

Negative selection is one of the immune system’s most intricate solutions to the potential menace of autoreactive T cells. Unfortunately, developing thymocytes have nuanced ways of avoiding deletion, and the wily surviving thymocytes pose a potential threat. Here again, the adroitness of the immune system saves the day. For example, MHC I-specific CD8⁺ thymocytes that survive negative selection can develop into immunosuppressive PD-1⁺CD8⁺ T cells.⁴¹ Similarly, KALxBM3 and KBBR post-selection PD-1⁺CCR7⁺ and PD-1⁺CCR7⁻ DN thymocytes are plausible precursors of PD-1⁺ DN T cells and IELs found in lymph nodes and intestines. We anticipate that future studies will demonstrate that these PD-1⁺ DN T cells and IELs are not renegades but rather that they are enforcers of self-tolerance to H-2K^b on tissue epithelial cells.

Author contributions

Martha C. Zúñiga (Conceptualization [Lead], Data curation [Lead], Formal analysis [Lead], Funding acquisition [Lead], Investigation [Lead], Methodology [Lead], Project administration [Lead], Resources [Lead], Software [Lead], Supervision [Lead], Validation [Lead], Visualization [Lead], Writing—original draft [Lead], Writing—review & editing [Lead]), Sangwon Hyun (Conceptualization [Supporting], Data curation [Supporting], Formal analysis [Equal], Investigation [Supporting], Methodology [Equal], Resources [Supporting], Software [Equal], Supervision [Supporting], Visualization [Equal], Writing—review & editing [Supporting]), Jacob Du (Data curation [Supporting], Formal analysis [Supporting], Writing—review & editing [Supporting]), Shahar Dubiner (Formal analysis [Supporting], Visualization [Supporting], Writing—review & editing [Supporting]), Alicia Freedman-Goretsky (Investigation [Supporting], Validation [Supporting], Visualization [Supporting]), Yitzhar Goretsky (Investigation [Supporting], Validation [Supporting]), Anita Pothukuchi (Data curation [Supporting], Formal analysis [Supporting], Writing—review & editing [Supporting]), Nicholas A. Y. Fong (Data curation [Supporting], Formal analysis [Supporting], Methodology [Supporting]), Alexander Berg (Data curation [Supporting], Validation [Supporting]), Caitlin Davis (Investigation [Equal]), Megumi Barata (Investigation [Supporting]), Tyler-Marie Deveau (Investigation [Supporting]), Alisa Sas (Investigation [Supporting]), Bryan Kim (Investigation [Equal]), Stefan Abreo (Investigation [Supporting]), An Nguyen (Investigation [Supporting]), and Jordan P. Schneider (Investigation [Supporting]), Alanna White(Investigation [Equal])

Supplementary material

Supplementary material is available at ImmunoHorizons online.

Funding

This work was supported by National Institutes of Health Grant R01.AI39055-06.

Conflicts of interest

The authors have no financial conflicts of interest.

Data availability

All data are provided in the figures and tables. Raw data are available from the corresponding author upon request.